Long Term Vegetarian Diets, Disease, and Stupid Headlines

(For my TL;DR'ers, the study in question is looking at population level genetic variation, with a focus on an insertion in the FADS2 gene that appears to have been under positive selection in traditionally vegetarian populations. The paper posits some potential functional consequences of this variant related to increased levels of polyunsaturated fatty acids that are implicated in several disease processes. It has nothing to with an individual's choice to be vegetarian in their own lifetime; it has to do with your ancestors' dietary decision and the genetic variants that you've inherited. Any potential risks associated with the variant likely depend on the amount of dietary precursor available for the enzyme to act on but further research is required from randomized controlled trials.)

This week's annoyingly misleading headlines tout the 'harms' of a long term vegetarian diet; some outlets had better headlines than others, but after seeing enough misleading interpretations of the study involved, I figured I'd cover what it does and does not suggest. 

http://www.dailymail.co.uk/health/article-3515293/Vegetarian-diet-raises-risk-heart-disease-cancer.html

Even the studies main authors' on twitter were a bit annoyed at the interpretation:

https://twitter.com/yekaixiong/status/715976627713359878

The actual study can be found in the journal Molecular Biology and Evolution and is titled 'positive selection on a regulatory insertion-deletion polymorphism in FADS2 influences apparent endogenous synthesis of arachidonic acid'; several of its main authors are from my home university, Cornell**. As you can see, the title says nothing about vegetarians dying a lot more. 

What the paper actually discusses is evidence for positive selection on an indel in the fatty acid desaturase 2 (FADS2) gene in populations that have been traditionally vegetarian vs those with more mixed diets (including animal products). Some background - humans require 2 'essential' fatty acids: linoleic and alpha-linolenic acid. The body enzymatically enlongates and desaturates these precursors to long chain polyunsaturated fatty acids (LCPUFA), like arachidonic acid, eicosapentanoic acid, and docosahexaenoic acid (which are arguably the truly essential things). However, these longer, more unsaturated fats can also be found in the diet, in animal foods (animals can eat the precursors and do the elongation/desaturation for us). Populations that don't consume animals rely on endogenous synthesis to make LCPUFA. To make these LCPUFA, mammals rely on different desaturases and elongases; the focus of this paper is on fatty acid desaturase 2, the rate limiting step in LCPUFA synthesis. The authors previously identified  a 22 base pair insertion/deletion (rs66698963)) near the sterol response element in FADS2; homozygosity for the deletion (DD)in vitro is associated with lower FADS1 expression and lower levels of its immediate product, arachidonic acid. In this paper, the authors present hypotheses and supporting data for 1) the idea that genotypes that confer greater conversion of precursors to LCPUFA have been driven by selective pressures to a higher frequency in populations that don't traditionally eat much preformed LCPUFA e.g. vegetarian populations  and 2) that individuals with different genotypes at this locus will have altered substrate and product fatty acid levels.

To test these hypotheses, the authors take to a dataset of U.S. and Indians, as well as the thousand genomes project data to look at the distributions of this genotype in different populations. The frequency of the D allele (homos and hets) in the U.S. was .62 and .18 in India; the overall frequencies deviated from Hardy Weinberg equilibrium. When looking at the 1000 genomes dataset, the authors find that South Asians and Africans had the lowest frequencies of the D allele and higher frequencies of homozygosity for the insertion allele (I), whereas Europeans and East Asians had lower frequencies of I/I. Noting these differences in frequency by geography, the authors calculate Fst values, a common metric in population genetics that quantifies the degree of allele frequency differences across populations; their calculations reveal a moderate level of genetic differentiation when comparing all 4 populations. However, when performing pairwise calculations, high levels of genetic differentiation were determined between South Asians and Europeans. South Asians also showed significantly higher levels of positive selection based on haplotype scoring, with the haplotype carrying the insertion showing a high frequency in this population; Africans and East Asians also showed some evidence of positive selection. The authors looked further into different populations of East Asians since divergence and haplotype testing produced conflicting results for evidence of selection, and find support for selection only in the Han Chinese populations from Beijing and Japanese (Tokyo). The authors conclude that the insertion allele (I) at this loci, or a variant nearby that it is in linkage disequilibrium with, are under positive selection in these (traditionally vegetarian) populations. 

To address potential functional outcomes of different genotypes at this FADS locus, the authors genotype individuals from their Kansas cohort (n=199) and measure their red blood cell fatty acid concentrations. The authors correlate genotype with differing levels of precursor and LCPUFA products, observing some consistency with their expectations based on their in vitro work. Individuals who are homozygous for the D allele show the highest levels of 20:3n-6, the substrate for FADS1 and immediate precursor to arachidonic acid, 20:4n-6. Individuals with the I/I genotype have higher levels of 20:4n-6, suggesting greater conversion of 20:3n-6 to 20:4n-6. Linoleic acid levels were not different between genotypes. 

So where are these headlines coming from? The cancer and heart disease claims come from the fact that Arachidonic Acid metabolism is a pharmacological target in aspects of these disease processes; AA is the precursor for a number of eicosanoids (many of which are pro-inflammatory, some that are anti-inflammatory) and have been implicated in cardiovascular disease and certain cancers (e.g. colorectal). It has been hypothesized that higher basal levels of AA might confer higher susceptibility to these diseases, but this isn't totally clear. Further analyses looking at this variant and its association with lifelong cancer/CVD risk are needed. 

How do we interpret this data? For one, the fatty acid analyses, while consistent with the in vitro work, result from observational data that isn't well controlled. Future trial evidence that randomizes individuals with these genotypes to differing high/low levels of linoleic and alpha-linolenic acid will further elucidate the effects of these genotypes on PUFA metabolism. Since general recommendations from groups like AHA focus on consuming preformed LC n-3 PUFAs, it will be interested to see whether individuals with I/I genotypes have differing responses to the arachidonic-acid displacing effects of eicosapentanoic acid. While the headlines raise a lot of concern about the higher AA levels associated with the I/I genotype seen in this study, the functional consequences of this slightly higher basal level (~8percent higher) is not known; again, large population studies correlating this genotype with some outcome are needed.

The most interesting interpretation for me is that this paper provides further evidence that diet has acted as a selective pressure over time and that there is no one ideal human diet #paleo. 

How should we not interpret this data? Eating a vegetarian diet will kill/harm you. 

** full disclosure: JTB is on my dissertation committee.