Audio recording of this lecture

Previous or Next lecture

Readings for this lecture:

• Björnsson L, Hugenholtx P, Tyson GW & Blackall LL. 2002 Filamentous Chloroflexi (green non-sulfur bacteria) are abundant in wastewater treatment processes with biological nutrient removal. Microbiol. 148:2309-2318.
• Huber JA, Mark Welch DB, Morrison HG, Huse SM, Neal PR, Butterfield DA & Sogin ML 2007 Microbial population structures in the deep marine biosphere. Science 318:97-100

Molecular Ecology: Advanced Sequence-based Approaches

The basic approach introduced in the last lecture, that of PCR amplifying rRNA sequences from DNA extracted from natural microbial populations and sequencing from this bank, followed often by FISH, can be used in a wide variety of ways. Identifying a predominant specific uncultivated organism in an environment, or doing a basic survey of a few hundred organisms, are simple and straightforward. Today we'll see a couple of ways this approach can be pushed to the limit. In the first, the survey approach and FISH identifications are combined; the survey is used to collect a set of organisms know to to be in the environment, then FISH is used to enumerate these in this environment. In this case, they push the FISH even further by using confocal laser microscopy, which allows them to construct a 3-dimensional image of the sample and the fluorescently-labeled cells within. In the second, the authors take the survey approach to the extreme - obtaining not hundreds, not thousands, not tens of thousands, but hundreds of thousands of sequences from individual environmental samples.

Fluorescent in situ hybridization (FISH) surveys

FISH

We've already talked a little about fluorescent in situ hybridization, or FISH, but it perhaps bears a general review. In this method, cells in an environmental sample are treated to make them permiable (e.g. with toluene) and mixed with an oligonucleotide probe that contained a fluorsecent tag such as Texas Red or Acridine Orange. The probe will find matches in the DNA and/or RNA of the permiablized cells and stick. Unannealed probe is washed out, and the sample is examined by fluorescent microscopy. If enough probe accumulates in a cell, i.e. if it contains the target to which the probe is designed, it should be fluorescent.

The most common probes for FISH in the microbial world target the rRNA. There are two reasons why rRNAs are a good target for fluorescent in situ hybridization, Firstly, it allows you to search using phylogenetically relevant sequences - often rRNA sequence is all you know about an organism. This also allows you to 'tune' the range of organisms you will label - the more conserved the target region of the rRNA, the wider the phylogenetic raange of cells you will label. Secondly, because there are thousands of ribosomes in each cell, a lot of probe can bind to each cell, giving a strong signal. In fact, it is generally seen that only metabolically-active cells contain enough ribosomes to be labeled with an rRNA probe.

Confocal laser scanning microscopy

Confocal laser scanning microscopy (CLSM) is a method that allows the collection of 3-dimensional data from a fluorescent microscope sample. The laser light is focused on a single point in the field of view (about 1um in diameter), and the resulting fluorescence from that point is visualized. The laser focal point is scanned back-and-forth across the field of view to collect a single plane image, then moved slightly up or down in the sample and another plane image is collected...and again, over and over. The result, then, is a 3-D digital image of the sample.

• For more details, see: http://en.wikipedia.org/wiki/Confocal_laser_scanning_microscopes
• And here is a great simulation of a CLSM image: http://micro.magnet.fsu.edu/primer/virtual/confocal/index.html

A phylogenetic FISH survey of Chloroflexi in wastewater sludge

Björnsson L, Hugenholtx P, Tyson GW & Blackall LL. 2002 Filamentous Chloroflexi (green non-sulfur bacteria) are abundant in wastewater treatment processes with biological nutrient removal. Microbiol. 148:2309-2318.

The aim of this study was to design phylum-and subdivision-specific oligonucleotide probes for the Chloroflexi and to evaluate them on sludge samples using FISH, to determine the abundance, morphology and spatial distribution of Chloroflexi in activated sludges.

Wastewater is commonly treated to remove the large part of the organic & nitrogenous (and phophate) material in the form of sludge; the remaining water is disinfected (chlorine, chloramine, ozone or UV are the most common in the US) and released into some waterway. First the raw wastewater is passed through a rough grating to remove larger foreign objects (condoms, guns, dead bodies, alligators, &c), then it undergoes an aerated digestion; in this step, dissolved organics, &c, are taken up by microbes & converted into biomass. Then the wastewater is allowed to settle; the bacteria collect in flocs and settle into a relatively thin layer at the bottom of the pool. The clarified water is pumped from the top of the settled pool, and the settled sludge is pumped into a cesspool (usually called a "lagoon"; this is the most blantant Orwellian misuse of the English language I know of) for anerobic digestion. During anaerobic digestion, a large part of the biomass is converted by microbial activity to CO2 and CH4, and the residue is applied to the soil as fertilizer.

A problem that can occur in the settling stage is the overgrowth of filamentous Bacteria that trap air and prevent flocculation; the result is "bulking" and foaming, in which the settling stage is thwarted.

It is commonly believed that a mesophilic member of the Chloroflexi (green non-sulfur Bacteria), Herpetosiphon, is a major contributor to wastewater bulking. However, the vast majority of filamentous Bacteria in wastewater sludges have not been cultivated or identified; they are most usually catagorized only by microscopic morphology or staining.

In this paper, the authors design a series of phylogenetic probes (fluorescently-labeled oligonucleotides complementary to the ssu-rRNA sequence of particular groups of organisms) to examine Chloroflexi in normal wasterwater after aerobic digestion.

Fig. 1. Evolutionary distance dendrogram of the bacterial phylum Chloroflexi based on comparative analysis of 16S rDNA data. Sequences from isolates are shown in bold. SBR clones fully sequenced in the present study are boxed. Branch points supported by bootstrap resampling are indicated by black circles (bootstrap proportion values shown). The scale bar indicates 0.05 changes per nucleotide. Probe specificities are shown to the right of the figure as are the subdivision affiliations of the sequences. Black squares indicate a perfect match between the probe and target sequences, white squares indicate one or more mismatches between the probe and target sequences and grey squares indicate unresolved bases in the target sequence which are likely to match the probe sequence.

Fig. 2. Confocal laser scanning micrographs of FISH of selected BNR sludge samples. In all cases, the colours of the different probes are indicated in parentheses after the probe. In superimposed images, the overlap between red and green is yellow, between red and blue is magenta, and between red, green and blue is white. All bars are 20 ?m. (a) Rouse Hill. The two phylum probes show good overlap and cells responding to EUBMIX (blue), CFX1223 (green) and GNSB-941 (red) are shown as white. (b) Lab scale 2. The numbers of filaments responding to EUBMIX (blue) and GNSB-941 (red) are high (magenta) whereas only a few larger white filaments are also binding CFX1223 (green). (c) Gibson Island. Filaments binding both Chloroflexi phylum probes GNSB-941 (red) and CFX1223 (green) but not EUBMIX (blue) appear yellow (white arrow). Some filaments are red (yellow arrow) due to binding GNSB-941 (red) but not EUBMIX (blue). (d)
Noosa 3. Chloroflexi binding EUBMIX (blue) and CFXMIX (both red and green) are shown as white filaments inside the floc. (e) Noosa 3. White filaments (arrowed) are Chloroflexi-1 due to binding EUBMIX (blue), CFXMIX (red) and CFX784 (green), while all other Chloroflexi appear magenta (EUBMIX and CFXMIX). (f) Luggage Point. White filaments are Chloroflexi-3 due to binding EUBMIX (blue), CFXMIX (red) and CFX109 (green), some Chloroflexi appear magenta (EUBMIX and CFXMIX) and some are red (arrowed) due to binding only CFXMIX (red).

The majority of filaments seen in the activated sludge were Chloroflexi, and the differentially-labeled probes were able to sort many of these into sub-catagories on the basis of signal color. The use of confocal micrscopy allows the authors to examine these flocs in three-dimensions, to peer into the depths of the flocs using their probes.

This paper is a great example of the right way to use rRNA FISH to study a microbial community, but it also points out the uncertainties and the work involved - this represents a huge investment in time and effort, for the analysis of a single or a few samples. Ecologists need to study samples taken from lots of locations in an environment, and at various times. It is really not yet reasonable to do this using this approach. What's really needed are two tools that have proven to be technically difficult to deal with: rRNA-based phylogenetic DNA arrays (to replace the PCR/cloning/sequencing/analysis part), and standardized multiplex FISH primer sets (so the second can be done efficiently and without waiting for results from the first part).

High-throughput ssu-rRNA sequencing surveys

As DNA sequencing has become easier, faster and cheaper over time, rRNA-based surveys have grown from dozens of sequences to hundreds and now thiusands of sequences. But in every case, the researchers keep getting new sequences - they never reach the bottom of the diversity barrel. So, how deep is microbial diversity in any given sample? One approach to finding out is very high-throughput sequencing of lots of sequences from a sample.

454 Sequencing

454 sequencing is a new technique that allows the researcher to get hundreds of thousands of sequences in a single experiment. The main drawbacks of this method are that (1) the sequences obtained are short, less than 100 bp, and (2) it's expensive and requires one-of-a-kind machines.

The method (as applid in this paper) starts with a standard PCR using primers that will generate a short (<100bp) product. Each of the primers have extra sequences on the end, one for annealing to beads, the other serves as a primer binding-site for the sequnecing reactions.

The DNA is denatured, and mixed with a numerical excess of small beads coated with oligonucleotides that bind to one end of the PCR product. Because there is an excess of beads, most of the beads get only one (or none) DNA molecule annealed to them. The slurry of beads in PCR reaction mixture is then emulsified with oil. Each bead ends up in a droplet of PCR reaction solution in this oil emulsion - each of these droplets is a microscopic PCR reaction tube. The emulsion is run through a set of PCR reaction cycles as usual. The PCR product generated in each drop annals to the tags on the bead, so each bead ends up covered in DNA with a single sequence. The emulsion can then be broken (separating the oil and aqueous phases), an the DNA-covered beads separated out.

The beads are then washed over a silicon wafer honeycomb "picotiter plate". Each cell of the honeycomb is big enough for one and only one bead, so the honeycomb ends up filled with one bad per "well". Each of these then serve as a sequencing reactor. The wells are filled with very small beads coated in 2 enzymes: sulfurylase and luciferase.

The honeycombs are then percolated with reaction mixture of the sequencing primer, DNA polymerase, and APS (adenosine 5'-phosphorothioate) and luciferin, followed by sequential rounds of dGTP, dATP, dCTP and dTTP, over and over again. When the next nucleotide in the sequence matches the dNTP added, the base is added to the growing DNA chain, releasing pyrophosphate. Sulfurylase replaces the thio grup on APS with the pyrophosphate, creating ATP, which is used by luciferase to react with luciferin to generate a pulse of light.

In other words, the well "blinks" a flash of light to signal that the next nucleotide in the sequence matches the current dNTP. If there is a stretch of 2 or more of the same base, proportionally more pyrophosphate is generated, and the flash of light is proportionally stronger. So, as the nucleotides flow past in repeated sequence, each well signal the sequence of the DNA on that bead in a kind of Morse code. A magnifying CCD video camera watches the honycomb, collecting the data from all of the wells over time, and a computer then separates the data from each spot to read the sequences.

(this image is false-colored red, green blue and yellow for each base in a single round of sequencing)

BTW, "454" comes from the address of the startup company that developed this method. Refinement of the techonology is increasing the length of the sequence reads, and there is no conceptual reason why they couldn't ultimately be as long on average as regular sequencing methods.

• "How it works" web page from 454/Roche: http://www.454.com/products-solutions/how-it-works/index.asp
• A more complete explanation of how it works: http://en.wikipedia.org/wiki/454_Life_Sciences

Massive phylogenetic survey of two similar microbial environments

Huber JA, Mark Welch DB, Morrison HG, Huse SM, Neal PR, Butterfield DA & Sogin ML 2007 Microbial population structures in the deep marine biosphere. Science 318:97-100

• Slide show (downloadable PDF)
• Slide show (web-friendly html)

Purpose: There are really four purposes/questions for this paper:

1. This is a proof-of-principle for this method
2. How big is microbial diversity?
3. What is the microbial composition of these two environments?
4. How similar are the community compositions of these two nearby similar environments?

In this paper, the authors use 454 sequencing of a 50-75 bp region of the ssu-rRNA (V6) from DNA isolated from two similar, nearby hydrothermal vent communites. The V6 region doesn't contain enough phylogenetic information in it to generate robust trees, but does contain enough information for individual sequences to be "binned" into particular taxonomic groups - just like the "Classifier" program on the RDP site. The advantage of 454 sequencing is that you get so many sequences; in this study, they get almost a million sequences between two samples. The single most prevalent group in both samples are epsilon-proteobacteria, which were already know to be abindant in hydrothermal environments. But they are twice as abundant (over half) in the Marker 52 sample, and the particular kinds of epsilon-proteiobacteria in the two samples are very different - primarily Arcobacter in Bag City, primarily Sulfurovum in Marker 52.

Perhaps more interesting, the rarifaction curves (which show the rate you're getting novel sequences as the number of sequences increases) do not plateau. They show this in the paper for the sequences overall, but in the supplement, they show the same thing for any particular group of bacteria within the dataset (e.g. gamma-proteobacteria). The implication is that you would expect to continue to get novel sequences at about the same rate no matter how many additional sequences you did. Of course ultimately you'd run out of DNA molecules or out of unique sequence posibilities, but the these would be huge numbers, and the curves predict that if you id another equivalent sample, or did a bigger peice of DNA, the number of new sequences would just keep coming. In other words, microbial diversity is a botomless pit!

Questions for thought

• Can you think of any target other than rRNA than might make a good target for fluorescent in situ hybridization?
  
  
• Given that fluorescent in situ hybridization combined with confocal microscopy allows the 3 dimensional phylogenetic analysis of an environment, can you imagine the four-dimensional (3 in space and 1 in time) analysis of a sample?
  
  
• What are the differences in the data you get from a molecular phylogenetic analysis of a microbial population using FISH vs the clone-and-sequence approach? Under what circunstances you you chose one of these approaches over the other?
  
  
• Given rRNA phylogenetic arrays and standard phylogenetically-targeted FISH primers, how would you repeat this experiment? How would you change it to take advantage of the technology?
  
  
• What do you think you'd get if you did V6 454 sequencing of a more mundane environment, say dirt or a fecal sample? Why do you suppose the uthors chose to do a hydrothermal vent analysis?
  
  
• Given a 50-75 bp sequence and a gram (say 10^12 cells) of bacterial biomass, which would do you think would run out first on the rarifaction curves above, the number of actual rRNA molecules in the sample, or the number of sequence possibilities?
  
  
• In 454 sequencing, you use an excess of beads over DNA molecules to minimize the number of beads with more than one DNA molecule attached. But this is going to happen anyway with some frequency. What might the data from these beads look like? How would you be able to identify these and throw them out? Or would you need to?
  
  
• What do you think the data would look like if, instead of comparing two similar sites the authors had chosen to samples the same site at two different times a week apart?
  
  
• What systematic biases do you think this approach might have?
  
  
• Looking at the data, the "census" is pretty hard to interpret. Can you think of a better way to present the data more clearly? How useful do you think this sort of a census is?

Previous or Next lecture