SEARCH

SEARCH BY CITATION

Keywords:

  • automated pattern analysis;
  • person-independent data evaluation;
  • bacterial communities;
  • microbial community dynamics;
  • DNA pattern analysis;
  • bioprocesses

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. Supporting Information

Altering environmental conditions change structures of microbial communities. These effects have an impact on the single-cell level and can be sensitively detected using community flow cytometry. However, although highly accurate, microbial monitoring campaigns are still rarely performed applying this technique. One reason is the limited access to pattern analysis approaches for the evaluation of microbial cytometric data. In this article, a new analyzing tool, Cytometric Histogram Image Comparison (CHIC), is presented, which realizes trend interpretation of variations in microbial community structures (i) without any previous definition of gates, by working (ii) person independent, and (iii) with low computational demand. Various factors influencing a sensitive determination of changes in community structures were tested. The sensitivity of this technique was found to discriminate down to 0.5% internal variation. The final protocol was exemplarily applied to a complex microbial community dataset, and correlations to experimental variation were successfully shown. © 2013 International Society for Advancement of Cytometry

The determination of highly diverse microbial community structures is challenging and frequently assessed only with molecular techniques (1,2). Although those techniques allow insight into phylogenetic or metabolic gene capacity, they lack quantitative and often functional information. Hence, optically derived cell parameters are an option to characterize individual cells and to monitor changes in complex microbial communities, even when obtained from natural environments (3,4). However, high-resolution flow cytometry has gained little attention so far in microbiology because of the lack of standard procedures.

Apart from deficient infrastructure and protocols, another drawback is the limited accession to reliable cytometric pattern analysis. Until now, pattern analysis requires individual and manual gating decisions, which is time consuming and results in person-dependent data evaluation (3,4). Expertise is needed to gate subcommunities regarding specific cell characteristics (5). In addition, natural microbial communities do not always cluster in clearly separated subsets of cells but can show merged distributions. Therefore, histograms may comprise an unlimited number of categories (cell subsets) and category overlap (6). As a result, different persons will define different gates. Automatic, that is, person independent, gating tools have already been described for applications in human research (7,8); however, they have not been shown to work for complex microbial community analysis so far. To our knowledge, the only gating-independent strategy for microbial community analysis is offered by the FlowFP package (9). It uses a kind of probability binning (as already described in Ref.8) and was successfully applied for monitoring changes in drinking water quality using viability markers (10).

To overcome that problem and make flow cytometric community monitoring easier accessible to microbial ecologists, we developed the Cytometric Histogram Image Comparison (CHIC) tool. The CHIC tool rapidly determines changes in natural community structures using flow cytometric 2D histograms that are converted into images and then processed on a standard computer. Another approach based on image comparison was recently developed by our group; however, it still needs gate information (11,12). In contrast, the CHIC tool allows a comparison of cytometric datasets without gating decisions and any cytometric pre-experience due to an automated image comparison procedure. Therefore, CHIC is person independent in its application and useful for automated evaluation.

The CHIC approach relies on the two-parametric acquisition of the cell size-related forward scatter signal (FSC) and chromosome number staining by using the small and highly DNA-specific dye molecule 4′,6-diamidino-2-phenylindole (DAPI; Refs.13 and14). DAPI is an all-cell-labeling probe, and the fluorescence intensity depends on the DNA content of a cell. The two-parameter information results in fingerprint-like cytometric patterns, which differ in distribution and cell abundance within the histogram. Changes in the community structure become visible if, for example, the number of cells with high DNA content rises while the cell number with low DNA content declines. These changes can easily be detected as variations in event positions in dot plots and their numbers per dot.

We tested the different steps of the protocol for their robustness to optimize the protocol. Conclusively, by applying the protocol on predefined mixtures of microbial communities, we assessed the sensitivity of the tool to detect very small changes in the community structure. To demonstrate the applicability of CHIC for monitoring changes in microbial communities, we used a cytometric dataset of microbial communities originating from biogas reactors. The CHIC results can be correlated with environmental and experimental parameter variations, thus allowing sensitive trend interpretation on community behavior and determination of driving forces of microbial community variation.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. Supporting Information

Principle of CHIC

CHIC determines the difference between cytometric measurements based on picture-based matrix comparison. A detailed manual and necessary macros are provided in the Supporting Information (Supporting Information Protocol 1 and Supporting Information Files 1–3). The commercial software Summit Ver. 3.1 (Beckman-Coulter, Brea, California) for flow cytometry and free software packages ImageJ (15) and R (16) need to be installed. The latter two programs are available for all computer platforms.

For application, the following steps are necessary (details see Supporting Information Protocol 1):

  1. The flow cytometric measurements of each sample need to be visualized with Summit 3.1 using the dot-plot option FSC against UV-induced fluorescence and image settings using 236 gray values (Supporting Information File 1 “Part_I_grey_shades.txt”). One image per sample is exported as bitmap (.bmp) file, and all files are stored in a single folder. This part can also be performed using any other cytometric software followed by a conversion of every histogram into gray shadows or using typical image analysis programs like the open-source software IrfanView (http://www.irfanview.com/).

  2. In the second step, image analysis is performed. Therefore, ImageJ needs to be run and the provided macro “Part_II_image_analysis.txt” (Supporting Information File 2) to be initialized. Installing the macro enables ImageJ to perform several analysis steps.

  3. Following, semiautomatic cutting of the images needs to be performed. This step is necessary to remove instrumental noise and automatic axis labels exported by Summit 3.1. The area of interest is selected, and the cut images are stored in a new subfolder. In the next step (image analysis process), a pairwise comparison between all cut images, based on exclusive disjunction (XOR) of single-pixel values, is performed. The exclusive differences found in each position between two images are displayed and stored in new (temporary) image files (Supporting Information Fig. S1a). Thus, the XOR algorithm highlights differences between two images. If there is no difference between two compared pixel values, the resulting pixel in the XOR image is black. The pixel color in the XOR image changes to gray and finally to white in maximum the more differences exist between the two original images. Small differences will therefore result in a few dark gray pixels, whereas white pixels mark huge differences in the resulting XOR image. Finally, the script determines the difference between two images by the average gray value of the XOR image. Calculating the average gray value directly from the images would result in that principally dominating areas without any events (black areas) cause a too low dissimilarity between images. This effect can be circumvented by restricting the compared pixel number to only the overlapping area of both cytometric measurements. The overlay of the images (Supporting Information Fig. S1b) is also automatically performed by the provided macro, and the number of pixels with pixel values, which are not white, is also given. The accumulated results from all comparisons are stored within the image folder.

  4. Now, the statistical software “R” has to be run for conducting the statistical analysis of the results from the image analysis. The analysis can be simply performed by drag-and-drop the script provided in the “Part_III_similarity_calculation.r” file (Supporting Information File 3). During this procedure, a dissimilarity matrix is calculated, and its results are visualized using nonmetric multidimensional scaling (nMDS) and cluster analysis.

The dissimilarity (Psim) between two images is estimated by using the total sum of the pixel values from the XOR images and the information of the number of informative pixel from the overlay. It can be calculated by the following equation:

  • equation image(1)

with equation imagewith A ij and B ij as the referring pixels at positions i and j of the compared images A and B, and PNW as the sum of the informative (=nonwhite) pixels of both images. In addition, the provided script allows performing correlation analysis with selected experimental parameters.

Defining the Best Histogram Resolution

We tested whether color setting and pixel resolution of the cytometric histogram images influence the potential of the CHIC to detect differences in the microbial community patterns. As a histogram is partitioned into channel numbers because of the binary code of electronic signal recording, image resolutions in the range of 64, 128, 256, 512, or 1,024 channel numbers are usually provided by conventional cytometric software like Summit 3.1. Equal cell-related signals will be accumulatively binned on the same pixel position. As a result, images with low gray-scale resolution provide high between-pixel-contrast but only low within-pixel-information, whereas images with high gray-scale resolution provide high within-pixel-information but low between-pixel-contrast. An optimal combination of both aspects is also relevant to discover small differences between two images. The area of binned signals is defined by the channel resolution and the color scale. In a low-resolution 2D histogram of 64 vs. 64 channels, the information of all measured events is downgraded obliterating local variations in event density and prone to cause an event overflow. Higher binary code-related resolutions guarantee that slight variations can be visualized more easily.

To determine the best histogram resolution for the image computation, a dataset was tested with different export settings (Summit 3.1 channel resolutions 64, 128, 256, 512, and 1,024). Forty-five measurements from the “degree of receptiveness” test (see below) were chosen. They represented 15 different microbial community structures, each measured thrice. The dot plots of the complete dataset are exemplarily represented by four samples and their XOR images (Supporting Information Fig. S 2). It is recommended to perform a similar procedure when other cell numbers are measured or other data export values are available. The use of smoothing options is not recommended.

Estimation of Methodical Variations When Using CHIC

A CHIC analysis mirrors alterations. The most aspired one is the reflection of a real change in community structure. Although the procedure is always performed with maximum reproducibility effort, small variations cannot be avoided and will be visible using CHIC. Small variations can also be caused by the measurement itself. With even 200,000 cells per measurement, cells with very low abundance may not be in the span of analyzed cell numbers in complex microbial communities.

To determine variations caused by technical handling of the samples, a specific experimental setup was created (Supporting Information Fig. S3). Two microbial community samples taken within a time interval of about 4 h were treated and prepared for measurement in three parallels (T1A-C and T2A-C) as well as analyzed repeatedly (thrice per sample). The resulting dataset was evaluated by CHIC, and the dissimilarity values passed onto cluster analysis.

As a recommendation, a similar test procedure should be performed before applying the CHIC approach to assess artificial internal variations for every unknown natural community investigated.

Degree of Receptiveness

The degree of receptiveness of the CHIC tool to monitor variations in cytometric patterns was tested. Three different microbial communities obtained from bench-top cultivated natural microbial biogas communities and defined mixtures of them were prepared and cytometrically analyzed (all files were uploaded to the FlowRepository: FR-FCM-ZZ42). Afterwards, the histograms were converted into images (channel resolutions 128 and 256), and the dissimilarity analysis was performed using the CHIC protocol and compared by nMDS in R (16). The three community samples differed in the number of cell clusters, position of clusters in the histogram, and abundance of cells in clusters. The unchanged three samples served as control patterns A, B, and C. The three communities were used to prepare defined ratios of cell numbers (Sample A with Sample B: 99.5:0.5, 95:5, 85:15, 70:30, 50:50, 30:70, 15:85, 5:95, 0.5:99.5, and the same ratios for Sample A with Sample C and Sample B with Sample C). Every combination was measured thrice.

The ability of CHIC to distinguish very similar samples was tested with R. A 95% confidence interval was calculated for repeated samples to determine statistical differences using “ordiellipse()” command provided by the R-package “vegan” (17). The measurements of each mixing ratio were defined as group, and nonoverlapping of the confidence interval ellipses was defined as a significant difference.

Cultivation of Microorganisms

Complex microbial communities were anaerobic enrichment cultures originating from laboratory-scale biogas reactors, with Community A grown on maize silage and Community B on distillers grains. All enrichments were cultivated on DSM 120 medium in 55-ml serum bottles under nitrogen atmosphere. As growth substrates, either acetate (20 mM) or methanol (0.5%, v/v) was used. The pH varied from 6.5 to 8.5. Individual sample parameters are given in Supporting Information Table S1. The cytometric files were uploaded to the FlowRepository: FR-FCM-ZZ43.

Sample Preparation

Sample fixation was performed with paraformaldehyde solution (2% in PBS, containing 0.4 M Na2HPO4/NaH2PO4, 150 mM NaCl, pH 7) at 4°C overnight. Then, the samples were washed thrice (centrifugation at 3,200 g, 10 min) with PBS and stored at 4°C until analysis. Besides a strong vortex step of 10 s, no additional cell detachment treatment was performed. The cells were stained with DAPI [0.68 μM (Sigma-Aldrich, St. Louis, MO), 400 mM Na2HPO4, pH 7.0] following the procedure described in Ref.14.

Instrumental Details

All samples were measured with a MoFlo cell sorter (DakoCytomation, Glostrup, Denmark), which is equipped with two lasers. The 488-nm argon laser (400 mW) was used for the measurement of FSC (488/10) and side-scatter signal (488/10, trigger signal) and a ML-UV laser (333–365 nm, 100 mW) for UV-induced fluorescence (450/65). Photomultiplier tubes were purchased from Hamamatsu Photonics (Models R928 and R3896; Hamamatsu City, Japan). Daily and in-between-day calibration of the instrument was performed with fluorescent beads [yellow-green fluorescent beads: 2 μm, FluoSpheres 505/515, F-8827, crimson fluorescent beads: 1 μm, FluoSpheres 625/645, F-8816 (Molecular Probes, Eugene, OR); Fluoresbrite BB Carboxylate microspheres: 0.5 μm (Polyscience)]. Cytometric data acquisition was performed with Summit Ver. 3.1 (Beckman-Coulter, Brea, CA).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. Supporting Information

Microbial systems are highly dynamic, and the detection of structural variation and especially the coverage of these dynamics are challenging. Cell-based markers representing proliferation, cell size, and granularity are well-suited parameters. Based on these parameters, CHIC allows the determination of differences between histograms, a trend interpretation of community behavior, and the correlation to factors responsible for community structure variation.

The procedure can be easily performed with the provided workflow (Supporting Information Protocol 1). The transformation of cytometric histograms to images was shown, followed by defining the best settings for sensitive pattern evaluation. The degree of methodical variation was also determined. In addition, the sensitivity of CHIC to detect structural community changes was verified. Finally, CHIC was applied to a dataset from a natural community to test its function.

Development of the CHIC Workflow

The choice of the image resolution during the first step of the CHIC procedure impaired the potential of tracking changes in the microbial patterns. The 64-channel gray-scale resolution showed only minor detectable differences between images, whereas the 128-channel resolution already provided more information. A further increase to 256 channels caused a higher within-pixel-resolution but less-defined clustering (lower between-pixel-differences). These graphic effects became more obvious when looking at the XOR images of exemplary samples (Supporting Information Fig. S 2). To find out which channel resolution might be the best for the CHIC application, a dataset of 45 measurements was used. Five channel resolutions were tested, and the complete values for the highest and lowest dissimilarities are presented in the table of Supporting Information Figure S2. The highest dissimilarity range (0.335) was achieved with the 64-channel resolution. This range decreased with increasing channel resolution down to 0.023.

In addition, the replicates of all samples were evaluated toward the spreading of Psim values using their standard deviation. The lowest value was found for the 256-channel resolution with 2.461% of the total dissimilarity range (other values see Supporting Information Fig. S2). The value was higher for the 128-channel resolution (3.122%). Although the dissimilarity range was lower for the 256-channel resolution in comparison with the 128-channel resolution, the lower relative standard deviation supports choosing the 256-channel resolution for creating images when 200,000 events were measured.

To determine the community structure variations caused solely by technical handling of the sample, the following setup was used (Supporting Information Fig. S3): two different samples were prepared and analyzed thrice per sample (T1A-C and T2A-C). The sample preparation procedure caused dissimilarity values Psim of 0.148–0.157 (T1, mean: 0.152) and 0.152–0.177 (T2, mean: 0.164), indicating that variation up to a value of 0.177 can be introduced artificially. For comparison, the deviation caused by the cytometric measurement itself showed lowest and highest dissimilarity values of 0.149 (T1) and 0.158 (T2). These values are in the range of the lowest dissimilarity values of the 256-channel resolution and will therefore not blanket a real change in community structure. They are caused by intrinsic noise (Supporting Information 1) and would be used as minimum dissimilarity Psim min for normalization (Supporting Information 2).

Degree of Receptiveness

The degree of receptiveness describes the ability to distinguish two samples that are different. The term is equal to the resolution of a method. Because we already used the term resolution for the image creation (see above), we stay with the term receptiveness at this point.

Three microbial communities and mixtures of them were cytometrically analyzed and proceeded with CHIC. Overall, 90 measurements were performed. The receptiveness of the CHIC tool was demonstrated in a dissimilarity comparison approach (Fig. 1 using 128-channel resolution and Supporting Information Fig. S4 for 256-channel resolution). The black circles displayed the dissimilarity values resulting from the CHIC analysis. Pie charts were used to display the varying contributions of the Samples A, B, and C in the different mixtures. The end points of the “triangle” were defined by the unmixed samples A, B, and C. The sample mixtures were arranged between the unmixed samples regarding their similarity to each other strongly reflecting the different sample proportions. Samples A and B resulted in a dissimilarity value Psim of 0.240. Sample C was less different to Sample A (0.234) or Sample B (0.234) as Sample C had higher overlapping areas with both samples.

thumbnail image

Figure 1. The degree of receptiveness was tested with three microbial communities and defined mixtures thereof. DAPI-stained samples were measured: first, the pure samples A, B, C (dot plots presenting FSC vs. UV-induced fluorescence), and then their mixtures (A with B: 99.5:0.5, 95:5, 85:15, 70:30, 50:50, 30:70, 15:85, 5:95, 0.5:99.5, and the same for A with C and B with C), each thrice to include technical variation. The complete dataset of 90 measurements was then processed with CHIC. Pie charts were used instead of data labels to visualize the varying sample contributions. The complete analyzing procedure from image export to nMDS visualization was performed within 20 min.

Download figure to PowerPoint

thumbnail image

Figure 2. CHIC reveals community differences regarding growth conditions. We used CHIC to analyze a dataset of 17 cytometric measurements. The effect of type of inoculum and substrates as well as pH on community structures was evaluated. Differences between the samples were found. They could be significantly explained by the type of inoculum (inoculum A: closed symbols; inoculum B: open symbols) and the choice of the substrate (acetate: squares; methanol: circles). The pH of the growth medium did not cause any significant difference.

Download figure to PowerPoint

To test the ability of CHIC to reveal small sample differences, a significance test was performed between unmixed samples (A, B, and C) and those mixed at low ratios of 99.5:0.5 as well as 95:5 (Table 1). The significance test was performed with the images exported both from the 128- and 256-channel resolutions. Based on the 128-channel resolution, CHIC clearly separated the unmixed samples from those mixed at ratios 95%:5%. Even the 99.5%:0.5% ratios of mixed samples were detectable aside the unmixed samples, with exception of combinations using Sample C. As this sample was less different to Samples A and B, it was not surprising that not all 99.5%:0.5% mixtures containing Sample C were distinguishable. However, statistical significant community differences between mixtures were detectable in eight of 12 mixed samples when only 0.5% of another sample was added. Thus, depending on the location of variance within a microbial community pattern, a variation of only 0.5% is most likely, whereas 5% variation is confidently detected by CHIC. Using images based on the 256-channel resolution, the number of significantly different samples was also good, but somewhat lower (23 compared to 26 with the 128-channel resolution; Table 1) despite the better dissimilarity range to variance ratio resulting from the resolution test. Therefore, both resolutions can be regarded as sensitive for detecting changes in community structures.

Table 1. Significance test for the ability of CHIC to distinguish highly similar samples thus detecting small variations in community structure
256-channel resolution100% A99.5% A + 0.5% B95% A + 5% B99.5% A + 0.5% C 100% B99.5% B + 0.5% C95% B + 5% C99.5% B + 0.5% A 100% C99.5% C + 0.5% A95% C + 5% A99.5% C + 0.5% B
  1. The most similar samples of the degree of receptiveness test (pure samples, mixing ratios of 99.5:0.5 and 95:5) were compared with each other using two different channel resolutions and marked with “+” if significant statistical differentiation was possible and “o” if differentiation was not possible.

99.5% A + 0.5% B+   99.5% B + 0.5% C+   99.5% C + 0.5% A+   
95% A + 5% B+o  95% B + 5% C++  95% C + 5% A++  
99.5% A + 0.5% C+++ 99.5% B + 0.5% Ao++ 99.5% C + 0.5% B+o+ 
95% A + 5% C++o+95% B + 5% A++++95% C + 5% Boo+o
128-channel resolution100% A99.5% A + 0.5% B95% A + 5% B99.5% A + 0.5% C 100% B99.5% B + 0.5% C95% B + 5% C99.5% B + 0.5% A 100% C99.5% C + 0.5% A95% C + 5% A99.5% C + 0.5% B
99.5% A + 0.5% B+   99.5% B + 0.5% Co   99.5% C + 0.5% A+   
95% A + 5% B++  95% B + 5% C++  95% C + 5% A++  
99.5% A + 0.5% Co++ 99.5% B + 0.5% A+++ 99.5% C + 0.5% Boo+ 
95% A + 5% C++++95% B + 5% A++++95% C + 5% B++++

Applicability of the Method—CHIC Visualizes Community Differences in Enrichment Cultures

To test the CHIC tool, a dataset of 17 cytometric measurements (S1–S17) was used. The complete dataset can be found in Supporting Information Table S1.

CHIC analysis revealed that the community structure varied depending on the type of inoculum and the substrates (Fig. 2). Communities originating either from inoculum A (closed symbols) or B (red symbols) clustered apart because of different structures. In addition, these differences were obviously supported by the type of the substrate (methanol: circle; acetate: square). It seems that communities growing on the same substrate and originating from the identical inoculum may show the highest similarity in structure.

When correlating the experimental parameters (pH, substrate, and inoculum) with the biotic parameters (cytometric images), the type of inoculum and the substrates were in fact identified as main factors causing structural differences between the microbial communities. The correlation analyses also showed that the varying pH values had no significant effect on the community structures. Details on the individual P values are shown in Supporting Information Table S2.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. Supporting Information

Flow cytometric FSC versus DNA content measurements mirror alterations in microbial community structures reflecting changes in environmental parameters as was already demonstrated for communities from wastewater treatment processes (3,5).

In ecological microbiology, fingerprinting as well as deep sequencing technologies are used to obtain knowledge with regard to phylogenetic affiliation and functional capacity information of community members to various resolution degrees. Deep sequencing techniques nowadays provide several ten thousands of sequences covering a single sample. The creation of such data assemblies is extremely expensive and requires experience and time to extract the information on the differences between two or several of these vast datasets (18,19). We, therefore, see a huge potential for the application of community flow cytometry in monitoring dynamics in community structures because ample datasets can be obtained very easily, fast, and nearly unlimited. In contrast to the molecular techniques, flow cytometry measures community changes every few minutes lowering months of data processing and evaluation to minutes in order to detect community structure variations.

We therefore developed CHIC, a quick tool to determine changes in the structure of natural communities. The analysis is fast and easy to perform. The provided macros and the instruction manual allow also nonspecialists in cytometric pattern analysis to perform community monitoring in complex natural samples.

We would like to stress our main intention to provide an easy-to-use, person-independent analyzing and monitoring tool. The same dataset analyzed by a number of different persons should result in very similar findings. However, the CHIC result matrix is still flexible allowing different visualization schemes or further data analyzing procedures. Additional evaluation tools regarding estimation of diversity indices and dynamics, terms that are widely used in microbial ecology, can also be derived from the CHIC tool (manuscript under preparation). We are aware that our approach in the presented way does not provide information on taxonomic changes in the process of a monitoring campaign. However, we would like to emphasize that this knowledge is not the main priority when monitoring certain environmental areas or complex bioreactor systems like groundwater or wastewater communities. Most often, rather stability or instability of microbial systems is of interest, that is, how fast changes appear and if these changes are linked to environmental causes or stressors, thereby providing means to adjust communities and their performances quickly and nearly on line. Microbial communities are known to react very fast and are sensitive on microenvironmental stressors. CHIC, therefore, might also provide a means to pre-evaluate hot-spot activities in vast microbial sampling campaigns.

CHIC is a tool that enables for interpretation of vast microbiological datasets from environment. It uses a gating-independent strategy for microbial community analysis. FlowFP (9) can be applied to similar questions despite it is primarily configured for quality assessment. Regarding the data analyzing procedure, CHIC and FlowFP are different in the event-binning procedure. The FlowFP procedure defines bins with equal event numbers, whereas CHIC has fixed bins (based on the channel resolution) and compares changes in the event numbers per bin (corresponds to pixel intensity). The definition of bins is computationally intensive. For microbiological questions, we need to record 200,000 or 250,000 cells per measurement, whereas in medical applications, 10,000–20,000 cells are more common. We therefore circumvent the definition of bins in the CHIC procedure by using images that have defined numbers and positions of bins. A primary training dataset is therefore not required for the CHIC analysis. In addition, the CHIC procedure is easier to learn for beginners in comparison with the R-based tools provided by the Bioconductor platform (9, 20, 21).

A prerequisite for using CHIC is to perform reliable measurements including constant cell numbers. Careful adjustment of the instrument and proper controls are crucial (Supporting Information 3). We suggest to use various fluorescent beads when setting up the instrument and to include overlap measurements between different measuring days. It is also necessary to keep maximum reproducibility in mind when preparing the samples. We defined our internal maximum dissimilarity that can be introduced by the sample preparation. Other sample preparation techniques, for example, including more aggressive cell detachment strategies or different fixation and staining procedures, can result in higher values (data not shown). To define this methodical dissimilarity value, it is essential to differentiate a real change in community structure from variations introduced by the sample preparation procedure.

Despite method-inherent noise detection, we demonstrated the high sensitivity of CHIC in detecting small community structure variations even in the range of 0.5%. We have also demonstrated the applicability of CHIC using a biological dataset. Microbial community structure differences were clearly found in response to differences in selected growth conditions. In conclusion, CHIC allows sensitive, quick monitoring, and person-independent evaluation of structural community changes, and we hope to encourage microbial ecologists to use CHIC for their microbial monitoring experiments.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. Supporting Information

The authors thank Thomas Hübschmann for helpful discussions, Thomas Schmidt for biogas reactor samples, and Birke Brumme for technical assistance. This work was integrated in the internal research and development program of the UFZ and the CITE program (chemicals in the environment).

LITERATURE CITED

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. Supporting Information
  • 1
    Talbot G, Topp E, Palin MF, Massé DI. Evaluation of molecular methods used for establishing the interactions and functions of microorganisms in anaerobic bioreactors. Water Res 2008; 42: 513537.
  • 2
    Morales SE, Holben WE. Linking bacterial identities and ecosystem processes: Can ‘omic’ analyses be more than the sum of their parts?. FEMS Microbiol Ecol 2011; 75: 216.
  • 3
    Günther S, Koch C, Hübschmann T, Röske I, Müller RA, Bley T, Harms H, Müller S. Correlation of community dynamics and process parameters as a tool for the prediction of the stability of wastewater treatment. Environ Sci Technol 2012; 46: 8492.
  • 4
    Günther S, Trutnau M, Kleinsteuber S, Hause G, Bley, T, Röske I, Harms H, Müller S. Dynamics of polyphosphate-accumulating bacteria in wastewater treatment plant microbial communities detected via DAPI (4′,6′-diamidino-2-phenylindole) and tetracycline labeling. Appl Environ Microbiol 2009; 75: 21112121.
  • 5
    Koch C, Günther S, Desta AF, Hübschmann T, Müller S. Cytometric fingerprinting for analysing microbial intra-community structure variation and identifying sub-community function. Nat Protocols 2013; 8: 190202.
  • 6
    Boddy L, Wilkins MF, Morris CW. Pattern recognition in flow cytometry. Cytometry 2001; 44: 195209.
  • 7
    Bashashati A, Brinkman R. A survey of flow cytometry data analysis methods. Adv Bioinformatics 2009; Article ID 584603.
  • 8
    Roederer M, Moore W, Treister A, Hardy RR, Herzenberg LA. Probability binning comparison: A metric for quantitating multivariate distribution differences. Cytometry 2001; 45: 4755.
  • 9
    Rogers WT, Holyst HA. FlowFP: A bioconductor package for fingerprinting flow cytometric data. Adv Bioinformatics 2009: 193947.
  • 10
    De Roy K, Clement L, Thas O, Wang Y, Boon N. Flow cytometry for fast microbial community fingerprinting. Water Res 2012; 46: 907919.
  • 11
    Bombach P, Hübschmann T, Fetzer I, Kleinsteuber S, Geyer R, Harms H, Müller S. Resolution of natural microbial community dynamics by community fingerprinting, flow cytometry, and trend interpretation analysis. In: Müller S,Bley T, editors. Advances in Biochemical Engineering and Biotechnology, Vol. 124. Berlin: Springer; 2011. pp 151181.
  • 12
    Müller S, Hübschmann T, Kleinsteuber S, Vogt C. High resolution single cell analytics to follow microbial community dynamics in anaerobic ecosystems. Methods 2012; 57: 338349.
  • 13
    Müller S. Modes of cytometric bacterial DNA pattern: A tool for pursuing growth. Cell Prolif 2007; 40: 621639.
  • 14
    Günther S, Hübschmann T, Rudolf M, Eschenhagen M, Röske I, Harms H, Müller S. Fixation procedures for flow cytometric analysis of environmental bacteria. J Microbiol Methods 2008; 75: 127134.
  • 15
    Schneider CA, Rasband WS, Eliceiri KW. NIH image to ImageJ: 25 years of image analysis. Nat Methods 2012; 9: 671675.
  • 16
    R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2011. Available at http://www.R-project.org/.
  • 17
    Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O'Hara RB, Simpson GL, Solymos P, Stevens MHH, Wagner H. Vegan: Community ecology package. R package version 1.17-10. 2011. Available at http://CRAN.R-project.org/package=vegan.
  • 18
    Hamady M, Knight R. Microbial community profiling for human microbiome projects: Tools, techniques, and challenges. Genome Res 2009; 19: 11411152.
  • 19
    Werner JJ, Knights D, Garcia ML, Scalfone NB, Smith S, Yarasheski K, Cummings TA, Beers AR, Knight R, Angenent LT. Bacterial community structures are unique and resilient in full-scale bioenergy systems. Proc Natl Acad Sci USA 2011; 108: 41584163.
  • 20
    Sarkar D, Le Meur N, Gentleman R. Using flowViz to visualize flow cytometry data. Bioinformatics 2008; 24: 878879.
  • 21
    Lee K, Hahne F, Sarkar D, Gentleman R. iFlow: A graphical user interface for flow cytometry tools in bioconductor. Adv Bioinformatics 2009: 103839.

Supporting Information

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
Supporting_Information_CHIC.pdf1496KSupporting Information
Part_I_grey_shades.txt4KSupporting Information_Part I
Part_II_image_analysis.txt4KSupporting Information Part II
Part_III_similarity_calculation.rtf36KSupporting Information Part III
MIFlowCyt_Item_Checklist.doc50KSupporting Information: MIFlowCyt

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.