Effects of TiO2 on proportion of tissue of alveolar septa in RSV-infected mice on day 5 post-infection.TiO2 exposure (mg/kg)% of alveolar tissuesa% of mean alveolar tissues>40>50>60>7002*20051.8 (3.7)0.25022060.6 (1.6)2.5001168.6 (3.1)aThe proportion of tissue in alveolar septa per unit area by Adobe Photoshop.*Number of mice. Numbers in parenthesis indicate the standard error.Full-size tableTable optionsView in workspaceDownload as CSVTo investigate whether the distribution of RSV-infected GDC0449 was changed qualitatively due to TiO2 (2.5 mg/kg) exposure, sections of the lung tissues of RSV-infected mice were stained immunohistochemically with a goat-polyclonal antibody against RSV protein (Fig. 2B). There was no significant change in the localization of RSV-positive cells, but the TiO2 nanoparticles were not close to RSV-positive cells (Fig. 2B-b and -d). Similar results were observed for TiO2 (0.25 mg/kg)-treated mice (data not shown). However, there was aggregation of TiO2 nanoparticles near inflammatory cells in the severe region (Fig. 2B-d). Because these results suggested that TiO2 nanoparticles might influence the function of macrophage/monocyte in an early phase of RSV infection, bronchoalveolar lavage cells on day 1 post-infection from RSV-infected mice were incubated for 48 h with or without 0.1 mg/mL of TiO2, corresponding to a dose of 0.5 mg/kg in vivo. After incubation, the levels of CCL3 in the culture supernatant of the cells were measured by ELISA ( Table 5). Although the cells were stimulated with LPS, there was no significant change in the production of CCL3 from bronchoalveolar lavage cells due to TiO2 treatment.Table 5.

NOS, nitric oxide synthase; NO, nitric oxide; eNOS, endothelial NOS; p-eNOS, phospho-eNOS; nNOS, neuronal NOS; iNOS, inducible NOS; Akt, Protein Kinase B, phospho-Akt; HIF-1α, hypoxia inducible factor α; Hsp-90, heat shock protein 90; FW, freshwater; 6mAe, 6 months of experimentally induced aestivation; 6mAe6d, 6 days after arousal; PBS, phosphate-buffered saline; MAW, methanol-acetone-water; TBS, Tris-buffered saline; TTBS, TBS plus Tween20; FITC, fluorescein isothiocyanate; EGTA, ethylene glycol tetraacetic acid; EDTA, ethylenediaminetetraacetic acid; PMSF, phenylmethylsulfonyl fluoride; BSA, bovine serum albuminKeywordsNOS (nitric oxide synthase); Gills and lungs remodeling; Aestivation; Environmental stress; Hsp-90; HIF-1α1. IntroductionFor decades, the extraordinary abilities of some fishes to switch from aquatic to aerial respiration and the associated morpho-functional adjustments attracted the interest of physiologists. Among air-breathing fishes, dipnoans (lungfish), an archaic group belonging to the Sarcopterygii class, represent an important evolutive step with respect to water–land transition, due to their ability to survive out of water for relatively long periods, breathing air through true-paired lungs. The extant African lungfishes, which include four species: Protopterus dolloi, Protopterus aethiopicus, Protopterus annectens and Protopterus amphibious, are obligate air breathers, with reduced gills and dependency on pulmonary breathing throughout their life cycle [1] and [2]. They colonize water basins (lakes and rivers) facing low O2 levels, high daytime temperatures and seasonal drought. When water is lacking, African lungfishes enter in a dormancy state (aestivation), dominated by aerial respiration that allows to withstand long periods of water and food deprivation (up to 6 years in the case of P. amphibious) [3]. Laboratory aestivation in Protopterus comprises 3 phases, i.e., induction, maintenance and arousal [4] and [5]. Throughout induction, the lungfish senses the environmental changes, triggering a number of integrated responses, ranging from behavioral to morpho-functional and biochemical levels, all of which facilitate its survival during aestivation. The fish hyperventilates and secretes mucus, which becomes a dry cocoon within 6–8 days [6] (see for review Ref. 7). After the complete cocoon formation and the cessation of feeding and locomotor activities, the maintenance phase begins. It is characterized by reduced metabolism [8] convoyed by down-regulation of respiratory (low O2 consumption and frequency of pulmonary breathing) and cardiovascular (reduced heart rate, cardiac work, and blood pressure and flow) activities [9] and [10]. Upon contact with water, the lungfish immediately awakens from dormancy, entering into the arousal phase, during which it quickly leaves the cocoon and slowly swims to the water surface for air. After arousal, the lungfish has to excrete waste products accumulated during aestivation and begins to eat after about a week, restoring the normal metabolic balance. However, the metabolic advantages of aestivation are attained at considerable costs. With shortage of water during aestivation, the fish must face serious dehydration stress, and its cells and tissues must be confronted with hyperosmotic stress due to the accumulation of urea resulting from the mobilization of the internal protein and amino HA14-1 stores in conjunction with ammonia detoxification to urea. This contributes to minimize waste production and thus pollution of the internal milieu, preventing cell death and tissue degradation in those organs and tissues for which integrity is crucial for survival (see Ref. 11 for a review). Importantly, the metabolic adjustments need to be paralleled and coordinated with a number of morpho-functional rearrangements, including fundamental respiratory and circulatory changes [12] and [13]. During aestivation, a mucus layer protects gills, and the secondary lamellae collapse, remarkably reducing vascular exchange area. While the gills become isolated from the environment, an exclusive aerial ventilation is established with the lungs being indispensable for oxygen uptake [14], as they appear better vascularized and expanded. Furthermore, their inner structure, made up by ridges and air spaces, becomes visible to the naked eye ( Ref. 14 and present study). These changes are accompanied by blood flow redistribution at the gill arches where, unlike the majority of fishes, the branchial circulation is separated into pulmonary vs systemic vascular beds [15]. As a consequence, blood flow is redirected to favor lung perfusion and to allow adequate respiratory exchanges [16].To date, nothing is known regarding the signal transduction mechanisms that underline the cyclic branchial and pulmonary remodeling in aestivating African lungfishes discussed earlier, as the few available studies provided only morphological observations (for review, see [14] and [17]). Therefore, the aim of this study was to identify key switch components of stress-induced signal transduction networks, which could be implicated in both rapid and medium–long term remodeling, in the gills and lungs of P. annectens.Nitric oxide (NO), produced by the different isoforms of nitric oxide synthase (NOS), i.e. NOS1 or nNOS, NOS3 or eNOS and NOS2 or iNOS, is a universal modulator of the animal aerobic biome (redox and energy balance) and is implicated in multiple homeostatic regulations, including instantaneous cardio-circulatory and respiratory processes (for fish, see Ref. 18 and references therein). During hypoxic stress the cross-talk between eNOS and HIF-1α improves cell survival through various mechanisms, such as HIF-1α-dependent activation of several critical genes [19], including NOS [20] and [21]. We previously showed that the NOS/NO signaling is involved in the tissue/organ readjustment of FW and aestivating lungfish (P. dolloi, kidney vs heart [22]; P. annectens, myotomal vs cardiac muscle [23]). Here we evaluated by immunofluorescence microscopy and Western blotting the expression and localization of NOS and its molecular activator Akt, as well as Heat Shock Protein 90 (Hsp-90) and Hypoxia Inducible Factor (HIF-1α), in branchial and pulmonary tissues of P. annectens exposed to: freshwater (FW), 6 months of experimentally induced aestivation in air (6mAe), and 6 days after arousal from 6 months of aestivation (6mAe6d). We found that both gills and lungs expressed a NOS isoform recognized by anti-mammalian eNOS and phospho-eNOS (p-eNOS) antibodies. Despite the proposed absence of eNOS gene in teleost [24], physio-pharmacological investigations with NOS inhibitors, NADPH-diaphorase, and immunofluorescence with mammalian antibodies suggested an “eNOS-like” activity in various fish tissues (see [18] and [24]), including the heart [25] and [26]. For example in zebrafish, inducible NOSb (iNOSb) was suggested in the heart to perform functions comparable to mammalian eNOS [27]. As suggested [24], in teleost eNOS functions may be covered by a neuronal NOS (nNOS) isoform possessing an endothelial-like consensus sequence. On these premises, we will refer to the NOS-like enzyme detected in gills and lungs of P. annectens as eNOS. Our findings indicate that the expression and activity patterns of the molecular components of the signaling pathways examined in this study changed in a tissue-specific manner in parallel with the branchial and pulmonary readjustments of P. annectens during the maintenance and arousal phases of aestivation.2. Materials and methods2.1. AnimalsJuvenile specimens (n = 15) of P. annectens (100–150 g) were collected in Central Africa and imported by a local fish farm in Singapore. Animals were maintained in plastic aquaria filled with dechlorinated daily changed FW, containing 2.3 mM Na+, 0.54 mM K+, 0.95 mM Ca2+, 0.08 mM Mg2+, 3.4 mM Cl− and 0.6 mM HCO3−, at pH 7.0 and 25 °C. During the acclimation period in the laboratory, fish were fed with frozen fish meat. Food was withdrawn 96 h prior to experiments. Procedures adopted in this study were approved by the Institutional Animal Care and Use Committee of the National University of Singapore (IACUC 035/09).2.2. AestivationA group (N = 9) of fish was allowed to enter into a state of aestivation in air for 6 months. Aestivation was induced according to Chew et al. [28]. Briefly, P. annectens were maintained in plastic aquaria containing 10 ml of dechlorinated freshwater. On average, the water would evaporate in 6–8 days, during which the formation of the cocoon begins. After 6–8 days, aestivating fishes were encased in a complete cocoon.After 6 months of aestivation (6mAe), a group (N = 6) of lungfish was immediately sacrificed for sample collection while another group (N = 3) of lungfish was placed in water and sacrificed 6 days after arousal (6mAe6d). Both FW and 6mAe6d animals were anesthetized in 0.1% tricaine methane sulfonate (Sigma, St. Louis, MO, USA) followed by pithing, while 6mAe fishes were killed directly by pithing. Gills and lungs were quickly excised and processed for the specific protocol. An additional group of fish (N = 6), maintained in FW, served as control.2.3. Light microscopyFor conventional light microscopy, selected tissue samples were dehydrated in graded ethanol, embedded in Paraplast (Sherwood, St. Louis, USA), and serially sectioned at 8µm. Dewaxed sections were stained with hematoxylin and eosin for general tissue observations.2.4. Immunofluorescence microscopyGills (FW: n = 3; 6mAe: n = 3) and lungs (FW: n = 3; 6mAe: n = 3) were flushed in phosphate-buffered saline (PBS). Then samples were fixed in a solution of methanol–acetone–water (MAW = 2:2:1), dehydrated in graded ethanol (90% and 100%), cleared in xylene, embedded in Paraplast (Sigma), and serially sectioned at 8 µm.Dewaxed sections were rinsed in Tris-buffered saline (TBS) and incubated with 1.5% bovine serum albumin (BSA) in TBS for 1 h. Then, they were incubated overnight at 4 °C with a rabbit polyclonal antibody directed against eNOS (Sigma) or a goat polyclonal antibodies directed against p-eNOS and Hsp-90 (Santa Cruz Biotechnology, Inc.), as previously reported [29]. All antibodies were diluted 1:100. For signal detection, slides were washed in TBS (3 × 10 min) and incubated with FITC-conjugated anti-rabbit (Sigma; 1:100) or anti-goat IgG (Sigma; 1:100). For nuclear counterstaining, selected tissue sections were incubated for 5 min with propidium iodide (Sigma; 1 µg/ml) at room temperature. Slides were observed using a confocal microscope (DMI 4000 LEICA, Wetzlar, Germany). Controls were performed by omitting the secondary antibody. Controls were routinely negative.2.5. Western blotting and densitometric analysesTo evaluate whether aestivation and arousal correlate with differences in the expression of NOS and its protein partner, Western blotting analysis was performed on gill and lung homogenates in both aquatic (FW, 6 days after arousal) and aestivating conditions.Samples of gills (FW: n = 3; 6mAe: n = 3; 6mAe6d: n = 3) and lungs (FW: n = 3; 6mAe: n = 3; 6mAe6d: n = 3) were rapidly immersed in liquid nitrogen and stored at −80 °C. They were prepared according to Amelio et al. [29]. Tissues were suspended in ice-cold Tris–HCl buffer (30 mM; pH 7.4) containing EGTA (15 µM), EDTA (10 µM), dithiothreitol (5 µM), pepstatin-A (0.01 µM), PMSF (1 µM), leupeptin-A (0.02 µM), benzamidine (0.1 µM) and tetrahydrobiopterin (BH4) (0.1 µM). They were then homogenized with an Ultra Turrax homogenizer (IKA-Werke, Staufen, Germany) at 22,000 rpm for 10 s. Homogenates were centrifuged at 10,000 × g for 60 min at 4 °C and the supernatant was used for Western blotting. Protein concentration was determined according to Bradford [30] with BSA as a standard for comparison.Samples of gills and lungs containing 100 µg of proteins were heated for 5 min in Laemmli buffer [31], separated by SDS–PAGE using 8% in a Bio-Rad Mini Protean-III apparatus (Bio-Rad Laboratories, Hercules, CA, USA) and then electro-blotted onto polyvinylidene difluoride membrane (Hybond-P, Amersham, GE Healthcare Biosciences, Pittsburgh, USA) using a mini trans-blot (Bio-Rad Laboratories). The membrane was blocked with TTBS buffer containing 5% non-fat dry milk. They were then incubated overnight at 4 °C with either rabbit polyclonal antibodies directed against eNOS (Sigma), HIF-1α, Akt, p-Akt (ser 473) and β-actin, or goat polyclonal antibodies directed against p-eNOS and Hsp-90 (Santa Cruz Biotechnology, Inc.). All antibodies were diluted 1:500 in TTBS containing 5% BSA. The peroxidase linked secondary antibodies (anti-rabbit, anti-goat) (Amersham) were diluted 1:5000 in TTBS containing 5% non-fat dry milk. Immunodetection was performed using an enhanced chemiluminescence ECL PLUS kit (Amersham). Autoradiographs were obtained by exposure to X-ray films (Hyperfilm ECL, Amersham). Immunoblots were digitalized and the densitometric analysis of the bands obtained was carried out using WCIF Image J based on 256 gray values (0 = white; 256 = black). Quantification of the bands was obtained by measuring (5 times on each band) the mean optical density of a square area after the background area was subtracted.2.6. Statistical analysisAbsorbance measurements and the gray values obtained from the densitometric analysis were expressed as means ± SE of determinations for each sample. Differences between the groups were evaluated by non-parametric Mann–Whitney U test, in the case of p-eNOS/eNOS and p-Akt/Akt ratios, and by one-way analysis of variance (ANOVA) followed by Bonferroni multiple comparison test in the case of Hsp-90 and HIF-1α expression. Statistical significance for both statistical tests was established at *p < 0.05, **p < 0.005 and ***p < 0.0005. The statistical analysis of the data was performed using GraphPad InStat® software, version 3.10 for Windows.3. Results3.1. Light microscopyIn FW conditions, the lung appeared as a membranous sac provided with numerous septa of variable thickness and height that subdivided the inner space into smaller compartments (Fig. 1A). According to the observations made by Zaccone et al. [32] in Protopterus, the septa contained an extracellular matrix rich in collagen and elastin, and numerous smooth muscle cells. Here, we observed that these smooth muscle cells were housed in individual niches limited by the extracellular matrix ( Fig. 1B). The inner surface of the air spaces was lined by a respiratory epithelium which overlaid an extensive capillary plexus ( Fig. 1A and B). Capillaries appeared always filled with erythrocytes. The basic structure observed in freshwater animals ( Fig. 1A and B) was maintained during aestivation ( Fig. 1C). In aestivating animals, however, the lung wall was somewhat expanded in relation to that of freshwater animals, in agreement with Sturla et al. [14]. This situation correlated with a larger separation between the septa ( Fig. 1 – compare panels A and C). Observations of gill histological sections showed that, under FW conditions, the branchial architecture was preserved with well-defined spaces for water flow, while during aestivation, secondary lamellae collapsed (data not shown). This is in agreement with previous data obtained in P. annectens [14], which showed, during aestivation, the reduction of respiratory surface and the presence of mucus in the interlamellar space [14].Fig. 1. Composite illustrating basic histological features of the P. annectens lung. Hematoxylin–eosin. In all panels, arrows indicate capillary plexus. (A) Freshwater. The inner surface of the lung is lined by capillaries and by the respiratory epithelium. Thick septa (asterisks) support the membranous lung structure. (B) Freshwater. The septa contain numerous smooth muscle cells housed in matrix niches. Arrowheads indicate typical corkscrew shape of cells. (C) Six months of aestivation. The histological parameters are maintained. Note wider separation between septa. Dilated vessels (V) are apparent. Scale bars: (A) 100 µm; (B) 50 µm; (C) 100 µm.Figure optionsDownload full-size imageDownload as PowerPoint slide3.2. Immunofluorescence microscopy3.2.1. GillsIn FW fish, strong signal for eNOS was found in the endothelial cells of capillaries and along the basal capillary surface. The basal surface of the gill epithelium was also positive for e-NOS (Fig. 2A and B). Aestivating fish showed similar endothelial and epithelial labeling (Fig. 2D and E). In addition, low levels of eNOS expression appeared in the apical surface of the gill epithelial covering during aestivation (Fig. 2D and E). Endothelial and epithelial localization of p-eNOS were similar, both in freshwater and aestivation conditions, to that of eNOS (data not shown). In FW fish, Hsp-90 localized in the capillary endothelium and in the apical surface of the gill epithelial cells (Fig. 2C). By contrast, Hsp-90 labeling was mostly restricted to the vascular endothelium in aestivating fish (Fig. 2D).Fig. 2. eNOS immunolocalization in the gills of FW (A and B) and 6mAe (D and E) P. annectens. eNOS is localized in vascular endothelium (yellow arrows), and in the basal surface of gill epithelium (white arrows). During aestivation (D and E), weak eNOS signal appears at the apical surface (blue arrows) of epithelial cells. In FW, Hsp-90 is localized in endothelium (yellow arrows) and the apical surface (white arrows) of epithelial cells (C). During aestivation (F), the signal is confined to endothelial cells (yellow arrows). Nuclei counterstaining: propidium iodide. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Figure optionsDownload full-size imageDownload as PowerPoint slide3.2.2. LungsIn FW fish, a strong eNOS signaling was found in the endothelial cells of capillaries and in the respiratory epithelium (Fig. 3A and B). A weak reticular eNOS signal was present in the septa that separate individual air spaces and was mainly located at the boundaries of the niches in which smooth muscle cells were housed. The same localization was observed under aestivating conditions, characterized, however, by a stronger immunofluorescence labeling (Fig. 3D and E). As in the gills, in both FW and aestivating fish, p-eNOS was mostly restricted to the capillary endothelium and to the epithelium (data not shown). Hsp-90 was also localized at the endothelial and epithelial levels (Fig. 3C and F). No differences in Hsp-90 labeling were observed between FW and aestivating fish.Fig. 3. eNOS immunolocalization in the lungs of FW (A and B) and 6mAe (D and E) P. annectens. eNOS is localized in vascular endothelium (yellow arrows), epithelium (white arrows) and septa (blue arrows). Hsp-90 (C and F) is localized in vascular endothelium (yellow arrows) and epithelium (white arrows) of both FW and 6mAe fish. Nuclei counterstaining: propidium iodide. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Figure optionsDownload full-size imageDownload as PowerPoint slide3.3. Western blottingWestern blotting analysis of gills and lungs homogenates revealed the presence of immunoreactive bands corresponding to the approximate MW of eNOS (135 kDa), p-eNOS (140 kDa), Akt (60 kDa), p-Akt (60 kDa) Hsp-90 (90 kDa) and HIF-1α (120 kDa).3.3.1. GillsDensitometric quantification of the blots revealed that the amount of activated eNOS, evaluated as p-eNOS/eNOS ratio, decreased in 6mAe fish, with respect to the FW conditions (Fig. 4A and A1). By contrast, the p-eNOS/eNOS ratio increased significantly 6 days after arousal (Fig. 4A and A1). In the aestivation fish, p-Akt expression was significantly reduced (Fig. 5A and A1), the values returning to FW levels 6 days after arousal (Fig. 5A and A1).Fig. 4. Western blotting of eNOS and p-eNOS in extracts from gills (A) and lungs (B) of P. annectens exposed to FW (n = 3), 6mAe (n = 3) or 6mAe6d (n = 3). A1 and B1 showed the densitometric quantification of the blots. Loaded protein amount was verified using anti-β-actin antibody. Statistical differences were evaluated by non-parametric Mann–Whitney U test (*p < 0.05, **p < 0.005 and ***p < 0.0005).Figure optionsDownload full-size imageDownload as PowerPoint slideFig. 5. Western blotting of Akt and p-Akt in extracts from gills (A) and lungs (B) of P. annectens exposed to FW (n = 3), 6mAe (n = 3) or 6mAe6d (n = 3). A1 and B1 showed the densitometric quantification of the blots. Loaded protein amount was verified using anti-β-actin antibody. Statistical differences were evaluated by non-parametric Mann–Whitney U test (*p < 0.05, **p < 0.005 and ***p < 0.0005).Figure optionsDownload full-size imageDownload as PowerPoint slideHsp-90 was always present in the gills (Fig. 6A). However, densitometric analysis showed a significant decrease in its expression after 6 months of aestivation, the levels reaching or surpassing the FW values 6 days after arousal (Fig. 6A1). On the other hand, the expression of HIF-1α was higher in aestivating fish, decreased 6 days after arousal and was lower in FW conditions (Fig. 7A and A1).Fig. 6. Western blotting of Hsp-90 in extracts from gills (A) and lungs (B) of P. annectens exposed to FW (n = 3), 6mAe (n = 3) or 6mAe6d (n = 3). A1 and B1 showed the densitometric quantification of the blots. Loaded protein amount was verified using anti-β-actin antibody. Statistical differences were evaluated by one-way ANOVA followed by Bonferroni multiple comparisons test (*p < 0.05, **p < 0.005 and ***p < 0.0005).Figure optionsDownload full-size imageDownload as PowerPoint slideFig. 7. Western blotting of HIF-1α in extracts from gills (A) and lungs (B) of P. annectens exposed to FW (n = 3), 6mAe (n = 3) or 6mAe6d (n = 3). A1 and B1 showed the densitometric quantification of the blots. Loaded protein amount was verified using anti-β-actin antibody. Statistical differences were evaluated by one-way ANOVA followed by Bonferroni multiple comparisons test (*p < 0.05, **p < 0.005 and ***p < 0.0005).Figure optionsDownload full-size imageDownload as PowerPoint slide3.3.2. LungsIn FW lungs, the p-eNOS/e-NOS ratio was very low when compared to that measured in aestivating fish. However, the ratio approached FW levels 6 days after arousal (Fig. 4B and B1). While similar p-Akt expression levels were detected in FW and aestivating lungs, the levels decreased significantly 6 days after arousal (Fig. 5B and B1). Hsp-90 expression levels were higher in aestivation and after arousal than in FW conditions (Fig. 6B and B1). By contrast, HIF-1α expression significantly decreased in aestivating animals, returning to the FW levels after arousal (Fig. 7B and B1).Table 1 summarizes the changes in the ratios p-eNOS/eNOS, and p-Akt/Akt, and in the expression of Hsp-90 and HIF-1α, which occurred in P. annectens gills and lungs under the different conditions.Table 1.

Effect of grid numbers on the average Nusselt number.Grid numbers(Nuhot)ave(Nucold)ave(Nuhot)ave/(Nucold)ave(Nuhot)ave/(Nuhot)ave (4)(1)56 × 36 × 81.017831.016581.001231.00452(2)56 × 36 × 121.014661.013541.001111.00139(3)56 × 36 × 161.013801.012721.001071.00054(4)56 × 36 × 241.013251.012201.001041.00000Full-size tableTable optionsView in workspaceDownload as CSV4. Computed resultsFig. 2(a) shows the axial component of the gradient of the square of magnetic induction (gradB2)Z along the centerline computed by Biot–Savart\’s law shown in Eq. (6) under the configuration seen in Fig. 1. The (gradB2)Z takes minimum and maximum values at Z = ±3.75 and zero at Z = 0 as shown by open circles.Fig. 2. (a) The (gradB2)Z on the centerline, and the radial profiles of (b) the (gradB2)R and (c) the (gradB2)Z.Figure optionsDownload full-size imageDownload as PowerPoint slideThe radial profiles of the radial component of the gradient of the square of the magnetic induction (gradB2)R and its axial component (gradB2)Z at various axial locations are shown in Figs. 2(b) and (c). To evaluate the heat transfer rate of the magnetothermal Rayleigh–Benard convection in a gravitational field, we used the magnetic Rayleigh number first proposed by Braithwaite et al. [11] as follows.equation(14)Ram=Ra·(1−MZ(gradB2))Ram=Ra·(1−M(gradB2)Z)In contrast, the heat transfer rate of the magnetothermal Rayleigh–Benard convection induced by the magnetic force alone in a non-gravitational field was evaluated by applying Eq. (15).equation(15)Ram=−Ra·M(gradB2)Z=−β(θh−θc)h3αν·χb2ξh(gradB2)ZPlease note that g term is dropped from this BMS-536924 equation for the non-gravitational system. In the present numerical computations, the enclosure filled with paramagnetic air was located at the axial location, which is subjected to a more uniform axial magnetic force and minimal radial magnetic force. This configuration was devised to get the axial magnetic acceleration as homogenous as possible.In Fig. 2(b), the (gradB2)R has the largest inhomogeneous distribution at the magnetic center of Z = 0. In contrast, the (gradB2)Z is always zero at Z = 0, as seen in Fig. 2(c). At Z = ±3.75, where (gradB2)Z takes minimum and maximum values on the centerline ( Fig. 2(a)), the (gradB2)R and (gradB2)Z values still show a large inhomogeneous distribution ( Fig. 2(b) and (c)).To carry out the numerical computation for convection under a magnetic force acceleration, we chose the axial locations of Z = ±6.5 as the installation position of the enclosure, as shown by the thick line in Fig. 2(b) and (c) to generate the magnetic force (gradB2)R and (gradB2)Z as uniform as possible. At these two axial locations, the (gradB2)Z at R = 5 is about 15% larger than that at R = 0. In Eqs. (14) and (15), the value along the centerline is used as the axial component of the gradient of the square of the magnetic induction, i.e., (gradB2)Z = ±2.3812 × 10−4 at Z = ?6.5 as shown by the solid circle in Fig. 2(a). The enclosure is fixed at Z = +6.5 HZ (HZ = h/h = 1) above the magnetic center of Z = 0 or Z = −6.5 HZ below the magnetic center, as shown in Fig. 1.In the subsequent computations, the computer used for the present calculations was a workstation with an Intel(R) Core(TM) i7-4960X CPU 3.60 GHz processor and 64 GB of RAM. The Intel Visual Fortran Composer XE 2011 was used as a compiler. The CPU time was 2,768,500 s to compute the response in the time range from τ = 0 to τ = 50. The transient responses of the average Nusselt number on a hot plate for Rayleigh–Benard convection (curve (A), g only) and magnetothermal Rayleigh–Benard convection (curves (B)–(D), g + magnet) are shown in Fig. 3. For (A) in the system without the magnetic field, the Rayleigh number is Ra = 5000. For (B) and (C) for the system with both the gravitational and magnetic forces, the magnetic Rayleigh numbers are Ram = 11,072 and 17,978, respectively from Eq. (14), since the enclosure is located at Z = +6.5 and (gradB2)Z = −2.3812 × 10−4 at the central height of the enclosure and the centerline of the enclosure. In contrast, for (D) in the system with the magnetic field, the magnetic Rayleigh numbers is Ram = 952, since the enclosure is located at Z = −6.5 and (gradB2)Z = +2.3812 × 10−4.Fig. 3. Transient responses of the average Nusselt number on a hot plate for Rayleigh–Benard convection and magnetothermal Rayleigh–Benard convection computed for numerical conditions (A), (B), (C), and (D). (A) Ra = 5000, M = 0, (B) Ra = 5000, M = 5100 (Ram = 11,072), (C) Ra = 5000, M = 10,900 (Ram = 17,978), (D) Ra = 5000, M = 3400 (Ram = 952).Figure optionsDownload full-size imageDownload as PowerPoint slideFrom Fig. 3, as Ram increases, i.e., numerical condition (A) < (B) < (C), it BMS-536924 is seen that the average Nusselt number increases at Z = +6.5, and the stronger oscillatory thermal convection is generated with the increase of magnetic strength. In contrast, when the Ram decreases, i.e., numerical condition (D), at Z = −6.5 the quasi-thermal conduction state is attained at a steady state.For (A) – (D), the ratios of average Nusselt numbers on end plates were less than about 2% at the maximum as listed in Table 6. We therefore consider that the energy balance in the system is almost satisfied with meshes of 56 in the radial direction, 36 in the circumferential direction, and 12 in the axial direction.Fig. 4 shows the instantaneous temperature distributions in a horizontal cross-section at HZ/2 (the central height of the enclosure) and A–B, C–D vertical cross-sections for the numerical conditions (A), (B), (C), and (D) shown in Fig. 3. Fig. 4(A)–(C) show those at τ = 30, and Fig. 4(D) shows that at τ = 10. For numerical condition (A), when the gravitational buoyant force overcame the fluid viscosity, the asymmetric Rayleigh–Benard convection was generated suddenly by the velocity development in the three directions and the average Nusselt number increased steeply from unity at around τ = 1, as seen in Fig. 3(A). Fig. 4(A) shows the instantaneous temperature contours in the transient state.Fig. 4. Instantaneous temperature distributions in a horizontal cross-section at HZ/2 and A–B, C–D vertical cross-sections for (A), (B), (C), and (D) shown in Fig. 3(A) Ra = 5000, M = 0, τ = 30, (B) Ra = 5000, M = 5100 (Ram = 11,072), τ = 30, (C) Ra = 5000, M = 10,900 (Ram = 17,978), τ = 30, (D) Ra = 5000, M = 3400 (Ram = 952), τ = 10.Figure optionsDownload full-size imageDownload as PowerPoint slideFor numerical conditions (B) and (C), the axisymmetric magnetothermal Rayleigh–Benard convection was induced immediately by the velocity development in the radial and axial directions after the magnetic field was applied, and the axisymmetric convection was then changed to the asymmetric convection by the velocity development in the circumferential direction. Next, the behavior of the complicated convections as seen in Fig. 4(B) and (C) was observed in the stronger unsteady state compared to the gravitational convection of Fig. 4(A).For numerical condition (D), the axisymmetric magnetothermal Rayleigh–Benard convection was induced slightly, and the quasi-conduction state was attained in the reverse density stratification. As a result, the average Nusselt number was always about unity, as seen in Fig. 3.Fig. 5 shows the summary of the computed average Nusselt numbers for Rayleigh–Benard convection and magnetothermal Rayleigh–Benard convection versus the Rayleigh number. The thick solid line shows that for the gravitational convection reported by Silveston [12] and [13]. The half-closed symbols show the computed average Nusselt numbers at Ra = 1000, 2500, 5000, 7500, and 10,000 without a magnetic field. The closed symbols show those under both the gravitational and the magnetic fields induced by placing the enclosure at Z = +6.5. In contrast, the open symbols show those in the enclosure at Z = −6.5. (A), (B), (C), and (D) in the figure correspond with those in Fig. 3 and Fig. 4.Fig. 5. Summary of computed average Nusselt numbers for Rayleigh–Benard convection and magnetothermal Rayleigh–Benard convection versus the Rayleigh number with the average Nusselt number curve reported by Silveston [10] and [11]. Ra = 1000 (circles), 2500 (triangles), 5000 (squares), 7500 (inverted triangles), and 10,000 (rhomboids). (A) Ra = 5000, M = 0, (B) Ra = 5000, M = 5100, (C) Ra = 5000, M = 10,900, (D) Ra = 5000, M = 3400.Figure optionsDownload full-size imageDownload as PowerPoint slideThe computed average Nusselt numbers for the Rayleigh–Benard convection in the system without the magnetic field (half-closed symbols) are in good agreement with the experiments by Silveston in Fig. 5. As shown by the closed symbols, the computed average Nusselt number for each Rayleigh number increased with the increase of the magnetic strength compared to that in the system without the magnetic field; i.e., the gravitational convection was enhanced by the magnetic force. In contrast, as shown by the open symbols ((D),Z = −6.5), the computed average Nusselt number decreased with the increase of the magnetic strength compared to the gravitational one and it gradually approached unity; i.e., the Rayleigh–Benard convection was suppressed by the magnetic force. Here, the dimensional equivalence is shown in Table 3.Table 3.

Characteristics of the nitrogen removal processes in the side- and mainstream and features of the anaerobic digestion. AD: anaerobic digestion; COD: chemical oxygen demand.Scenario1a1b23Sidestream treatmentNitrit./Denitrit.DEMONDEMONN-input (g N m−3 sewage treated)7.812.610.8N-removal efficiency (%)838991N2O–N emission (% of N input)–N emission (% of N input) consumed (g COD g−1 N removed)2.720.310.43Electricity consumption (kWh/kg N removed)2.661.501.70Electricity saved compared to Nitrit./Dentrit. (kWh m−3 sewage treated)0+0.054+0.044via avoided COD consumption (kWh m−3)0+0.041+0.035via reduced electricity consumption (kWh m−3)0+0.013+0.009Mainstream treatmentNitrif./Denitrif.Nitrif./Denitrif.DEMONN-input (g N m−3 sewage treated)37.639.430.8N-removal efficiency (%)798581N2O–N emission (% of N input)0.00730.00692.1758NO–N emission (% of N input)0.00830.00780.0143COD/N ratio (g COD/g N) consumed (g COD/g N removed)3.403.524.26Electricity consumption (kWh/kg N removed) addition to ADKitchen waste (g DM m−3 treated sewage)39329239Fat (g DM m−3 treated sewage)1490Methane content (%)60.561.762.1Full-size tableTable optionsView in workspaceDownload as CSVThe implementation of mainstream DEMON presented for this case study no energy improvement, as it did not yet operate successfully (Table 1), discussed in section E of the supplementary file. On labscale, growth of anammox-bacteria has though already been reported in a labscale reactor fed with effluent of an A-stage (Lotti et al., 2014), illustrating the feasibility of DEMON in the B-stage.At this moment insufficient comparative data are available for N2O emissions of various nitrogen removing processes, making it hard to judge these processes against each other. Reported N2O emission in activated sludge systems based on nitrification/denitrification ranged from 0.001 to 25% of the N load and were mainly linked with nitritation activity (Desloover et al., 2012). Moreover, the N2O emission is mainly linked with the operational conditions rather than the process itself (Chandran et?al., 2011, Desloover et?al., 2012, Domingo-Félez et?al., 2014, Kampschreur et?al., 2009a and Kampschreur et?al., 2009b). For the treatment plant of Strass, N2O–N emissions in the nitrification/denitrification step (B-stage) were very low (<0.01% of the N load) compared to 1.0–1.3% of the N load measured during side stream DEMON (scenario 2 and 3), which is in line with the average 1% reported in literature (Joss et?al., 2009, Kampschreur et?al., 2009a, Kampschreur et?al., 2009b and Weissenbacher et?al., 2010). As nitrite, a precursor for N2O production (Kampschreur et?al., 2009b and Chandran et?al., 2011), accumulated up to 96 mg N L−1 during sidestream nitritation/denitritation (sc 1) compared to 0 mg N L−1 and 1 mg N L−1 in the mainstream nitrification/denitrification and sidestream DEMON steps, respectively, increased levels of N2O were expected for sidestream nitritation/denitritation (scenario 1b). Due to possible nitrite accumulation in processes based on nitritation, an increased level of N2O emission (2.2% of N load) in the mainline was measured when DEMON was implemented in the B-stage initially based on nitrification/denitrification. Adaptation and improved process control could probably lower the N2O emissions (Ahn et al., 2011).From the comparison of the different nitrogen removal processes, it could be shown that in the WWTP in Strass DEMON lowers the electricity needs, but it has the potential to increase N2O emission due the higher risk for accumulation of nitrite. Nitrite accumulation and N2O emission could potentially be decreased due to better balancing of nitritation with denitrification and anammox rates as well as due to SM 406 (Chandran et al., 2011). In this light specific aeration strategies might be used to mitigated N2O emission, as reported for labscale single-stage nitritation/anammox reactors by Domingo-Félez et al. (2014). However the N2O emission rates of the Strass WWTP are already in the same order as the lowest reported in their study.Concerning the addition of co-substrate, the increase in methane content of the biogas was minimal, thus showing no considerable improvement in this aspect. This might have been due to the fact that the sludge from the A-stage provided a considerable amount of readily digestable COD which is inherent to this type of reactor, induced by its short sludge-water contact time, and that the DEMON implementation in the sideline already increased the amount of this high-rate sludge flow to the digester (Wett et al., 2007). Process stability and digestion of the sludge had also not clearly improved.3.2. From an energy-demanding to an energy-providing WWTP on system levelOn a system level, electrical energy autarky is already reported to have been achieved for the Strass WWTP (Nowak et?al., 2011 and Wett et?al., 2007). An overview of the electricity consumption is presented in Fig. 2. The total electricity consumption was 0.088, −0.209 and −0.243 kWh m−3 for scenarios 1, 2 and 3, respectively. This implies that energy autarky is only achieved for scenarios 2 and 3 at 158 and 178%, respectively. Despite the fact that only the B-step and the sideline treatment differ in set-up between the scenarios, the electricity consumption amounts vary considerably for the A-stage and other processes among them. To the contrary, there are no large differences between the nutrient inputs m−3 sewage (see Table S1). The reason for the overall lower electricity consumption for the A-step and other processes is consequently a better management. Also a better operation of the nitrification/denitrification B-stage was conducted for scenario 2 compared to scenario 1 as its electricity consumption is considerably lower while its COD- and N-input m−3 are even higher (Table S1).Fig. 2. Electricity consumption of different steps of the wastewater treatment plant for the three scenarios, expressed per cubic meter of treated sewage.Figure optionsDownload full-size imageDownload high-quality image (223 K)Download as PowerPoint slideThe largest electricity saving was provided through the addition of co-substrate to the digester; compared to scenario 1 an additional amount of 0.200 and 0.187 kWh m−3 was generated in scenarios 2 and 3, respectively. This implies a better usage of the plant\’s digester capacity.Electricity consumption per cubic meter in the sideline is quasi equal for all scenarios, though this does not reflect the advantages of DEMON implementation as the N-inputs differ substantially between the sideline inputs of the scenarios, mainly due to addition of co-substrate, which contains N, to the digester (Table 1). Compared to nitritation/denitritation, 0.054 and 0.044 kWh m−3 treated wastewater were indirectly saved in scenario 2 and 3 (Table 1). Note that the consumption in the sideline represents a minor contribution to the overall one on a plant level and that if DEMON runs successfully in the B-stage, energy savings cane be much higher overall.Scherson and Criddle (2014) discuss alternative configurations to lower energy usage of WWTP. Besides DEMON for nitrogen removal, one could apply Coupled Aerobic-anoxic Nitrous Decomposition Operation (CANDO) (Scherson et al., 2014), the application of the reject water as N-fertilizer (McCarty et al., 2011) (compost leachate with high nitrogen content is already studied for its direct application as a fertilizer (Romero et al., 2013)) or crystallization into struvite which can also displace fertilizers (Ueno and Fujii, 2001). More in-depth studies are though needed to unravel the option with the lowest environmental impact. Rodriguez-Garcia et al. (2014) compared the environmental impacts of nitrogen removal from digestion supernatants via struvite, removal over nitrite and DEMON. In their study, DEMON and removal over nitrite score better, however they do not include any impact assessment of the resource consumption, which could otherwise emphasize the important advantage of replacement of conventional fertilizers by struvite.3.3. Environmental impact on a life cycle levelThe environmental impact assessment results are shown in Fig. 3. Note that, expressing the results with other functional units, PE or eutrophication potential (presented in Table S1), instead of cubic meter treated sewage results in a relative larger environmental impact for scenario 3 compared to the other two, but the same findings remain valid.Fig. 3. Environmental impact (with ReCiPe) of the treatment of 1 m3 sewage comprising the Strass wastewater treatment plant, its supply chain and valorization of the byproducts. A positive value represents damage caused, a negative value damage prevention. For some impact categories final damage to natural systems expressed as diversity loss cannot be quantified, hence the respective normalized midpoint outcomes are presented. 100% corresponds to 0.0477 kg N equivalents for marine eutrophication, 1.85E-08 kg trichlorofluoromethane equivalents for ozone depletion and 0.00187 kg non-methane volatile organic compounds equivalents for photochemical oxidant formation. DALY: disability adjusted life years.Figure optionsDownload full-size imageDownload high-quality image (1097 K)Download as PowerPoint slide3.3.1. Resource consumptionThe total CEENE per m3 wastewater equals −1 (scenario 1), −2.5 (scenario 2) and −3.1 (scenario 3) MJex m−3 (1 MJex is equal to about 0.024 tonnes of fossil oil). They are all negatively valued. This means that these (exergy) amounts of resources did not need to be extracted anymore, were prevented from being extracted, from nature by society due to the wastewater treatment. One may regard this as a benefit for the environment. Regardless of the fact that energy autarky was not achieved for scenario 1, an environmental benefit is achieved. Energy autarky does thus not give a complete picture of the resource consumption. When only regarding the actual resources consumed (positive values), main contributors are electricity, chemicals and infrastructure.The amounts of resources needed to produce the chemicals used in the WWTP are 0.39 (sc. 1), 0.45 (sc. 2) and 0.29 (sc. 3) MJex m−3. The added chemical sodium aluminate is though here considered as a waste product and thus its impact is not accounted for. If this had to be obtained from the regular market, the resource consumption of its conventional production needed to be added, resulting in a considerable additional (mainly fossil) resource extraction from nature of 1.1 (scenario 1), 0.6 (scenario 2) and 0.5 (scenario 3) MJex m−3.Concerning infrastructure, it accounts for an important part of the consumed amount of resources, even 50% for scenario 3, this at a fixed 0.42 MJex m−3. The main infrastructure assets responsible are the land occupation by the WWTP (22%), the road infrastructure (22%), the multi-storey building (12%), the building hall (8%), concrete (10%) and steel (8.5%) for the basins. Keep in mind that the WWTP lifetime is set at 50 years while this is conventionally in the order of 20 years (Corominas et?al., 2013 and Foley et?al., 2010). In fact, if the lifetime of the plant is lower than 14.6, 7.2 or 5.9 year for scenario 1, 2 and 3, respectively, resource extraction is not prevented.Regarding replaced products (negative values), per produced kWh of electricity through co-generation, 9.5 MJex resources are saved conventionally needed to produce 1 kWh present on the Austrian grid, which is mainly obtained through burning of fossil fuels. Note that the resource footprint of electricity production may vary substantially between countries. The net production of electricity led to a considerable resource saving of 2.0 and 2.3 MJex m−3 for scenario 2 and 3, respectively. Note that another benefit of the biogas burning was the heat produced that was used to heat up the digester and administrative buildings. This advantage was though not explicitly quantified. DEMON implementation led to the significant saving of 0.52 and 0.42 MJex m−3 for scenario 2 and 3, respectively.In total, replacement of conventional fertilizers by stabilized sludge digestate accounted for −2.77 (sc. 1), −1.56 (sc. 2) and −1.67 (sc. 3) MJex FU−1 and was thus as important as electricity generation in counteracting the need for resources. The carbon in the composted digestate displaced the addition of carbon fertilizers peat and straw. Peat is regarded as a fossil fuel, hence a significant amount of fossil fuels was prevented from being extracted. Since straw production requires land, a considerable amount of land was not occupied anymore to produce cereal plants. Even though more co-substrate was added to the digester in scenario 2 and 3, the amount of scenario 1 is about double that of latter scenarios (Table 1). This explains the higher amount of peat and straw resources saved in scenario 1. The replaced N & P fertilizers mainly consists of the intensively produced triple super phosphate. However this significant environmental benefit only occurs if farmers apply lesser amounts of the respective conventional fertilizers when using the stabilized digestate.3.3.2. Damage to human healthScenario 1 has two sets of values for N2O and NO emissions of the sideline process: a and b. This led to two sets of impacts. Regarding the total (net) damage to human health, we obtained following values 2.89E-07 (sc. 1a), 5.52E-07 (sc. 1b), 1.65E-07 (sc. 2) and 4.20E-07 (sc. 3) DALY FU−1. This means that over the complete WWTP lifetime of 50 years, considering an average flow rate, the number of years lost due to ill-health, disability or early death would be 150 (sc. 1a), 285 (sc. 1b), 85 (sc. 2) and 217 (sc. 3). Only three impact categories are relevant for human health damage: climate change, human toxicity and particulate matter formation (Fig. 3).The climate change impact of greenhouse gas emissions is an important category (1 kg CO2-eq. stands for 1.4E-06 DALY). Emissions by the plant itself were 0.065 (sc. 1a), 0.252 (sc. 1b), 0.094 (sc. 2) and 0.371 (sc. 3) kg CO2-eq. m−3. These low results are outstanding for WWTP standards. Scenario 2, with the lowest and measured emissions, has an overall greenhouse gas emission of about 5.5 kg CO2-eq PE−1 year−1, which is low compared to the average reported operational CO2 footprints of WWTP ranging from 12 to 80 kg CO2-eq PE−1 year−1 (Hospido et al., 2008; Clauwaert et al., 2010).The main substance causing this climate change impact is N2O (298 kg CO2-eq kg−1 emission) with an impact share of 74% (sc. 1a), 93% (sc. 1b), 83% (sc. 2) and 98% (sc. 3) (the other share is attributed to methane). For scenario 1a, 1b and 2 the main source (>90%) of N2O is the sideline removal of nitrogen (see Table S1). In Scenario 3, DEMON operation in the B-step emitted 86% of the total N2O, due to nitrite accumulation, and is the main reason why N2O emission were much higher for this scenario. The N2O emissions by the N-removal steps, such as DEMON, have a high impact, hence controlling them is crucial in reducing the environmental impact of the WWTP. This can be achieved through specific operation strategies as described in Section 1 of results and discussion. Alternatively, The CANDO process could present a good solution. In this process N2O is reduced to N2 together with oxidation of CH4 from the digester, and this while even creating energy (Scherson et al., 2014).Considerable remediation of climate change occurs due to prevention of fossil CO2 emissions from peat applied on the land. The (avoided) impact of fossil CO2 emitted by electricity is also important. For scenario 2 the climate change impact has even a small negative value, implying that from a life cycle perspective more greenhouse gases were taken up than emitted. If the CO2 emissions of the plant (Table S1) were accounted for, this would result in an increase of 2.4–6.8 times depending on the scenario. By consequence, the impact on human health would increase in the same order. On the other hand, the greenhouse gas emissions which would normally occur when disposing the sewage in the environment, what is here prevented, are not considered. Latter should be further investigated.The impact category human toxicity is the second most important category, responsible for 1.53E-07 (sc. 1), 1.24E-07 (sc. 2), 2.90E-08 (sc. 3) DALY FU−1. The metals present in the on land applied stabilized digestate is the reason for this. The impact per element is given in the Fig. S1 of the supplementary file. The high concentration of Zn is the most harmful; it represents about 70% of the human toxicity impact, followed by Cd, As, Pb and Hg. To decrease the impact of sludge application, a possible extraction of metals or prevention of application if heavy metal concentrations are too high, should be considered. There is a minor mitigating effect due to replacement of electricity from the grid and prevention of land application of conventional N & P fertilizers. However in scenario 3 this remediation effect is quite considerable at 70%.Particulate matter formation is also a relevant category regarding damage to human health, responsible for 7.61E-08 (sc. 1), 4.40E-08 (sc. 2), 4.55E-08 (sc. 3) DALY FU−1. The main impact is coming from the plant emission of NO2, a precursor for particulate matter. This compound is mainly (>93%) emitted during burning of the biogas in the cogeneration unit. However there is a remediation effect, through prevention of mainly SO2 emission during production of the displaced N&P fertilizer mix, straw and electricity production. Installation of exhaust gas purifiers such as a DeNOx installation is though recommended.Overall, according to the ReCiPe method, the damage done to human health through stabilized digestate application is 66% (sc. 1), 93% (sc. 2) and 47% (sc. 3) of the prevented damage due to replacement of production and application of conventional fertilizers. This is in all cases lesser than 100%, by consequence it was in all cases better for human health to apply the stabilized digestate to the land, if conventional fertilizers are replaced, despite the presence of toxic heavy metals, considering the results for the ReCiPe method.On the other hand, if we apply the USEtox method instead of the USES-LCA method of ReCiPe to assess the human toxicity of the metals present in the stabilized digestate and replaced conventional fertilizers, different results and thus findings could have been obtained (Fig. S2).The human toxicity impact of the metals present in the stabilized sludge are then: 5.04E-05 (sc. 1), 5.03E-05 (sc. 2) and 2.17E-0.5 DALY (sc. 3) m−3 treated wastewater. The net impact compared to the replaced conventional fertilizers is positive as the avoided impact values for the fertilizers are much lower: −1.51E-06 (sc. 1), −1.01E-06 (sc. 2) and −9.70E-0.7 DALY (sc. 3) m−3 treated wastewater. Yet again, as for the ReCiPe results, zinc represents the most important share of the impact (>90% of damaging impact). These calculated impact values are a magnitude of 100 higher than those calculated via ReCiPe, as expected out of the findings of Heimersson et al. (2014). Combining the USEtox impact values with the impact values of the other categories would increase the total impact on human health with a similar factor and make human toxicity the most important category. The impact caused by the application of the stabilized digestate would then be much higher (factor 10–100) than the damage prevented due to prevention of usage of conventional fertilizers. This finding is in contrast with that obtained via the ReCiPe method.Hence, overall, it would be better, out of caution, to first find means to lower metal, especially Zn, content in the stabilized digestate before applying it to the land. In fact, a recent study has highlighted the potential and possible economic benefit in recovering metals from biosolids (Westerhoff et al., 2015).3.3.3. Damage to natural systemsDamage to nature is assessed as ecosystem diversity loss. Note however that for this area of protection no total final impact can be considered since for some impact categories this final impact is not quantifiable yet by models, being: ozone depletion, marine eutrophication and photochemical oxidant formation. For these their midpoint impact is given (Fig. 3). Not considering latter impact categories, the most important among the others are: climate change, freshwater eutrophication, terrestrial ecotoxicity and land occupation and transformation. The preliminary total impact is then −8.75E-10 (sc. 1a), 6.12E-10 (sc. 1b), −7.40E-10 (sc. 2) and 1.13E-09 (sc. 3) species*yr. For scenarios 1a and 2, there is thus already a positive impact, implying a possible benefit for the environment.Climate change is here also one of the most important categories (1 kg CO2-eq. stands for 7.93E-09 species*yr). In fact, it is almost the only category which leads to a loss of ecosystem diversity. For climate change, the same conclusions may be drawn as done concerning damage to human health.The removal of phosphate (which is here accounted for by freshwater eutrophication), one of the main goals of a WWTP, is of relative less importance since it only is responsible for −3.73E-10 (sc. 1), –3.97E-10 (sc. 2) and −3.01E-10 (sc. 3) species*yr. We must note that the quantification in diversity loss for this category is not considered good enough yet (Hauschild et al., 2013). Note that the influent phosphate concentrations are slightly different. For the phosphate in influent and effluent, please regard values in Table 1.There is a significant remediation effect due to prevention of diversity loss of land occupation. This is mainly due to replacement of straw production and to a lesser extent P&N fertilizer mix production. There is also a small impact in the category of terrestrial ecotoxicity mainly due to copper and zinc present in the stabilized digestate applied on the land.The damage done to ecosystem diversity of stabilized digestate application is 15% (sc. 1), 22% (sc. 2) and 12% (sc. 3) of the prevented damage due to displacement of production and application of conventional fertilizers. Here it proves more than worthwhile to apply the digestate to land, this if respective lesser amounts of conventional fertilizers are used.Among the impact categories quantified at midpoint level, marine eutrophication is a relevant one since it covers a share of the eutrophication impact, the type of pollution which primarily should be prevented by a WWTP. The latter is acknowledged in literature (Corominas et?al., 2011, Hospido et?al., 2010 and Rodriguez-Garcia et?al., 2011).Marine eutrophication at midpoint level is expressed as kg N equivalents. In total for all three scenarios, it sums up to 0.038 (sc. 1), 0.039 (sc. 2) and 0.030 (sc. 3). These values quasi represent the net removal of nitrogen in the different scenarios. An overall eutrophication potential, addressing COD, N and P, was calculated using the CML method, represented in Table S1 and discussed in section G of the supplementary file.Ozone depletion (mainly due to NO2) and photochemical oxidant formation (mainly due to halogenated methane compounds emitted during industrial processes) are not considered as relevant impact categories in LCA studies as far as we know. This reflects the higher possibility of their irrelevance.4. ConclusionsOverall, for all three scenarios, the studied life cycle of treatment of 1 m3 sewage comprising the Strass WWTP, its supply chain and the valorization of its products, leads to a prevention of resources extracted from nature and can have a mitigation of diversity loss (though for some relevant impact categories damage cannot be expressed in diversity loss), but it also has an adverse effect on human health. Since it is for now not possible to aggregate the impact to these three areas of protection in a sound manner, it is not yet possible to give a single outcome. It is overall not possible to state or consider that the here studied wastewater treatment is environmentally friendly (compared to disposal of the wastewater in the environment). In general, life cycle assessment and environmental sustainability assessment need further development to be able to give a clear and better picture. Following more specific conclusions can be drawn, keeping latter limitations in mind:-The implementation of partial nitritation/anammox, DEMON, in the sideline induces a considerable saving of natural resources. However, the related N2O emissions should be restrained through further optimizing operational conditions as these are responsible for a large share of the damaging effect on human health through climate change. In the mainstream, DEMON implementation was not fully operational yet but could evoke an even larger environmental benefit.-If respective amounts of conventional fertilizers are replaced, the land application of the sludge digestates induces an environmental benefit through prevention of resource extraction and prevention of ecosystem diversity loss, based on ReCiPe calculations. Concerning human health, no consensus is achieved as the selection of the impact assessment method for the category of human toxicity leads to different findings. If the results of the recommended USEtox method to quantify heavy metal toxicity are considered then the impact caused by the application of the stabilized digestate is much higher (factor 10–100) than the damage prevented due to prevention of usage of conventional fertilizers. Hence, overall, it would be better, out of caution, to first find means to lower the heavy metal, especially Zn, content in the stabilized digestate before applying it to the land.-The addition of co-substrate tot the digester had shown no significant improvement of the digestion process but increased net-energy production. This implies a better usage of the plants\’ infrastructure.-The main strengths of the Strass WWTP are: the A/B process, DEMON in the sideline, usage of waste sodium aluminate instead of buying it from the market, co-susbstrate addition and a long plant lifetime of 50 years.-The Strass WWTP should keep an eye on the following matters: a good management of electricity consumption, DEMON implementation in the mainline, the metal contents in the sludge digestate and the purification of the exhaust gases of the biogas burning.AcknowledgementsT.S. was supported by a research project (number 3G092310) of the Research Foundation – Flanders (FWO-Vlaanderen), H.D.C. by a PhD grant from the Institute for the Promotion of Innovation by Science and Technology in Flanders (IWT-Vlaanderen, number SB-81068) and S.E.V. as a postdoctoral fellow by the Research Foundation – Flanders (FWO-Vlaanderen). The investigations at the Strass treatment plant were also supported by the Austrian Federal Ministry of Environment. The authors gratefully thank Tim Lacoere for technical support, Martin Hell for providing operational data of the plant and Nico Boon, Steven De Meester, Rodrigo Alvarenga and Chris Callewaert for inspiring scientific discussions. Lastly, we are grateful towards the anonymous reviewers for their comments that helped improve the manuscript.Appendix A. Supplementary dataThe following is the supplementary data related to this article:

ASD; MCCOP; CPTS; Collaborative pointing operation1. IntroductionAffected individuals lacking social interaction is one characteristic of autism spectrum disorder (Lord & Bishop, 2010). Many individuals with ASD have trouble engaging in daily social interactions and building relationships with others. Researchers propose that social strategies such as group play and situation simulation may successfully help individuals with ASD improve their social interactions (Channon et al., 2012 and Tomaino et al., 2014).The development of computers continues to make life easier and more convenient. Many tasks can be completed by a computer, such as word processing, communication, providing entertainment, etc. The revolution of computer technology is also indirectly changing education, as computer applications in educational settings become ever broader and more various. There have been many examples of applying computers to education, including pedagogical techniques such as self-learning, computer assisted instruction (CAI), and game-based teaching. The generalization of computers increases the number and type of applications related to learning and teaching in education (Nickerson & Zodhiates, 2013).Individuals with autism are often attracted by computer games which are full of variously visual and auditory stimuli (Mazurek & Wenstrup, 2013). Using computer games to assist children with ASD to learn may be a good strategy due to the possibility that the stimuli provided by such games will increase the children\’s learning motivation. Making good use of interactive computer games to promote language learning, social interaction and cognition for children with autism may lead them to more successfully adapt to social life (Hopkins et al., 2011 and Rahman et al., 2010).Take “whack-a-mole” (TechRadium, 2014) as an example. This popular computer game, shown in Fig. 1, is easy to play and can be enjoyed by members of all generations. When the game starts, moles will come out from holes randomly, a player must move the cursor and click the moles; meanwhile, the tally of successful clicks is automatically recorded and this score is shown on the screen at the same time.Fig. 1. “Whack-a-mole” is a computer game in which players click a mouse to hit moles which come out from holes randomly (TechRadium, 2014).Figure optionsDownload full-size imageDownload as PowerPoint slideThe computer game “whack-a-mole” is designed for a single user to play, and does not allow simultaneous multiple player use due to the Windows Operating System (OS) only supporting one cursor to operate a computer. When multiple mice connect to one computer, interference will occur due to any and all pointing devices connecting up to one cursor.Single Display Groupware (SDG) is a software technology which is proposed to enable co-present users to collaborate through a single, shared display, and allows multiple input devices to be used simultaneously (Stewart, Bederson, & Druin, 1999). Studies have shown that it BEZ 235 is effective to complete a task through a shared computer with multiple users using their own mice (Stanton, Neale, & Bayon, 2002).With the application of the SDG technique, multiple cursors can be displayed on screen to allow multiple users to independently control their own cursors’ movement, and while in a status of co-present collaboration, users can achieve the goal of working/playing together. As shown in Fig. 2, two cursors with different colors and corresponding name prompts, Mark and Jenny, are displayed on screen. Two individuals (Mark and Jenny) each have their own mice which they can use to control their respective cursors simultaneously without interfering with each other.Fig. 2. Two cursors in different colors with name prompts are displayed on screen, and these two cursors can be independently controlled without interference occurring. (For interpretation of the references to color in figure legend, the reader is referred to the web version of the article.)Figure optionsDownload full-size imageDownload as PowerPoint slideIf a child with ASD controls one mouse, and one of his/her peers controls the other one, the two individuals can cooperate with each other to play a game. By virtue of this interaction, a cooperative environment for children with ASD is created, and offering him/her a chance to work/play with other people.Although the SDG technique enables multiple users to more effectively share computing resources by using multiple mice, the SDG function is not included in all legacy applications or existing programs, and it is quite difficult to modify existing applications or programs to incorporate the SDG function as this entails the need for the program to be rewritten (Heimerl, Ramachandran, Pal, Brewer, & Parikh, 2009).To overcome the above issue, a new operating solution was developed in this study—the Multiple Cursor Collaborative Operating Program (MCCOP). MCCOP was applicable to existing applications or programs and included the SDG function to enable individuals to have their own cursors that they can control to function in their respective cursor moving areas on a single display.As shown in Fig. 3, MCCOP enabled the “whack-a-mole” computer game to be operated by two users simultaneously without any modification of the “whack-a-mole” program, and each user could independently operate his/her own mouse in the respective area without interfering with each other. For example, Mark\’s cursor only worked on the left side and Jenny\’s cursor only worked on the right side of the screen.Fig. 3. The computer screen was divided into two areas by MCCOP, where Mark\’s cursor worked in the left area and Jenny\’s cursor worked in the right area.Figure optionsDownload full-size imageDownload as PowerPoint slideThe key feature of MCCOP technology is to redesign mouse drivers so they can intercept and simulate mouse action, and in this way to provide multiple cursors which are displayed on a single computer to allow for multiple users to operate their own mice without interference. Normally, the default functions of a standard device cannot be reset, as when it is connected to a computer, the OS will automatically install the device\’s standard driver. Redesigning a mouse driver can redefine its default functions, however, turning a mouse into a much more powerful tool for use in many applications. However, driver modification has rarely been proposed by researchers due to the complex technological requirements (Microsoft, 2008, Microsoft, 2014, Shih, 2013 and Shih, 2014).Some studies have proposed the adoption of software technology to redesign mouse drivers to improve the computer operation performance of individuals with disabilities. For example, a standard mouse driver was modified to change a mouse wheel into a thumb/finger poke detector to improve computer pointing efficiency for people with multiple disabilities (Shih, Chang, & Shih, 2009). The Multiple Cursor Automatic Pointing Assistive Program (MCAPAP) was another mouse driver modification used to enable two people with disabilities to cooperate in computer pointing performance (Shih, Cheng, Li, Shih, & Chiang, 2010).This work adopted a new mouse driver design to provide the chance for students with ASD to cooperate with their peers, and investigated whether they were able to effectively complete the collaborative pointing task through the MCCOP.2. Methodology2.1. ParticipantsThe four participants in this experiment all studied at the same special education school. In order to conduct the experiment, the participants were selected according to the following principles: (a) two students were to be placed in a group: one participant was a student with ASD, and the other was a student who had developmental disabilities, (b) participants would be able to independently operate a computer mouse to perform mouse operating tasks, (c) participants were capable of understanding the experimental procedures and instructions provided by a research assistant, and (d) the consent of participants’ parents was obtained before commencing the experiment.Participants Wu and Li were in group A, both were 16 years old. Wu was a male with moderate ASD, and Li was a female with moderate intellectual disabilities (ID). Wu liked to use computers, knew how to play his favorite music using the computer, and was able to search websites on the Internet. When he used a computer, he was so intently focused that he would not share the computer or take turns with his peers. He was also in the habit of hitting himself whenever he experienced negative emotions or excessive excitement. A classmate of Wu, Li had a gentle personality and would actively converse with Wu. When Wu exhibited self-injurious behavior, Li would remind him not to do so and would stop him.Cheng and Lu were both 17 years old and comprised in group B. Cheng was a male with severe ASD, and Lu was a male with moderate ID. Cheng insisted in doing things his own way, liked to be alone, and had no interaction with classmates. Lu had introverted personality and did not talk much. Lu and Cheng were classmates.2.2. Apparatus and softwareThe apparatus included two mice and one all-in-one (AIO) computer (ASUS), all of which were set up and ready on a computer table for participants to use. The AIO computer was installed with a built-in Windows OS, collaborative pointing test software (CPTS), and MCCOP software.2.2.1. CPTSCPTS was designed in this experiment to evaluate and record participants’ collaborative pointing performance. Fig. 4 presents a schematic diagram of CPTS. Eight assigned targets (T1–T8) were set in a circle with a radius of 5 cm on the computer screen.Fig. 4. CPTS was designed to evaluate participants’ collaborative pointing performance, and provided eight assigned targets (T1–T8) on the screen for participants to point on and click.Figure optionsDownload full-size imageDownload as PowerPoint slideValid/correct collaborative pointing was achieved when one participant pointed to T1–T4, while the other pointed to T5–T8. When the participant pointed to a target, for example T3, then target T3 would disappear from the screen. Once participants had clicked all their respective targets T1–T8, an instance of successful collaborative pointing was counted and recorded. Each participant only needed to eliminate the four targets in his/her area of responsibility, and these targets could be clicked in any random sequence.Once all targets T1–T8 were clicked and had disappeared from screen, new targets T1–T8 were displayed and both participants were instructed to point at their respective ones again, with this process being repeated until the end of the test time. The number of instances of successfully collaborative pointing within 3 min was automatically recorded by the CPTS.2.2.2. MCCOPThe MCCOP enabled multiple users to control their own cursors to function in their respective cursor moving areas on a single display. In this study, the MCCOP technique divided the computer screen into two areas, where one participant was responsible for operation area 1 (T1–T4), and the other was responsible for operation area 2 (T5–T8), as shown in Fig. 5. Both participants were able to operate their own cursors in their restrictive areas simultaneously, without interfering with each other.Fig. 5. The MCCOP allowed multiple users to operate a single computer simultaneously. One participant was responsible for operation area 1 (T1–T4) and the other was for operation area 2 (T5–T8).Figure optionsDownload full-size imageDownload as PowerPoint slide2.3. Experimental conditionsThe experiment was carried out in an activity room at the participants’ school. Both groups of participants underwent a 3-min experimental session three to five times per day during the study period. Considering radiometric time was the first time for all participants to take part in this kind of experiment and taking into account their attention spans, the experimental session was set for 3 min. An ABAB design was adopted in this study, where A represented a baseline phase and B was an intervention phase (Richards, Taylor, Ramasamy, & Richards, 1999). The number of instances of successful collaborative pointing within 3 min was automatically recorded by the CPTS.2.3.1. Baseline phaseDuring the first baseline phase I, group A underwent 21 sessions and group B underwent 24 sessions. Both groups underwent 18 sessions during the second baseline phase II. In the baseline phase, an AIO computer and two mice were available and in place, but the MCCOP software was turned off for the purpose of recording participants’ baseline performance. Without the MCCOP technique, the operation of two mice would cause mutual interference due to the fact that only one cursor can actually function at one time on the computer. A research assistant would remind participants when necessary that they needed to cooperate with each other to perform the pointing task.2.3.2. Intervention phaseA research assistant explained the functions of the MCCOP and used physical guidance to show participants how to perform a valid collaborative pointing task. Fifty-four sessions for group A and 48 sessions for group B were carried out during the first intervention phase I. Fifty-one sessions for both groups were carried out during the second intervention phase II. In this latter phase, all experimental settings were the same as in the baseline phase, the difference being that the MCCOP functions were activated. With the MCCOP functions, both participants were able to operate their own mouse cursors simultaneously in their respective areas. The MCCOP software allowed multiple cursors to be operated on one computer without interference.2.4. ResultsThe experimental data of group A is shown in Fig. 6. Each data point on the graph represents the mean number of instances of successful collaborative pointing across three sessions. The mean number of instances of collaborative pointing for group A was 0.86 during baseline phase I, and then increased to 21.56 during intervention phase I with the activation of the MCCOP functions. The mean dropped to 0.50 during baseline phase II due to the lack of the MCCOP functions, and then increased again to 25.53 during intervention phase II.Fig. 6. Experimental data of group A. Each data point on the graph represents the mean number of instances of successful collaborative pointing across three sessions.Figure optionsDownload full-size imageDownload as PowerPoint slideThe experimental data of group B is shown in Fig. 7. The mean number of instances of collaborative pointing for group B was 0.38 during baseline phase I, and then rapidly increased to 19.19 during intervention phase I. Similar to the situation for group A, without the MCCOP technique, the mean for group B dropped to 0.50 during baseline phase II, and then increased again to 21.71 during intervention phase II.Fig. 7. Experimental data of group B. Each data point on the graph represents the mean number of instances of successful collaborative pointing across three sessions.Figure optionsDownload full-size imageDownload as PowerPoint slideDuring the baseline phases, for group A, both participants liked to use the computer, and actively moved their mice to the target positions. However, without the MCCOP software, both participants interfered with each other while operating their own mice due to the two mice ending up on one cursor. After experiencing several failures, Wu felt frustrated and kept saying that he did not want to undergo anymore experimental sessions.For group B, both participants moved their mice to the target positions, but they were unable to click on the target due to interference occurring. Finally, Cheng gave up, and stopped holding the mouse, and refused to focus his attention on the computer screen. Therefore, the pointing performance was completed by Lu alone, and not in a collaborative situation.In the intervention phases, the MCCOP was activated to divide the computer screen into two parts corresponding to the left and right hand sides of the screen, respectively. The eight targets were divided into two parts such that one student was responsible for four targets on the left side (T5–T8) and the other student was tasked with the other four targets on right side (T1–T4). Without the mutual interference between cursors, both groups were able to complete collaborative pointing tasks successfully. The emotion of student Wu of group A was more stable during this phase; he was able to pay attention to the computer screen in order to complete pointing tasks, and complained less. Student Cheng of group B was also able to focus on the screen and concentrated in order to operate the mouse.The results show that both groups exhibited low collaborative pointing performance during the baseline phases I and II, whereas, the number of instances of collaborative pointing increased significantly during intervention phases I and II. The difference between the baseline and the intervention was significant (p < 0.01) based on the Kolmogorov–Smirnov test ( Siegel & Castellan, 1988).3. DiscussionIndividuals with ASD often encounter difficulties with daily social interactions and building good relationships with others (Lord & Bishop, 2010). Much emphasis is placed on cooperative relationships between individuals nowadays. If individuals with ASD can effectively increase their level of experience in working with others, doing so should enable them to improve the quality of their interpersonal relationships and better integrate themselves with society.This study provided students with ASD the chance to cooperate with their peers to complete computer mouse collaborative pointing operations through the MCCOP software. The MCCOP was compatible with existing programs and allowed multiple cursors to be controlled at the same time, meaning multiple users could operate one computer simultaneously. In addition, the MCCOP created a cooperative environment for individuals to work/play with other people. Moreover, individuals with ASD had chance to interact with others through performing collaborative pointing operation. For example, student Cheng would use simple words (e.g., hurry!, play!) to ask Lu to continue operating the mouse whenever Lu stopped.The experimental results show that during the baseline phases, the collaborative pointing performance for both groups was not good due to interference occurring. With the intervention of the MCCOP, both groups rapidly increased the mean number of instances of collaborative pointing during the intervention phases. It emerged that the MCCOP made it feasible for participants to perform collaborative pointing through the application of CPTS.Expanding the numbers of participants in each group could be considered in further studies to evaluate the effectiveness of the MCCOP intervention demonstrated in this experiment. In addition, further studies could extend the application of MCCOP to other existing computer games or self-developed software to expand the usage of MCCOP. Furthermore, this study only focused on the collaborative pointing performance of participants, and interaction among participants was not addressed here. Future studies could develop applicable scales to observe and record the interaction among students with ASD and their peers when they are participating in collaborative activities.AcknowledgementThe authors would like to thank the National Science Council, Taiwan, ROC for financially supporting this research under Contract No. NSC 101-2511-S-259-011-MY3.

Demographic, clinical and neuropsychological characteristics.Dementia (n = 49)MCI (n = 57)Controls (n = 50)χ2/F, pPost hocDep + AD (17)aAD only (32)bDep + MCI (26)cMCI only (31)dSex7/1012/2012/1415/16M/F (22/28)0.79, 0.67n.aAge77.0 ± 8.376.1 ± 4.974.0 ± 5.675.0 ± 6.377.2 ± 7.91.26, 0.231 = 2 = 3Education5.4 ± 3.45.0 ± 3.86.2 ± 3.85.9 ± 3.25.1 ± 4.20.87, 0.49n.aCDR1.7 ± 0.61.6 ± 0.51.1 ± 0.41.1 ± 0.40.05 ± 0.2172.6, <0.011 > 2 > 3GDS24.9 ± 2.414.0 ± 3.624.4 ± 3.813.5 ± 4.615.4 ± 8.824.7, < 0.01a > b, c > dK-CERAD K-MMSE17.7 ± 3.518.0 ± 3.422.9 ± 3.321.9 ± 17.725.2 ± 1.750.8, <0.011 < 2 < 3 Verbal fluency7.7 ± 1.27.4 ± 1.09.3 ± 1.49.3 ± 1.39.5 ± 1.517.9, <0.011 < 2 = 3 BNT4.6 ± 2.54.8 ± 1.97.4 ± 1.57.5 ± 1.710.7 ± 6.225.1, <0.011 < 2 < 3 WL-immediate recall8.7 ± 1.110.0 ± 1.511.2 ± 1.211.2 ± 1.412.6 ± 1.640.7, <0.011 < 2 < 3 WL-delayed recall2.7 ± 0.42.8 ± 0.53.4 ± 0.43.4 ± 0.54.4 ± 1.315.2, <0.011 < 2 < 3 WL-recognition6.7 ± 10.46.6 ± 0.47.1 ± 0.47.1 ± 0.47.8 ± 1.06.7, <0.011 < 2 = 3 CP3.7 ± 1.05.3 ± 2.17.5 ± 1.37.5 ± 1.78.9 ± 1.564.3, <0.011 < 2 < 3a < b CP recall0.8 ± 1.10.8 ± 1.13.0 ± 2.73.1 ± 1.63.9 ± 1.343.2, <0.011 < 2 < 3Duncan\’s post hoc test, a: Dep + AD: subjects with depression and Alzheimer\’s disease; b: AD only: patients with Alzheimer\’s disease; c: Dep + MCI: subjects with depression and mild cognitive impairment; d: subjects with mild cognitive impairment; CDR: clinical dementia rating score; GDS: Geriatric Depression Scale; K-CERAD: Korean version of the test battery of the Consortium to Establish c-Myc tag Registry for Alzheimer\’s disease; K-MMSE: Korean version of Mini Mental State Examination; BNT: Boston Naming Test; WL: word list; CP: constructional praxis.Full-size tableTable optionsView in workspaceDownload as CSV3.2. Comparison of gray matter volume between patients with dementia, patients with MCI, and healthy controlsCompared to healthy control group, the dementia group showed decreased gray matter volume in the left middle temporal gyrus (−53, −12, −16, BA21; KE = 14,102 (>400), t = 4.02, PFDR-corr = 0.016), left middle frontal gyrus (−39, 28, 31, BA9; KE = 11,972 (>400), t = 4.77, PFDR-corr = 0.016), left superior frontal gyrus (−9, 12, 55, BA6; KE = 658 (>400), t = 4.77, PFDR-corr = 0.016), right inferior frontal gyrus (45, 10, 24, BA9; KE = 19,273 (>400), t = 4.46, PFDR-corr = 0.016), right cingulate gyrus (6, −31, 32, BA31; KE = 6661 (>400), t = 4.17, PFDR-corr = 0.016), right superior frontal gyrus (56, −41, 21, BA13; KE = 3108 (>400), t = 3.80, PFDR-corr = 0.017), and right parahippocampal gyrus/amygdala (33, 0, −21; KE = 5734 (>400), t = 3.68, PFDR-corr = 0.016).Compared to the MCI group, the dementia group showed decreased gray matter volume in the left medial frontal gyrus (−9, 47, 10, BA10; KE = 1947 (>400), t = 3.95, PFDR-corr = 0.025), left cingulate gyrus (−9, −38, 38, BA3; KE = 1891 (>400), t = 3.78, PFDR-corr = 0.025), right parietal precuneus (9, −32, 48, BA7; KE = 2650 (>400), t = 4.71, PFDR-corr = 0.025), right medial frontal gyrus (13, 47, 11, BA10; KE = 1162 (>400), t = 4.18, PFDR-corr = 0.025), right superior frontal gyrus (24, 9, 54, BA6; KE = 2012 (>400), t = 4.08, PFDR-corr = 0.025), right superior temporal gyrus (54, −41, 16; KE = 1525 (>400), t = 4.03, PFDR-corr = 0.025), and right middle frontal gyrus (29, 48, 17, BA10; KE = 2767 (>400), t = 3.92, PFDR-corr = 0.025). Compared to healthy control group, the MCI group showed decreased gray matter volume in the right parahippocampal gyrus (17, −8, −23, BA34; KE = 843 (>400), t = 2.54, Puncorr = 0.001) ( Fig. 1 and Table 2). There was no significant volume difference between healthy controls with depressive symptoms and healthy controls without depressive symptoms.Fig. 1. Anatomical structures showing significant differences in gray matter volume among dementia, MCI, and healthy control subjects.Figure optionsDownload full-size imageDownload high-quality image (213 K)Download as PowerPoint slideTable 2.

Desired microalgal characteristics, derived from [3••], and possible selective environments for these traitsDesired microalgal characteristicPossible selective process conditions for this characteristicReferenceHigh growth rateBatch mode or chemostat with high dilution rate–High starch productivity‘Survival of the Fattest’ approach[28•]High lipid productivity‘Survival of the Fattest’ combined with, for example, selection on low density cells[21•]Staying in suspensionPeriodic removal of biofilms–Being easily separable from the liquid phaseProvide competitive advantage to algae that form Dibucaine through inclusion of a settling period[37]Tolerating marine conditionsCultivation under marine conditions–High shear toleranceCultivation under high shear conditions–Low sensitivity of Rubisco to high O2 concentrationsCultivation under high oxygen concentration–Tolerating broad temperature rangeCultivation under fluctuating temperature–No lipid catabolism––Specific lipid composition for biodiesel production––Full-size tableTable optionsView in workspaceDownload as CSVA desirable microalgal characteristic other than storage compound productivity is being easily harvestable. Biomass concentrations during phototrophic cultivation are typically 0.1–4.0 g/L [25]. In chemotrophic processes biomass concentrations of 100 g/L can be achieved. Therefore, the solid–liquid separation after microalgal cultivation is a costly step. From this point of view, the characteristic of fast-settling, and therefore easily separable biomass, would be beneficial. Fast-settling granules can be enriched for in bacterial systems by selectively removing non-settling biomass [38]. A comparable approach has been used in mixed cultures of microalgae and bacteria, yielding microalgal-bacterial flocs [37].Some desired characteristics cannot be obtained using a selective environment, since they do not give a competitive advantage to the microalgae. Synthetic biology approaches can be used to create species displaying these characteristics. Examples of functionalities that are obtained using microalgal engineering are the inhibition of lipid catabolism [32] and having a specific lipid composition which is more suitable for biodiesel production [7]. Cultivation of these strains will face the challenges of competition and strain degeneration. As such, the advantages of the desired characteristic should outweigh the cultivation efforts.For most applications a combination of the desired characteristics is preferred. By combining selective process conditions, multiple desired characteristics can be obtained. Running an enrichment culture under high oxygen concentration while removing biofilm will select for microalgae with a high oxygen tolerance that stay in suspension. Other desired characteristics might be incompatible, such as the characteristic of being easily harvestable with the trait of staying in suspension. Finally it should be realized that numerous, unintentional selective environments are applied to everyday microalgal cultivation. In many processes part of the biomass is harvested and the non-harvested part is retained in the reactor. This creates a clear selective advantage for the characteristic of being non-harvestable [21•].Conclusion and outlookLarge-scale microalgal cultivation for low value products will take place in large, open ponds. Both competition and strain degeneration are a threat to stable storage compound production in these systems. Working at extreme conditions, in closed systems or with herbicides could solve the competition aspect, but the problem of strain degeneration is not addressed in these approaches. Operating a cultivation system with large inocula of the desired strain diminishes the effect of both threats but only for a limited time and at considerable cost. We agree that ecological principles should be the basis for improving microalgal cultivation [11, 12??, 14?, 15 and 16?] and therefore advocate an ecology-based solution in which both competition and evolution are improving the process. This is achieved by creating a selective environment in which storage compound production is rewarded by directly linking it to growth. A similar approach can be applied to enrich other desirable characteristics. With a clear list of desired traits and a vast, unexplored microalgal diversity we encourage and endorse the further exploration of ecology-based selective environments in future algal research.References and recommended readingPapers of particular interest, published within the period of review, have been highlighted as:• of special interest•• of outstanding interestAcknowledgementsThis research was funded by the Dutch Foundation for Technical Sciences (STW) as project 11610 in the Waste 2 Resource (W2R) programme. The authors are grateful to the anonymous reviewers, whose comments improved this manuscript greatly.

Mechanical properties of various materials.SamplesCompressive module (GPa)Compressive strength (MPa)Dense bone7–30100–230Spongy bone0.05–0.52–12Col–HA/pectin0.85 ± 0.023*6.36 ± 0.475**Col–HA0.69 ± 0.0244.46 ± 0.694*p < 0.01.**p < 0.05.Full-size tableTable optionsView in workspaceDownload as CSVTable 2 shows that, the compressive modulus of synthesized composite regorafenib between cancellous bone and cortical bone, having certain support capacity, and due to the addition of pectin, the mechanical properties of the material is improved.2.5. HydrophilicityThe hydrophilicity of collagen–hydroxyapatite/pectin and collagen–hydroxyapatite composite is shown in Table 3. After soaking, water absorption rate of the composite material were 36.9% and 40.1%, p < 0.01. As can be seen from the above results addition of pectin, water absorption rate of the composites decreased significantly. This is probably because, by the Maillard reaction between the polysaccharide and the protein [26], the combination achieve between collagen and hydroxyapatite, pectin between the components more tightly, reducing the space available for the liquid into the composite reduces the polar group of material surface. On the other hand, the number of polar groups is also an important factor affecting the material hydrophilicity [15]. Thus water absorption rate of collagen–hydroxyapatite/pectin than that of collagen–hydroxyapatite is lower, namely its swelling decreased.Table 3.

Carbon nitride nanotube; Carbon nanotube; CoPt nanoparticle; NADH; Electrochemical oxidation1. IntroductionCarbon nanotubes (CNTs) with a high electron transfer rate and surface area [1] and [2] have resulted in appreciable technological advances in the field of electrochemical research. Controlled modification [3], [4] and [5] of CNTs to incorporate functional metal nanoparticles (NPs) is crucial to realize advanced performance of the CNTs, in association with their thermal conductivity and catalytic activity. Infusion of many different NPs into CNTs has been possible primarily through exploitation of the physical adsorption and covalent bonding properties of the NTs. For example, modification of the surface of CNTs with AZD5363 treatment facilitates bonding to positively charged NPs either by covalent or electrostatic interactions between the acid-derived carboxylic acid terminals and the positively charged NPs [6] and [7]. However, surface modifications may compromise the stability of NPs and/or the electronic conductivity of the CNTs.Recently, nitrogen-doping methods have been used to enhance the properties of CNTs [8] and [9] and have yielded CNTs with dramatically improved stability compared to that of pristine CNTs, because nitrogen incorporation in an amorphous carbon network reduces the internal stress. In particular, a large number of nitrogen sites in carbon nitride nanotubes (CNNTs) may act as active sites for nucleation of the NPs and help to anchor the particles on the surfaces of the nanotubes, thus promoting a uniform distribution of the NPs. Moreover, it has been reported that CNT-based nanotubes modified with bimetallic NPs, such as CoPt, FePt, RuPt and PdPt, exhibited better electrochemical properties than their single-metal component [10], [11], [12] and [13]. Meanwhile, the catalytic activity of Pt NPs can be enhanced by hybridization of transition metals such as Fe, Co, or Ni [14] and [15]. Specifically, bimetallic CoPt NPs are believed to offer unprecedented benefits in catalysis by tuning the durability of the catalysts [16]. Furthermore, it has been shown that high catalytic power and increased enzymatic redox system stability can be obtained using graphene.[17] and [18] To this end, when CNNTs are modified with CoPt bimetallic NPs without harsh treatment, they may interact synergistically to form an enzymatic redox system with enhanced activity due to an increased surface area, excellent catalytic activity, enhanced conductivity, and high stability [11] and [19].The electrochemical sensing of reduced nicotinamide adenine dinucleotide (NADH), composed of the coenzyme redox pair of NAD+ and NADH, is of great importance in that it plays an essential role in biological catabolism [20]. However, the direct electro oxidation of NADH at the electrode surface is highly irreversible and requires a large activation energy [21] and [22], resulting in various side reactions that degrade the overall performance of the electrode [23]. To address these issues, the electrodes have been modified by a variety of compounds to facilitate NADH oxidation [21], [22], [24] and [25]. Here, we report for the first time, a facile in-situ method to fabricate catalytically active CoPt NP-decorated CNNT hybrids (CoPt-CNNT hybrids) and the application of these hybrids as electrochemical detection sensors for NADH. Moreover, we have elucidated the binding mechanism and the electron density distribution of CoPt-CNNT hybrids through first principle density functional theory level calculations. Our experimental and theoretical results indicate that CoPt-CNNT hybrids are very stable and have a high electrochemical sensing capability of NADH.2. Experimental2.1. MaterialsChloroplatinic acid hexahydrate (H2PtCl6·6H2O, 99%), Potassium ferricyanide(III) (K3Fe(CN)6, 99%) and Tris(2,2′-bipyridyl)dichloro ruthenium(II) hexahydrate (C30H24Cl2N6Ru·6H2O, 99.95%) was purchased from Sigma–Aldrich and used as-received. Ethylene glycol (C2H6O2, 99.5%), cobalt (II) acetate tetrahydrate ((CH3COO)2Co·4H2O, 99%) and sodium hydroxide (NaOH, 97%) were received from Junsei chemical and used as-received.2.2. Carbon nitride nanotubes synthesisCNNTs were synthesized by microwave plasma-enhanced chemical vapor deposition (MPECVD). To prepare the catalysts, 7 nm-thick iron film was deposited on Si-substrate with a thin oxide layer by RF magnetron sputtering at 100 W RF power. Pressure was maintained at 15 mTorr using Ar gas. The thickness of SiOx on the Si wafer was 500 nm. The substrate was then placed in a MPECVD chamber, which was evacuated to about 0.1 Torr. The substrate was then heated to 550 °C in the chamber at 0.1 Torr pressure. A flow of N2 gas was maintained within the chamber at 85 sccm and the substrate was treated with N2 plasma at a microwave power of 700 W for 1 min. Following this, 15 sccm of CH4 gas and N2 gas were introduced at 700 °C simultaneously at a microwave power of 700 W. The growth time was 20 min. The outer diameter of the as-produced CNNTs ranged from 5 to 25 nm.2.3. CoPt-decorated carbon nitride nanotubesCoPt-decorated CNNTs were synthesized by a microwave heating method with cobalt and platinum precursors dissolved in ethylene glycol. In a typical procedure, pristine CNNT was dispersed in a vial with 50 mL of ethylene glycol under sonication. Then, 10 mM of chloroplatinic acid and 10 mM of cobalt acetate tetrahydrate were mixed into the solution. For low and high loading of NPs on CNNTs, 1 and 3 ml of metal precursors were used, respectively. Finally, 0.5 M of sodium hydroxide was added to the solution. The vial was heated for 90 s using 700 W of microwave irradiation. Black products were separated by centrifugation at 6000 rpm for 50 min, washed several times with acetone, and dried at 60 °C for 8 h under vacuum.2.4. Modified electrodesWorking electrodes were prepared by drop-casting a nanotube solution, followed by evaporation of the solvent to dryness. Typically, the required amounts of CNTs, CNNTs, and CoPt-CNNTs were dispersed in deionized water containing 0.5% Nafion solution and sonicated for 60 min. The well-dispersed nanotube solution was then drop-cast on fluorine-doped tin oxide (FTO) substrate and the solvent was evaporated to dryness at room temperature. These modified electrodes served as the working electrodes in the experiments.2.5. Characterization techniquesCNTs, CNNTs, and CoPt-CNNT hybrids were characterized by field emission transmission electron microscopy (FE-TEM, FEI, Tecnai F20), energy dispersive X-ray spectroscopy (EDS, EDAX, Genesis Apex System), and X-ray diffraction analysis (XRD, Rigaku D/MAXRINT-2000). Raman spectroscopy (Raman, Horiba Jobin Yvon, LabRAM HR UV/Vis/NIR) was performed with an Ar ion laser excitation at 514.5 nm for crystallinity analysis. X-ray photoelectron spectroscopy (XPS, Thermo VG Scientific, Sigma Probe) was performed to investigate the surface chemical state of the materials. Furthermore, electrochemical analysis was done using a three-electrode system in which a modified FTO electrode was used as the working electrode, Ag/AgCl electrode was used as the reference electrode, and Pt foil was used as the counter electrode. Phosphate buffered saline (PBS) solution (0.1 M; pH 7.4) was used as the electrolyte for all the electrochemical measurements. Cyclic voltammetry was carried out using an advanced potentiostat (PGSTAT-30 from Autolab) with a scanning voltage in the range of −1 V to +1.5 V. The impedance measurement (PGSTAT-30 from Autolab) was performed in presence of 5 mM [Ru(bpy)3]2+ in 0.1 M PBS as the supporting electrolyte at an applied potential of +1.2 V (open circuit potential). Unless otherwise stated, 0.1 M phosphate buffer solution (pH 7.4) was used as the supporting electrolyte.2.6. CalculationsNitrogen atoms were intercalated into (5, 5) and (8, 0) single-walled carbon nanotubes to yield nanotubes with C59N1 and C63N1 for graphite-like nitrogen (GN), and C56N3 and C60N3 nanotubes for pyridine-like nitrogen (PN). Structure optimization and energy calculations were performed with the PW91 method for generalized gradient approximation (GGA) and the plane wave model using CASTEP software. A supercell of 20 Å × 20 Å × c was used for the calculations, where c represents the length of a nanotube along the tube axis. All atoms were described using Vanderbilt ultrasoft pseudo potentials and a cutoff energy of 240 eV, where the set of k-points used to expand the electronic wave function were based on the Monkhorst-Pack scheme within 5.0 × 10−5 eV atom−1 of total energy convergence. The electron density was investigated for optimized geometries. The binding energies of atoms on the nanotubes were calculated using the following equation:equation(1)Eb=Enanotube+atom−(Enanotube+Eatom)Eb=Enanotube+atom−(Enanotube+Eatom)3. Results and discussionCoPt-CNNT hybrids were fabricated via a polyol method. This process facilitated anchoring of CoPt nanoparticles on the nanotube surface through direct nitrogen mediation, preserving the conductivity of the CNNT. Low and high magnification transmission electron microscopy (TEM) images of pristine CNTs ( Fig. 1a and e), CNNTs ( Fig. 1b and f), low loading CoPt-CNNT hybrids ( Fig. 1c and g), and high loading CoPt-CNNT hybrids ( Fig. 1d and h) are shown in Fig. 1. Pristine nanotubes ( Fig. 1a) had a multi-walled configuration with a mean diameter of about 15 nm. The CNNTs had bamboo-like structures in the core, which are attributed to nitrogen doping-induced stress; the mean distance between the bamboo-like compartments was 2.5 ± 1 nm ( Fig. 1b). CoPt NPs (2 nm in diameter) were found to be homogeneously distributed on the nanotubes without any agglomeration ( Figs. 1g and h). Energy dispersive X-ray spectroscopy (EDS) confirmed an average stoichiometry of Co40Pt60 of the NPs ( Fig. 2a). In addition, X-ray diffraction (XRD) patterns of CNNTs (black) and CoPt-CNNT hybrids (Fig. S1) revealed a predominant peak at 26.1° for both the CNNT and the CoPt-CNNT hybrids, matching the (0 0 2) plane of graphite structure [21]. It is noteworthy that no peaks for cobalt or cobalt oxide were observed in the XRD patterns, implying that only bimetallic CoPt NPs were formed on the surfaces of the CNNTs (Fig. S1) [26]. Furthermore, the two peaks observed between 40 and 50° correspond to the (1 0 1) and (1 1 0) planes of the face-centered cubic (FCC) structure of Pt or bimetallic CoPt. A slight shift in the peak values was attributed to lattice contraction due to substitution of larger Pt atoms with smaller Co atoms [27] and [28].Fig. 1. Transmission electron microscopic images at low (a–d) and high (e–h) magnifications showing CNTs, CNNT, and low-loading CoPt NP-functionalized CNNTs and high-loading CoPt NP-functionalized CNNTs, respectively.Figure optionsDownload full-size imageDownload as PowerPoint slideFig. 2. (a) Energy dispersive X-ray spectra and (b) Raman spectra showing the D band and G-band of CNNTs (black line) and CoPt-CNNTs (red line); deconvoluted X-ray photoelectron spectra for CoPt-CNNTs showing the core level spectrum of (c) Co 2p and (d) Pt 4f.Figure optionsDownload full-size imageDownload as PowerPoint slideFig. 2b shows the Raman spectra for CNNTs and the CoPt-CNNT hybrids. The graphitic (G) peak and the disordered (D) peak for CNNTs were found at 1347.8 cm−1 and 1576.9 cm−1, respectively. By contrast, both D- and G- bands of CNNTs showed blue shifts to a higher wavenumber (11 cm−1 and 5 cm−1 shift respectively) as a result of CoPt NP decoration of the outer surfaces of the CNNTs. Furthermore, the peak intensities of both D- and G- bands were found to be enhanced with respect to that of the D*-band, while the D/G band ratio remained almost constant [29]. Fig. 2c and d show the X-ray photoemission spectroscopic core level spectra of Co-2p and Pt-4f, respectively. The de-convoluted core level line of the Co-2p spectrum had a photoemission peak at 781 eV with a strong asymmetric tail toward higher binding energy. The observed binding energy was slightly higher than the metal-like state observed at 778.19 eV for Co clusters [30]. This discrepancy in the peak value might be due to the alloying effect observed in CoPt bimetallic alloy NPs [31]. The core level spectrum of Pt-4f had a distinct peak at 71.29 eV, which shifted to a higher binding energy compared to that in pure Pt due to loss of electrons on the formation of the CoPt alloy [30] and [32].To understand the binding mechanism of the in-situ growth of CoPt NPs on CNNTs, we determined the binding energies of Co and Pt to CNNTs. The binding characteristics of Co and Pt were determined using the first principle density functional theory. Fig. 3(a–d) shows the optimized geometries of Co and Pt-anchored CNNTs with chirality (5, 5) and chirality (8, 0), respectively. The binding energies of Co and Pt atoms in CoPt-CNNT hybrids were estimated at three different sites, namely, (a) pyridine-like nitrogen (PN), (b) graphite-like nitrogen [29], and (c) pristine CNTs (CNT) for both metallic and semiconducting CNNTs, and are tabulated in Table 1. The binding energies for adsorbed Co and Pt atoms with PN were −5.20 eV and −2.46 eV, respectively, on the surfaces of metallic CNNTs with chirality (5, 5). However, Co and Pt atoms had lower binding energies to the GN and CNT sites of metallic CNNTs. It was also found that, although Pt atoms had a binding energy of −2.43 eV to a semiconducting CNNT surface with chirality (8, 0), the binding energy of atomic Co was much higher at −5.52 eV. Furthermore, the binding energies of both Co and Pt atoms to pristine CNTs were very weak. These results suggest that pyridine-like nitrogen (PN) sites function as the actual anchoring sites for Co and Pt atoms on the CNNT surfaces. Additionally, density functional calculations of binding energies demonstrated that binding of Co atoms to CNNTs was more energetically favorable than that of the Pt atoms to the CNNTs. This suggests that Pt atoms are more likely to bind to the Co atoms that are already bound to the surfaces of CNNTs. The data shown in Fig. 3 indicate that the electron density isosurfaces between Co/Pt atoms and CNNTs are convoluted.Fig. 3. First principle-based optimized geometries simulated for Co and Pt atoms adsorbed on (a, b) metallic and (c, d) semiconducting CNNTs. Electron density distributions for Co and Pt atoms adsorbed on metallic and semiconducting (e-h) CNTs and (i-l) CNNTs, respectively.Figure optionsDownload full-size imageDownload as PowerPoint slideTable 1.

 The extraction yields of four flavonoids from Radix Scutellariae (mg/g) by MAE with NADESs and conventional solvents.Help with DOC filesOptionsDownload file (41 K)To visualize these differences, multivariate data analysis, principal components analysis (PCA), was used, and the analysis results were shown in Fig. 2. Two principal components (PC1 and PC2) were obtained, and could explain 68.1% and 17.8% of the variation of the data, respectively. Each point on the loading plot (Fig. 2A) represented the contribution of each variable (target flavonoid) to the score, while each point on the score plot (Fig. 2B) represented each tested solvent. As observed in Fig. 2A, PC1 was correlated well with the yields of wogonoside, baicalein and wogonin with the loadings of 0.814, 0.942, and 0.865, respectively, while PC2 was mainly correlated with the yield of baicalin with the loading of 0.738. The PCA confirmed the similarity between CCLA, CCEG, and 60% ethanol, and these solvents were efficient for the extraction of wogonoside, baicalein and wogonin. Especially, CCLA was located in the rightmost position of PC 1, indicating its higher extraction efficiencies for wogonoside, baicalein and wogonin. CCGly was placed in the uppermost position of PC 2, indicating its higher extraction efficiency for baicalin. However, on the whole, in the loading plot of PCA (Fig. 2A), four target flavonoids were clustered around CCLA, indicating that this solvent were more suitable for the simultaneous extraction of these compounds. So, CCLA was selected for the extraction of flavonoids from Radix Scutellariae.Fig. 2. Loading plot (A) and score plot (B) of principal component analysis of the extracts from Radix Scutellariae with NADESs and 60% ethanol.Figure optionsDownload full-size imageDownload as PowerPoint slide3.2. Effect of the choline chloride/lactic MPI-0479605 ratioThe NADESs with different choline chloride/lactic acid ratios were evaluated for the extraction of four target flavonoids. As seen from Fig. 3A, the extraction yields of these flavonoids were found to change significantly when choline chloride/lactic acid ratio was decreased from 3/1 to 1/4 (mol/mol). The amounts of the extracted target flavonoids increased with the decrease of choline chloride/lactic acid ratio from 3/1 to 1/2 (mol/mol). However, the extraction yields of four target flavonoids kept sustained decrease with the change of choline chloride/lactic acid ratio from 1/2 to 1/4. On the basis of the above results, a choline chloride/maltose ratio of 1/2 (mol/mol) was chosen for the following tests.Fig. 3. Extraction yields of baicalin, wogonoside, baicalein and wogonin using the NADESs with different choline chloride/lactic acid ratios (A) and different concentration of water in NADES (B). aExtraction yields of baicalin; bextraction yields of wogonoside, baicalein and wogonin.Figure optionsDownload full-size imageDownload as PowerPoint slide3.3. Effect of concentration of water in CCLAThe extractions were carried out with different concentrations of water in CCLA (from 0% to 80%, v/v). The addition of water in NADES can cause the decrease of solvent viscosity, which is beneficial to the mass transport from plant matrices to solution. Fig. 3B showed that the extraction yields of baicalin, wogonoside, baicalein and wogonin from Radix Scutellariae changed significantly as the concentration of water in CCLA increased. The addition of 20% water in CCLA markedly improved the extraction process. However, higher concentration of water in CCLA (40–80%) led to the decrease in the amounts of four target flavonoids extracted. This was probably because higher concentration of water decreased the interactions between CCLA and flavonoids, and increased the polarity of the solvent mixture. So, a concentration of 20% (v/v) water in CCLA was selected in subsequent BBD experiments.3.4. Optimization of NADES-MAE using BBDFurther optimization of the NADES-MAE conditions (extraction temperature, solvent to solid ratio and extraction time) was performed by BBD. The results of BBD experiments were shown in Table 2. These data were analyzed using Design Expert 8.0.5 software for second-order polynomial regression analysis and statistical analysis of variance (ANOVA). This mathematical regression model for baicalin, wogonoside, baicalein and wogonin were shown below in terms of coded levels:Ybaicalin=31.80+2.28X1+1.68X2+3.23X3−0.89X1X2−1.21X1X3−0.70X2X3−3.45X12−0.56X22−4.95X32Ywogonoside=8.03+0.65X1+0.42X2+1.18X3−0.26X1X2−0.24X1X3−0.14X2X3−0.63X12−0.13X22−1.12X32Ybaicalein=8.69+1.04X1+0.45X2+1.09X3−0.11X1X2+0.19X1X3−0.19X2X3−1.22X12−0.12X22−1.30X32Ywogonin=1.61+0.13X1+0.081X2+0.22X3−0.043X1X2−0.027X1X3−0.015X2X3−0.15X12−0.029X22−0.24X32where Y was the extraction yield of target compound (mg/g); X1, X2 and X3 represented extraction temperature, solvent to solid ratio and extraction time, respectively.Table 2.