Quantification of Pseudomonas aeruginosa in multispecies biofilms using PMA-qPCR

View article
PeerJ

Introduction

Specific quantification of the different members in a multispecies biofilm is a challenging task. Cultivation-based approaches are time-consuming and can lead to an underestimation of cell numbers due to the presence of viable but nonculturable bacteria (VBNC). VBNC bacteria will not grow on routinely-used microbiological media, but are nevertheless still viable and potentially virulent (Li et al., 2014). A promising alternative for cultivation-based methods is quantification based on qPCR. However, a major drawback of qPCR-based quantification is the overestimation of cell numbers due to the presence of extracellular DNA and DNA originating from dead cells, and adjustments are required to differentiate between viable and dead bacteria. Treatment of bacterial samples with propidium monoazide (PMA) prior to DNA extraction has been proposed as an effective method to avoid the detection of extracellular DNA and DNA from dead cells (Alvarez et al., 2013; Kruger et al., 2014; Yasunaga et al., 2013). PMA only enters membrane-compromised cells, and once inside the cell, it intercalates into DNA between the bases (one PMA molecule per 4 to 5 base pairs DNA, with little or no sequence preference). Besides intercalating into DNA of membrane-compromised cells, PMA can also intercalate into extracellular DNA (Nocker, Sossa & Camper, 2007; Waring, 1965). After exposure to strong visible light, the photoreactive azido group of PMA is converted to a reactive nitrene radical. This nitrene radical forms a stable covalent nitrogen-carbon bond with the DNA, resulting in permanent DNA modification. The modified DNA is then lost together with cells debris during genomic DNA extraction and will not be amplified during the qPCR reaction (Nocker, Cheung & Camper, 2006; Nocker et al., 2009). Excess PMA is inactivated by reaction with water molecules in solution, prior to DNA extraction, and thus will not affect the DNA from viable cells after cell lysis (Nocker, Cheung & Camper, 2006). Nevertheless, the use of PMA has some limitations. The discrimination between viable and dead cells is only based on membrane integrity, and the effect of antimicrobial therapies that do not target the cell membrane can thus not be monitored using PMA (Nocker & Camper, 2009). Secondly, viable cells with a slightly damaged cell membrane will not be quantified (Strauber & Muller, 2010) and the presence of a high number of dead cells (>104 cells/ml) can affect the quantification of viable cells (Fittipaldi, Nocker & Codony, 2012). Finally, the presence of other compounds in the sample, e.g., environmental compounds that can bind to PMA, can subsequently prevent PMA to bind to DNA (Taylor, Bentham & Ross, 2014).

In order to determine whether there are differences in antimicrobial susceptibility of cells grown in mono-or multispecies biofilms (Dalton et al., 2011; Lopes et al., 2012), accurate quantification of the various members of these biofilms is required. In the present study, we evaluated the use of PMA-qPCR to quantify Pseudomonas aeruginosa in mono- and multispecies biofilms following exposure to various antibiotics, used to treat respiratory infections in cystic fibrosis (CF).

Materials and methods

Bacterial strains

P. aeruginosa ATCC9027, Staphylococcus aureus LMG10147, Burkholderia cenocepacia LMG16656, and Streptococcus anginosus LMG14502 were cultured overnight at 37 °C in Brain Heart Infusion broth (BHI) (Oxoid, Basingstoke, UK).

Antibiotic solutions

The difference in susceptibility of P. aeruginosa in a mono- and multispecies biofilm towards colistin (Sigma-Aldrich, Diegem, Belgium), tobramycin (TCI Europe, Zwijndrecht, Belgium), and levofloxacin (Sigma-Aldrich, Diegem, Belgium) was determined. The concentration used for colistin and tobramycin was 200 µg/ml, for levofloxacin 100 µg/ml. These concentrations were based on the levels achievable in CF sputum by inhalation therapy (Wu et al., 2013). All antibiotic solutions were prepared in physiological saline (0.9% [w/v] NaCl) (PS) (Keltner et al., 2014). Minimal inhibitory concentrations (Nusbaum et al., 2012) of colistin, tobramycin and levofloxacin were determined in duplicate according to the EUCAST broth microdilution protocol using flat-bottom 96-well microtiter plates (TPP, Trasadingen, Switzerland) as previously described (Peeters, Nelis & Coenye, 2008).

Quantification of P. aeruginosa cells in monospecies and multispecies biofilms

Formation of P. aeruginosa monospecies and multispecies biofilms

For formation of mono- and multispecies biofilms, round-bottomed 96-well microtiter plates (TPP) were used. Inoculum suspensions containing approximately 106 CFU/ml of P. aeruginosa alone or 106 CFU/ml of P. aeruginosa in combination with 106 CFU/ml of S. aureus, 107 CFU/ml of B. cenocepacia and 107 CFU/ml of S. anginosus, were made in BHI. The inoculum cell numbers were based on preliminary optimization experiments, and led to biofilms with the highest cell numbers (data not shown). BHI was supplemented with 5% (w/v) bovine serum albumin (BSA) (Kart et al., 2014), 0.5% (w/v) mucine type II, and 0.3% (w/v) agar. Mucine and agar were added to mimic the composition of CF sputum and to increase the medium viscosity, respectively. Sterile medium served as blank and was included on each plate. After 4 h of adhesion at 37 °C, wells were rinsed with 100 µl PS to remove non-adhered cells. A 100 µl amount of fresh medium was added to the wells and the plates were incubated for an additional 20 h. After 20 h, the supernatant was again removed, each well was rinsed using 100 µl PS and 100 µl of the antibiotic solution (colistin, tobramycin or levofloxacin) was added to the mature biofilms. To the wells of the control biofilm plate, 100 µl PS was added. The plates were then again incubated at 37°C for 24 h. For each test condition, 72 technical replicates were included. All experiments were performed on three different occasions. Confocal imaging was performed as described in Udine et al. (2013). The control plate cell numbers of P. aeruginosa, S. aureus and B. cenocepacia, respectively, determined on cetrimide agar, mannitol salt agar and tryptic soy agar supplemented with tobramycin (4 mg/ml) and nitrofurantoin (25 mg/ml), increased after 24 h, respectively with 1.30, 0.67 and 0.95 log cfu/biofilm, indicating that these species are actually growing in the multispecies biofilm. The control cell number of S. anginosus, determined on Mc Kay agar (Sibley et al., 2010) did not change, indicating that this species was continuously present in the multispecies biofilm. Viable but nonculturable cells of any species might not be detectable by plate counts, which might lead to differences between plate counts and PMA-qPCR.

Propidium monoazide cross-linking

After 24 h of experimental administration of antibiotics to the biofilm, the antibiotic solution in the test plate and the PS in the control plate were removed. The wells were rinsed with 100 µl PS. Next, biofilms were detached by vortexing (900 rpm) and sonication (both 5 min), followed by collection of the content of the wells in a sterile tube. The vortexing and sonication step was repeated after the addition of 100 µl PS to each well. The sterile tube was centrifuged (5 min at 3000 × g), and the pellet was resuspended in 1.5 ml of PS. For each treatment, 2 wells of a 24-well plate were filled with 600 µl of the cell suspension. 1.5 µl of a 20 mM PMA solution in dH2O (Biotium, Inc., California, USA) was added to the first well (final concentration of 50 µM). To the second well, 1.5 µl of MilliQ water (MQ water) (Millipore, Billerica, Massachusetts, USA) was added. The plates were vortexed (5 min, 300 rpm, room temperature) in the dark and exposed to light for 10 min, using a LED-lamp (Dark Reader transilluminator, Clare Chemical Research, Dolores, Colorado, USA) (output wavelength 465–475 nm) (Deschaght et al., 2013).

Effect of PMA on P. aeruginosa cell viability

To analyze the effect of 50 µM PMA on cell viability, overnight grown planktonic P. aeruginosa cells (OD600 = 1.0) were used. To 2 mL of this culture, 5 µl of a 20 mM PMA solution in dH2O was added (final concentration of 50 µM). As a control, 5 µl of MQ water was added instead of the PMA solution. The plates were incubated in the dark (5 min, 300 rpm) and exposed to light for 10 min. Next, the cell numbers of control and test group were determined via the plate count method (on tryptic soy agar) and by solid-phase cytometry (SPC) (ChemScan RDI; AES-Chemunex, Ivry-sur Seine, France), as described previously (Vanhee et al., 2010).

Extraction of genomic DNA

After incubation with PMA, 500 µl of cell suspension from each well was transferred to a sterile Eppendorf tube. The samples were centrifuged (10 min, 13.000 rpm) and DNA from Gram-negative organisms was extracted as described previously (Pitcher, Saunders & Owen, 1989). Briefly, the pellets were washed with 500 µl RS-buffer (0.15M NaCl [Sigma-Aldrich, Diegem, Belgium], 0.01 M EDTA [VWR, Leuven, Belgium], pH 8.0) and resuspended in TE-buffer (1 mM EDTA, 10 mM Tris–HCl [Sigma-Aldrich, Diegem, Belgium]). 500 µl GES-buffer (60% [w/v] guanidium thiocyanate [Sigma-Aldrich, Diegem, Belgium], 0.5 M EDTA, pH 8.0, 1% [w/v] sarkosyl [Sigma-Aldrich, Diegem, Belgium]) was added and the samples were placed on ice for 10 min. After the addition of 250 µl cold ammonium acetate (7.5 M [VWR, Leuven, Belgium]), the samples were placed back on ice for 10 min. Subsequently, 500 µl cold chloroform/isoamylalcohol (24:1) (Roth, Karlsruhe, Germany) was added. Samples were mixed thoroughly and centrifuged for 20 min at 13.000 rpm. Supernatant was then collected in a new tube and 0.54 volumes cold isopropanol (Sigma-Aldrich, Diegem, Belgium) were added to precipitate the DNA. Samples were then centrifuged (10 min, 13.000 rpm), and supernatant was removed. A 150 µl amount of ethanol (Sigma-Aldrich, Diegem, Belgium) (70% [v/v]) was added and samples were centrifuged for 1 min. This step was repeated. The DNA pellet was air-dried, 30 µl TE-buffer was added, the samples were placed at 4° C for 24 h and were then treated with RNase.

Following electrophoresis on 1% agarose gels, genomic DNA was visualized with GelRed (GelRed nucleic acid, Biotium, Hayward, California, USA), and the genomic DNA concentration was measured with Quantifluor dsDNA kit (Promega, Madison, Wisconsin, USA).

qPCR

Real-time PCR (CFX96 Real Time System; Bio-Rad, Hercules, CA, USA) was carried out with the PerfeCTa SYBR Green FastMix (Quanta Biosciences). Species-specific primer sequences for gyrB of P. aeruginosa were designed using primer-BLAST (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) using P. aeruginosa and B. cenocepacia sequences obtained from GenBank. The forward primer and the reverse primer were 5′-GGTGTTCGAGGTGGTGGATA-3′ and 5′-TGGTGATGCTGATTTCGCTG-3′, respectively. The specificity of the primers was evaluated by melting curve analysis.

To generate a standard curve, DNA extracted from serially-diluted and PMA-treated planktonic P. aeruginosa cultures was used for qPCR. The Cq-values obtained were plotted against the number of viable cells determined by SPC. The serial dilutions were prepared from a P. aeruginosa overnight suspension (OD 0.1). Cells were diluted from 109 CFU/ml to 104 CFU/ml in PS. Six independent biological repeats were included.

Effect of PMA on Cq-values of defined ratios of viable and dead P. aeruginosa cells

Planktonic P. aeruginosa cells (OD 1.0) were killed by heating for 15 min at 95°C. Complete killing was confirmed by SPC. Mixtures of viable and dead cells were prepared, in which viable cells represented 0%, 0.1%, 1%, 10%, 50%, 75% and 100% of the total population. Four wells of a 24-well plate were filled with 600 µl of each mixture. PMA was added to 2 wells and MQ water was added to the other 2 wells (PMA-negative control). Cells were then treated as described above (2.3.2.) and Cq-values were determined via qPCR. Six independent biological repeats were carried out.

Statistical data analysis

Statistical data analysis was performed using SPSS software, version 22 (SPSS, Chicago, Illinois, USA). The normal distribution of the data was verified using the Shapiro–Wilk test. Non-normally distributed data were analyzed using a Mann–Whitney test. Normally distributed data were analyzed using an independent sample t-test. Differences with a p-value <0.05 were considered as significant.

Results and discussion

Optimization of the PMA-qPCR

Treatment with PMA (50 µM) did not affect the number of viable cells as determined via SPC (Fig. 1), so it can be concluded that PMA itself has no inhibitory effect on P. aeruginosa. Therefore, all experiments were conducted with a PMA concentration of 50 µM. Increasing the fraction of viable cells in the mixture led to an increase in the genomic DNA yield after PMA treatment (Figs. 2A, 2B and 3). As shown in Fig. 2A, the DNA concentration increases with an increasing percentage of living cells, after PMA treatment. Without PMA treatment, the DNA concentration between all mixtures was more similar (Fig. 2B). This indicates that the DNA of the heat-killed cells is still present in PMA-untreated mixtures. The correlation between the DNA concentration and the percentage of living cells is shown in Fig. 3. The same trend can be seen as in (Figs. 2A, 2B and 3): the DNA concentration in the PMA-treated mixtures is increasing with an increasing number of living cells, while the DNA concentration in the PMA-untreated mixtures is higher for a lower percentage of living cells and more quickly reached a plateau phase. This indicates that the DNA concentration (and subsequently the viable cell number) is less overestimated in PMA-treated samples. Increasing the fraction of viable cells in the mixture also led to a significant decrease of the Cq-value determined via qPCR (Fig. 4). The Cq-values obtained with PMA-treated mixtures were significantly higher than the Cq-values of corresponding PMA-untreated mixtures (p < 0.05). The higher Cq-value after PMA treatment indicates that the qPCR amplification of DNA of heat-killed cells is efficiently inhibited by the addition of PMA. This is confirmed by the decrease in Cq-value after increasing the fraction of viable cells and was also described by Alvarez et al. (2013). Without PMA treatment, there is also a decrease in Cq-value after increasing the fraction of viable cells: actively proliferating bacterial cells have more DNA than inactive or dead cells, which lead to a lower Cq-value when the percentage of living cells is increased, even without the addition of PMA. Moreover, the dead cells in the samples were obtained using heat inactivation (95 °C, 15 min). This heat treatment can have an influence on the degradation of the DNA, and may subsequently lead to a lower amount of intact DNA and thus a higher Cq-value when the percentage of inactivated cells increases over live cells (Takahashi et al., 2004).

When plotting the log of the number of viable cells versus Cq-values obtained, a linear relationship was observed between both parameters (R2 = 0.9685) (Fig. 5). The linear range of this relationship is between 105 and 109 cells, indicating that the method used is limited to treatments that result in a number of surviving cells higher than 105. A viable cell number of 105 corresponds to a Cq-value of approximately 30. Nocker et al. (2009) described that signals from killed cells could not be suppressed completely by PMA at very low ratios of live/killed cells, with corresponding Cq-values of 30 or higher. This could be due to the sensitivity of exponential amplification, and could be a possible explanation for the lower limit of the linear range.

In native material of multispecies biofilms, for instance from patients’ material, divergent results might occur, and further testing of spiked samples may be useful to clarify this point of uncertainty also raised by Taylor, Bentham & Ross (2014).

Number of viable cells (determined using SPC) in PMA-treated (50 µM) and untreated samples.

Figure 1: Number of viable cells (determined using SPC) in PMA-treated (50 µM) and untreated samples.

Treatment with PMA (50 µM) did not affect the number of viable cells (Mann-Whitney test, p > 0.05).
Agarose gel with DNA from PMA-treated and untreated mixtures of living and dead cells.

Figure 2: Agarose gel with DNA from PMA-treated and untreated mixtures of living and dead cells.

Genomic DNA extracted from PMA-treated mixtures (A) and PMA-untreated mixtures (B), containing an increasing fraction of viable P. aeruginosa cells.
Correlation between the gDNA concentration (ng/ml) and the percentage of living cells.

Figure 3: Correlation between the gDNA concentration (ng/ml) and the percentage of living cells.

Data were obtained with the same samples used in Figs. 2A and 2B.
Effect of PMA treatment on Cq-values obtained following qPCR using DNA extracted from mixtures of viable and heat-killed P. aeruginosa cells.

Figure 4: Effect of PMA treatment on Cq-values obtained following qPCR using DNA extracted from mixtures of viable and heat-killed P. aeruginosa cells.

The addition of PMA leads to a higher Cq-value. This indicates that the amplification of DNA of heat-killed cells was inhibited by PMA. Error bars represent the standard error mean (n = 6). (: p < 0.05, Mann-Whitney test).
Correlation between log viable P. aeruginosa cells/ml determined via SPC and Cq-values determined via PMA-qPCR.

Figure 5: Correlation between log viable P. aeruginosa cells/ml determined via SPC and Cq-values determined via PMA-qPCR.

The equation for the linear trendline is y = − 0.2421x + 12.642 with R2 = 0.9685. Using this equation, the log viable cells/ml can be calculated from the Cq values obtained with PMA-qPCR. Since one biofilm represents a volume of 200 µl, the log viable cells/biofilm can be calculated by dividing the log viable cells/ml by five. Error bars represent standard deviations (n = 6).

Susceptibility of planktonic and sessile P. aeruginosa cells to colistin, levofloxacin, and tobramycin

The MIC of tobramycin for P. aeruginosa ATCC 9027 planktonic cells was 0.5 µg/ml, the MIC of colistin was 2 µg/ml and the MIC of levofloxacin was 1 µg/ml. These concentrations are below the breakpoint for P. aeruginosa (National Committee for Clinical and Laboratory Standards, 2007), indicating that P. aeruginosa is sensitive to the antibiotics used.

A confocal image of the mature multispecies biofilm is shown in Fig. 6. Both rod-shaped bacteria (presumed Gram-negative bacteria—P. aeruginosa and B. cenocepacia)—and coccal bacteria (presumed to be the Gram-positive cocci—S. aureus and S. anginosus)—were visible.

Microscopy image of a multispecies biofilm.

Figure 6: Microscopy image of a multispecies biofilm.

A confocal image of a mature multispecies biofilm (Live/Dead staining). Rod-shaped and coccal bacteria are clearly visible in the biofilm image.

The susceptibility of sessile P. aeruginosa cells in mono- and multispecies biofilms to antibiotics was determined with PMA-qPCR. The reduction of the number of viable cells was calculated by using the equation for the linear trendline describing the relationship between the log viable P. aeruginosa cells/ml and the Cq-value obtained with PMA-qPCR. Colistin treatment (200 µg/ml, 24 h) led to a significant decrease (p < 0.05) in the number of viable Pseudomonas aeruginosa cells, both in mono- and multispecies biofilms (Fig. 7 and Table 1). In multispecies biofilms, this average reduction was 1.26 log. Using the plate count method, an average reduction of 1 log was observed in P. aeruginosa monospecies biofilms. However, based on PMA-qPCR, more viable cells were present, suggesting that the use of the plate count method leads to an underestimation of the surviving cell numbers. The results also show that P. aeruginosa is significantly more sensitive to colistin in a multispecies biofilm with S. aureus, S. anginosus and B. cenocepacia than in a monospecies biofilm (p < 0.05). After treatment with levofloxacin (100 µg/ml, 24 h), there was also a significant reduction in the number of viable P. aeruginosa cells, both in mono- and multispecies biofilms (p < 0.05) (Fig. 7). Based on the equation for the linear trend line describing the relationship between the log viable P. aeruginosa cells/ml and the Cq-value obtained with PMA-qPCR, a 1.57 log reduction was observed for P. aeruginosa in monospecies biofilms, while in multispecies biofilms, this average reduction was only 0.94 log (p < 0.05) (Table 1). These results indicate that P. aeruginosa is less susceptible to levofloxacin in a multispecies biofilm.

Cq-values obtained with PMA-qPCR of DNA samples recovered from mono- or multispecies biofilms, after treatment with colistin (200 µg/ml), levofloxacin (100 µg/ml) and tobramycin (200 µg/ml) for 24 h.

Figure 7: Cq-values obtained with PMA-qPCR of DNA samples recovered from mono- or multispecies biofilms, after treatment with colistin (200 µg/ml), levofloxacin (100 µg/ml) and tobramycin (200 µg/ml) for 24 h.

Error bars represent standard error mean values (n = 3 × 2). (: p < 0.05)
Table 1:
Comparison of Cq values and calculated number of CFU.
Cq-values (±SEM) obtained with PMA-qPCR and calculated log values of viable P. aeruginosa biofilm cells (using the equation for the linear trendline), after treatment with colistin (200 µg/ml), levofloxacin (100 µg/ml) or tobramycin (200 µg/ml) for 24 h and incubation with PMA. The difference in the calculated number of viable cells after treatment is significantly different between mono- and multispecies biofilms (p < 0.05).
Biofilm type Treatment Cq-value (±SEM) Calculated log viable cells/biofilm Δlog
Multispecies 15.13 ± 0.28 8.27 1.26
Colistin 7.01
20.39 ± 0.64
Monospecies 15.98 ± 0.35 8.07 0.53
Colistin 7.54
18.18 ± 0.23
Multispecies 12.79 ± 0.38 8.85 0.95
Levofloxacin 7.90
16.71 ± 0.38
Monospecies 14.80 ± 0.15 8.36 1.57
Levofloxacin 6.79
21.29 ± 0.46
Multispecies 15.04 ± 0.44 8.30 −0.05
Tobramycin 8.35
14.84 ± 0.43
Monospecies 14.80 ± 0.15 8.36 −0.02
Tobramycin 8.38
14.70 ± 0.18
DOI: 10.7717/peerj.787/table-1

For tobramycin (200 µg/ml, 24 h), there was no significant increase in Cq-values after treatment (p > 0.05) (Fig. 7 and Table 1). Nevertheless, experiments using the plate count method showed an average reduction of P. aeruginosa in monospecies biofilms of 2.35 log after treatment with tobramycin (Fig. 8). A likely explanation is that tobramycin causes little or no loss of membrane integrity (Kim et al., 2008; Tack & Sabath, 1985). Bacterial cells can be killed by tobramycin, but their DNA can still be amplified in the qPCR reaction, as PMA cannot bind to the genomic DNA of intact cells. DNA of dead cells is then extracted together with the DNA of living cells in the DNA extraction procedure and amplified during qPCR, resulting in a lower Cq-value and an overestimation of the number of viable cells.

Number of cells in treated and untreated biofilms.

Figure 8: Number of cells in treated and untreated biofilms.

Log P. aeruginosa biofilm cells in a control biofilm and in a test biofilm, after treatment with tobramycin (200 µg/ml) for 24 h, determined with the plate count method. Errors bars represent standard deviation values (n = 3 × 2) (p < 0.05).

Conclusions

The present study shows that PMA-qPCR is a useful alternative for the plate count method to quantify P. aeruginosa in mono- and multispecies biofilms, after treatment with a membrane-compromising agent. This method can thus be used to avoid underestimating the cell number due to the presence of VBNC. The use of PMA, able to inhibit amplification of DNA of dead cells, avoids an overestimation of the viable cell number seen with conventional qPCR. However, there are some limitations: the number of cells surviving after treatment should be higher than 105 cells/ml and the treatment should compromise the integrity of the membrane. Nevertheless, the PMA-qPCR method was successfully used to determine the difference in susceptibility of P. aeruginosa in a mono- and multispecies biofilm towards colistin and levofloxacin: P. aeruginosa grown in a multispecies biofilm appears to be less affected by levofloxacin, and more sensitive to colistin than when grown in a monospecies biofilm. These data indicate that the effect of the presence of different members in a biofilm on the susceptibility of P. aeruginosa depends on the antibiotic used, and that P. aeruginosa in a multispecies biofilm is not always less susceptible to antibiotics than in a monospecies biofilm.

Supplemental Information

Influence of PMA on cell viability

Raw data for Fig. 1: Influence PMA on cell viability.

DOI: 10.7717/peerj.787/supp-1

Correlation between DNA concentration and % living cells

Raw data for Fig. 3: Correlation between DNA concentration and % living cells.

DOI: 10.7717/peerj.787/supp-2

Correlation between Cq-value and % living cells

Raw data for Fig. 4: Correlation Cq-value and % living cells.

DOI: 10.7717/peerj.787/supp-3

Correlation between viable cells/ml (log) and Cq-value

Raw data for Fig. 5: Correlation log viable cells/ml and Cq-value.

DOI: 10.7717/peerj.787/supp-4

Cq values obtained with and without treatment with colistin, levofloxacin and tobramycin

Raw data for Fig. 7/ Table 1: Cq values with and without treatment with colistin, levofloxacin and tobramycin.

DOI: 10.7717/peerj.787/supp-5

Treatment of a monospecies biofilm with tobramycin

Raw data for Fig. 8: Treatment of a monospecies biofilm with tobramycin.

DOI: 10.7717/peerj.787/supp-6
38 Citations   Views   Downloads