Faculty Highlight: Dr. Rob Knight

Workshop on Genomics 2014 Faculty Highlight

Rob Knight
University of Colorado, Boulder
Knight Lab Website
Howard Hughes Medical Institute

It would be highly unusual for someone working in the field of metagenomics to not have heard of Dr. Rob Knight and his lab at the University of Colorado. His projects are incredibly diversified ranging from prairie soil to the ocean, toothbrushes to frog skin and includes work on viruses, ticks, fungi and archaea, even flatulence! He’s been at the forefront of program development for metagenomics including the wildly popular UniFrac and QIIME, but also; PyNAST, UCHIME, MIxS-BE, EMPeror (QIIME on an Iphone!), and Boulder ALE among many others. He’s been a champion of the American gut project, QIIME, Earth microbiome project, Mal-ED, Genomics Standards Consortium and the list goes on (more complete listing at the bottom of this post). His more recent high profile work/involvement has been a comment on Hsaio and colleagues work on the link between autism and gut microbes and the link between gut microbiota and obesity. Michael Pollan wrote a great article for the NY Times called “Some of my best friends are germs” about sequencing his gut microbiome and what it said about his health and body. He talks extensively with Rob about their work and how their lab has risen to the top of the field in metagenomics. I highly recommend the article, it’s a great read. For this blog I wanted to highlight one of the paper’s Rob was most proud of, though as you’ll see below he has many favorites among his publications, and then get a little bit into his involvements and suggestions in the realm of metagenomics and bioinformatic analysis in general.

…to the Research!

Establishing causal links between our bodies microbiota and a particular physiologic or disease manifestation (phenotype) has been notoriously challenging…Rob Knight was involved in a publication that was rising to that challenge. Ridaura and colleagues had a publication in Science, 2013: Gut microbiota from twins discordant for obesity modulate metabolism in mice. In the work the group took separate groups of germ-free mice and colonized them with the fecal microbiota from each member of four twin pairs discordant for obesity (meaning one fit the metrics for obesity and the other did not, yet they were twins). The mice were fed diets reflecting the upper or lower levels of consumption of saturated fats, and fruits and vegetables per the US National Health and Nutrition Examination Survey. Mice inoculated with the ‘obese’ twin’s microbiota were noted ‘obese (Ob)’ and those inoculated with the ‘lean’ microbiota of the lean twin were noted as ‘lean (Ln)’. They wanted to see if mice inoculated with the Ob microbiota would thereby become obese themselves and vice a versa, if the mice inoculated with the lean microbiota became ‘lean’. They also did a series of experiments to see if co-housing mice obese/obese, lean/lean or obese/lean affected the gut microbiota and, if you will, reversed the affects apparent in the obese mice or lean mice per whichever initial lean or obese microbiota they were inoculated with.

What they used:

  • fecal samples from twins (discordant for obesity), inoculated into mice
  • Analysis of 16S rRNA datasets: UniFrac
  • Shotgun sequencing
  • Comparisons of genes with assignable enzyme commission numbers (ECs)
  • ANOVA for statistics
  • Quantitative magnetic resonance analysis to asses body composition of mice
  • RNA-seq of transplanted microbial communities
  • Transcript mapping to database of human gut bacterial genomes
  • Transcripts assigned to Kyoto Encyclopedia of Genes and Genomes (KEGG) and ECs.
  • ShotgunFunctionalizeR: to define significant differences and distinguishing characteristics (based on poisson model).
  • Flourescence-activated cell sorting (FACS) to asses inflammatory response in mice colonized by either the Ob or Ln microbiota.
  • Tandem mass spectrometry (MS/MS) to analyze amino acids
  • Gas Chromotagraphy mass spectometry (GC-MS) to confirm metabolic process/pathway products.
  • They followed up with culture work to see if they could recreate the bacterial member composition that would transfer the Ob or Ln phenotype to the mice they inoculated with it.
  • Glucose tolerance testing.

What they found?

  • They were successfully able to efficiently and reproducibly capture the features of the human’s microbiota in the mice.
  • Ln mice exhibited higher expression of genes involved in digestion of plant-derived polysaccharides, fermentation of butyrate and propionate
  • Differences in body composition were correlated with differences in fermentation of short-chain fatty acids, (increased in Ln mice), metabolism of branched-chain amino acids (increased in Ob mice, and microbial transformation of bile acid species (increased in Ln mice and correlated with down-regulation of host farnesoid X receptor signaling).
  • Co-housing the mice obese/lean prevented development of increased adiposity and body mass in Ob mice and tranformed their microbiota’s metabolic profile to match that of the Ln mice.
  • The most successful ‘invader’ from the Ln microbiota that was found in the Ob mice after co-housing was Bacteroides.
  • Co-housing didn’t change the Ln microbiota, bacteria from the Ob mice gut were unable to successfully invade Ln mice.
  • When metatranscriptomes were analyzed, Ob mice that were cohoused with Ln mice were seen to be more similar to the Ln mice…consistent with the invasion of the Ln microbiota into the Ob mice.
  • Not ‘one’ bacterium can solve the obesity phenotype and reverse it; invasivness and adiposity are dependent on bacterial community context.
  • Diet is only part of the equation: Ob mice that were fed the Ln diet were still had significantly increased body mass as compared to Ln mice.
  • Gut microbiomes influence systemic lipid metabolism.

Take Home Points…

  • Bacteroides may be the clear winner in terms of ‘conquesting’ gut microbiota, but inevitably, ‘it takes a village’ within the gut to control obesity in mice.
  • Having all the right bacteria still doesn’t mean you are clear and free to eat what you want, there is a strong diet factor that influences the success of Ln invading bacteria. Anotherwards it’s a combination of the right ‘parts’ + ‘lifestyle change’ that will make the difference in obesity outcomes.

Looking into the glass ball for the future…

  • Future work can address which members of the microbial community are really involved in transmission of donor phenotypes.
  • Test the affects of diet ingredients on the microbiota-associated body composition and metabolic phenotypes
  • Assist in identifying next generation probiotics, prebotics or a combination of both.
  • Provide safer, sustainable alternative to fecal transplants for microbiome-directed therapeutics

Along with many other studies that Rob Knight has been involved in:


…and the list goes on; we can clearly look at Rob Knight through the super-hero adoring eyes of a 6 year old and call him ‘GutMan!’…or perhaps given the breadth of his collaborations…’MetaMan!’…though my personal preference may be GutMan.Picture2

Rob was kind enough to talk to me about his research, his views on metagenomics, RNA-seq and other bioinformatic related madness…

Do you have a ‘personal favorite’ among all your papers or studies conducted?

“Yes, it’s actually the Lozupone & Knight PNAS paper. This paper was the motivation for developing UniFrac, and inspired QIIME. Second place is a lot tougher; Costello et al. 2009 Science, Fierer et al. 2011 PNAS, Smith et al. 2013 Science, Ley et al. 2008 Science and Ridaura et al. 2013 Science all make very strong claims, for different reasons.”

Your 2013 paper Ridaura et al. employed RNA-seq which I’ve heard is the ‘wild west’ of sequencing nowadays and some scientist are eschewing it in favor of microarray analysis until tools can be better developed to quality control RNA-seq (ChIP-seq as well). They cite problems with replication and arbitrary thresholds etc…do you have any advice for those diving into the world of RNA-seq based on your experiences?

“Microarrays only work if you have the reference genome, which we don’t in the gut. There is a pretty good emerging literature showing that RNAseq is competitive with microarrays on cost and reproducibility while allowing you to discover new things. Of course, there are always people who will prefer more established technology; there are still people using Sanger sequencing, after all.”

  • Mel: We will hear more about expression studies during the workshop with Dr. Manuel Garber when he talks about transcriptomics (for last years presentation on transcriptomics from Manuel you can go to my blog from last year)

What do you feel is the most important consideration when conducting a metagenomics study? Or set of considerations?

“Experimental design. Do you have enough samples, and enough biological and technical replicates, to draw valid conclusions? A close second is: do you actually have an analysis plan for what to do with your data once you collect it?”

  •  Dr. Mike Zody will also be discussing NGS experimental design in great detail during his sessions at the workshop…so stay tuned.

Where do you see your labs focus going next? Do you think your projects will continue to stay diversified or has a particular topic captured your imagination for the future?

“We will stay diversified and continue to develop technology platforms that seek to transform what the research community as a whole can accomplish. There is far too much interesting work to be done for any one lab to do it.”

Your CV mentions biochemistry, microbiology, ecology and evolution. Michael Pollan in his article mentioned along the way you found you had a knack for computational biology. So you are a biologist by training and taught yourself the computational side?  How did you go about this? Many of our students are also teaching themselves and last year had many questions about how to turn a ‘biologist’ with little computing background into a ‘programmer/data scientist’ necessary for their research. Advice?

“I did take an introductory CS course in college (in Pascal, not so useful going forward) but mostly it was an iterative cycle of trying to solve problems I cared about, talking to other researchers who had faced and ideally solved similar problems, and reading books. I especially benefited from Steve Freeland, a postdoc in the lab where I was a grad student, who was an excellent mentor in programming in C and VBA, and from extended periods visiting Mike Yarus’s lab in Boulder, which provided the freedom to explore new techniques.”

  •  Mel: Rob is right…it is important to know that no man is an island, utilize resources and people around you to help in your development as a bioinformatician.

Do you have a preferred programming language? 

“I like python (especially with IPython) and vim. However, there are certainly specific applications where you’d want to use javascript, C, R, etc. Basically it’s about using the right tool for the job, although some tools turn out to be good for a surprisingly wide range of jobs.”

  • Python 1, Perl 0…and the debate rages on…

There are a great deal of aspiring scientists that look up to you as they create and re-create themselves…who was the defining force/mentor or if you prefer defining moment in your life that set you on the scientific path you are on now?

Mike Yarus, then Norm Pace, then more recently Jeff Gordon, George Church and Steve Quake.”


And finally…Just because it’s Fun!

Three words that you feel sum up your experience(s) with metagenomics?

Do microculture instead (if you’re referring to shotgun metagenomics).”

What is your favorite microbe/organism?

“I prefer to work at the community/ecosystem level.”

Dr. Rob Knight is currently involved in the following projects and we look forward to the insights and great programming that will inevitably flow from the Knight Lab.

We are most fortunate and look forward to hearing from Rob at the workshop…until then we say farewell GutMan! MetaMan! Pursuer of all things microbial…


…Dr. Mel