1 of 12

Using phylogenetics to tackle uncertainty in delimiting microbial taxa

Austin Koontz, MSc Student

Dr. Bonnie Waring

Dr. William D. Pearse

Utah State University

iTOL: Interactive Tree of Life

2 of 12

Background

smithsonianmag.com

sciencephoto.com

(Fast!)

To give you a context of what I’m going to be discussing:�

As ecologists and biologists, we’re interested in species and communities, and the boundaries formed by species definitions. And in the entire field of biology, there is no subdiscipline that has spent more time wrestling with the species concept than microbial ecology, where a robust and widely accepted species-concept remains out of reach.

The talk I’m going to give is motivated from this intransigency, and so is centered around delimitation of microbial taxa. But the technique applies to all types of communities, and I hope that by the end, whether you’re interested in microbes or not, you’ll find it interesting with regards to speciation generally. So, even if you’re not interested in microbial ecology, hang tight, you might be surprised.

3 of 12

Overview

Problem: classification of microbial taxa�
The tools that exist: barcode genes, OTUs/ASVs�
What’s missing: moving from sequence clusters to species�
What we provide: a technique informing species thresholds based on shared evolutionary history

ecdc.europa.edu

www.bbc.co.uk

So, as a broad overview:

Problem: Microbial taxa (bacteria and fungi) are inherently difficult to classify, and and this leads to discrepancies between estimates of microbial diversity and issues with diversity metrics applied to microbial communities.
The tools that exist: Microbial ecologists have developed means of comparing certain genetic sequences in order to group microbial individuals together.
What’s missing: Despite the tools that we have, there’s contention over what threshold value to use to generate sequence clusters. Ultimately, the clusters defined in this statistical way aren’t robust enough to be unequivocally considered “species”.What we provide: By modeling the shared evolutionary history of species, we can diminish or enhance the effect that different species boundaries have on the community we want to understand. We can then use that effect to inform what range of species boundaries make sense for that community.�

4 of 12

Difficulties in microbial classification

Bacteria:

Fungi:

Taylor et al., Applied and Environmental Microbiology, 2016

Microbial ecologists have long wrestled with creating a reliable means of classifying microbial taxa:

unculturable taxa (Great Plate Count Anomaly)
rapid mutation rate
horizontal gene transfer

With the advent of molecular biological techniques came new means of determining species boundaries in microbes. Today, barcode gene regions, and their sequence similarities between different individuals, are used to classify microbial taxa.

Bacteria--16S rRNA gene
Fungi--internal transcribed spacer (ITS) region�

These genes are informative, because they are

universal,
comparable (they contain conserved regions), and
distinctive (they contain variable regions)

These gene sequences become the proxies for the microbial individuals we are trying to classify.

DNA is extracted from a community of individuals, and is sequenced. Groups of similar sequences (for each relevant gene) are clustered based on similarity into what are known as operational taxonomic units, or OTUS.

5 of 12

Difficulties in species delimitation: �OTUs, ASVs, and statistical thresholds

ASVs: 100% sequence similarity threshold

OTUs: statistical tools used to address ambiguity in microbial species delimitation. Defined by DNA similarity threshold values--historically, 97% has been a widely used threshold value due to its ties to DNA-DNA strand reassociation studies. However, over time the usage of different thresholds has shifted, moving to 98 and 99% for different reasons. �
This trend has continued to the extreme, with only 100% similar sequences being considered members of the same group (or species). Such groups are called amplicon sequence variants (ASVs). Researchers have called for the usage of ASVs over OTUs due to their supposedly greater reproducibility and their relatively higher information content (as compared to OTUs, which are associated with a loss of information) �
However, even with usage of ASVs, limitations remain:

Because of sequencing error, ASVs tend to inflate estimated microbial diversity, even after the application of denoisers
OTUs, clades, and shared traits
Historical usages of OTUs, communication between ecologists�

So, my research is motivated from 2 broad questions stemming from these problems with OTUs.�

6 of 12

Questions

Specifically: What percentage DNA threshold should be utilized?�
Generally: How can we move from sequence clusters, determined from a statistical value, to a more robust identity of species?

ecdc.europa.edu

www.bbc.co.uk

7 of 12

Our central assumption in species delimitation

Interspecific differences

Intraspecific differences

Sequencing error

>

Greater resolution of differences ➛ Greater phylogenetic diversity

A phylogenetic tree typically represents the shared evolutionary history of different species, and when in evolutionary time speciation events occur.�
However, we can model further levels of differences between different individuals by adding on branches that distinguish intraspecific differences (i.e., different populations). And crucially, our central assumption in species delimitation is that these intraspecific differences are less than the interspecific differences they’re added onto. Consider this in the context of population biology--if this weren’t the case, then there would be no distinction between different species and different populations. ��Similarly, we can model differences arising from random sequencing error on top of differences between different populations. And similarly, these differences due to random error must be less than the intraspecific differences they’re being added onto. This must be true in the context of genetic sequencing--if error were greater than any intraspecific differences, then the sequences couldn’t provide any valuable information. �
These changes in the structure of the phylogenetic tree of our community are reflected through metrics of phylogenetic diversity. With each additional level of difference, we increase the overall diversity in the community, because we’re increasing the number of branches. �
But differences between individuals can also be represented, not by just adding more tips, but by changing the proportion of shared branch length, or shared history. In a sense, this alters the weight of the distinctions between individuals, or the phylogenetic signal. This type of difference between individuals can be modeled using a classic family of branch length transformations--Pagel’s transformations.

8 of 12

Pagel’s Transformations: proportion of shared history represented through phylogeny

δ < 1

δ > 1

Greater proportion of unshared history ➛ Greater phylogenetic diversity

δ = 1

Pagel’s transformations: a classic family of branch length transformations which are measures of what’s called phylogenetic signal , or the extent to which differences between related species is the result of phylogeny. Specifically, I’m going to talk about delta transformations, which transforms the proportion of shared branch lengths between tips. We start with a baseline phylogeny, with a delta of 1. ��If delta is decreased, then speciation events become evolutionarily recent, and a greater proportion of branches are shared. This lowers the overall phylogenetic diversity of the community, because with greater proportion of a shared history, there’s less a proportion for evolutionary time to accrue individual differences.��Conversely, by increasing the value of this metric, we increase the proportion of branches in a tree that are unshared, thus increasing the level of distinction between individuals within a community, and thus increasing the overall phylogenetic diversity.

Both of these tools--branch addition and branch length transformation--are means of increasing the level of difference between individuals in a community--either by increasing the number of distinctions determining what those individuals are, or enhancing the weight of those distinctions. �
Using these two tools, we ran simulations creating various demo phylogenies and communities with species abundances based around these phylogenies. We then measured the phylogenetic diversity of these communities at different levels of branch transformations, and looked to see how patterns of that diversity changed. I’ll show you what the results for one such community looked like.

9 of 12

Species thresholds affect community phylogenetic diversity in informative ways

We have a demo phylogeny with additional branches indicating intraspecific differences and random differences generated sequencing error. Below, we have a graph plotting a metric of phylogenetic diversity (here, mean pairwise distance, or mpd) of the demo communities (simulated along with the phylogeny) against the delta transformation value of that phylogeny. The different lines represent different communities, with different species abundances, and therefore, different (but largely similar) initial phylogenetic diversities (as some species in the community may go extinct randomly).�
Here’s what we find:

At very low values of delta, phylogenetic diversity decreases, because this implies the entirety of evolutionary history to be shared
At high values of delta, phylogenetic diversity tapers off, indicating that at a certain point, greater distinctions between different lineages no longer contributes to the overall diversity of a community.
At certain values of delta (box here), the relative diversity of different communities shift, such that one community becomes more phylogenetically diverse than another. This crossover point indicates a threshold delta value that can be informative--below this shifting point, we’re grouping together taxa that may have important differences between one another. �

These delta points correspond to different OTU thresholds--at a certain low delta point, we’re clumping together species that should be considered separate--say, a 90% OTU. At a high delta point, we’re splitting apart groups that have only small differences between them--say, an ASV. What we find in our simulations is that, for scenarios in which intraspecific differences are less that interspecific differences, there’s a range of threshold values in which measures of the diversity of the community being analyzed remain constant. �
In addition to informing OTU boundaries for microbial taxa, this technique can also be applied to macrotaxa

10 of 12

Similarly, species thresholds affect perceived community diversity for macrotaxa

Macroinvertebrate community data for the Rio Laja, Mexico

Helmus et al., Oecologia, 2013

11 of 12

Conclusions & Next Steps

The extent of shared evolutionary history can inform meaningful species boundaries.�
In a range of possible threshold values, diversity remains constant.�
Beyond higher delta value thresholds, overall diversity becomes inflated.�

ecdc.europa.edu

www.bbc.co.uk

(Fast!)
Takeaway�

We can use phylogeny, or the understanding of a community’s shared evolutionary history, to inform the species boundaries used to distinguish individuals from one another within a community. This technique of informing species delimitation goes beyond comparison of shared nucleotides.�
For scenarios in which sequencing error is less than individual distinctions arising from intraspecific differences, there is a range of viable species threshold values.�
At certain transformation values, accounting for further genetic changes inaccurately inflates species diversity. �

As a next step to this approach, we’re wanting to take data we’ve collected from soils from different environments out at some fieldsites we have in Utah and apply this delimitation technique to the microbes we find in those soils.

12 of 12

Acknowledgements

Ivan J. Palmblad Graduate Research Award, USU Biology Department�
Graduate Research Award, �USU Ecology Center �
Dr. Lawrence Piette Graduate Scholarship, USU College of Science�
Margaret Williams Research Grant, �Native Nevada Plant Society