Are four postulated disease spectra due to evolutionary trade-offs?

I recently read Crespi et al.’s interesting paper on this subject. They describe eight diseases as due to four underlying diametric sets that can be explained by evolutionary/genetic trade-offs:

  1. Autism spectrum vs psychotic-affective conditions
  2. Osteoarthritis vs osteoporosis
  3. Cancer vs neurodegenerative disorders
  4. Autoimmunity vs infectious disease

Of these, #2 and #4 seem obviously correct to me based on my fairly limited med school exposure, and they describe the evidence in a systematic way. I don’t know enough about the subject matter to speculate on #1, but I would like to see more genetic evidence.

Finally, I found their postulated explanations for #3 somewhat weak and I personally think that it is a selection bias trade-off, i.e. a case of Berkson’s bias as applied to trade-off. That is, since both cancer and neurodegeneration are age-related conditions, you could think of aging as the “agent” that selects either neurodegeneration or cancer as the ultimate cause of age-related death. I could be persuaded to change my mind on the basis of genetic predisposition evidence or some other mechanism, but I found the mechanism of apoptosis to be weak since apoptosis occurs (or doesn’t occur when it should) in many, many diseases, and moreover it is far from clear that neurodegeneration is mostly due to apoptosis as opposed to some other mechanism of cell death. A mechanism that might be most persuasive to me is one related to immune cells, since they clearly play a large role in regulating cancer growth, and also have high expression for the most GWAS risk factors for Alzheimer’s disease. But I still suspect that the selection bias is primary.

Twelve Interesting Recent Papers

1) Wootla et al. discussing naturally occurring antibodies for treatment of CNS disorders. Naturally occurring antibodies are mainly IgM and bind to many different types of antigens with low affinity (that’s what happens when you don’t do any affinity maturation). One idea is that elderly people without AD (but with, say, risk factors such as APOE) may have more of these antibodies, that help clear amyloid, and that’s why they haven’t developed AD. In fact, one of the more promising current treatments for AD in trials, aducanumab, was originally derived from elderly donors without AD based on this hypothesis. A similar procedure is also being done in MS — e.g., the authors describe some antibodies that bind specifically to oligodendrocytes with the goal of promoting remyelination.

2) Cummings et al. describing good phase II trial results for dextromethorphan + quinidine for agitation in AD. Aside from being excited about a potential new treatment for an aspect of AD, I find this particularly interesting since I was previously involved in a project that evaluated the effects of recreational doses of DXM in the comments of YouTube videos. However, the recreational doses are much higher than the doses in this study (> 200 mg vs 30 mg, respectively), so the effects are probably radically different — as always, the dose makes the poison.

3) A couple of papers recently came out purporting to explain the role of the ApoE risk variant in AD, which is very important but still very much unknown. First, a really interesting paper from Zhu et al. shows that in APOE ɛ4 carriers, synj1 expression increases, which decreases the expression of phospholipids such as PIP2. This is similar to an ApoE-null phenotype, suggesting a loss of function phenotype. Second, Cudaback et al. show that ApoE allele status affects the astrocyte secretion of the microglial chemotaxis factor CCl3. Interestingly, the ɛ4 and ɛ2 alleles have a more similar effect than ɛ3 in their data.

4) Turner et al. present results from an RCT of resveratrol for AD, which finds some good effects in biomarkers, but is not a home run clinically. Although with only 119 participants, it is likely underpowered, and one of the four clinical measures had a p = 0.03 effect in the correct direction.

5) Tom Fagan at AlzForum does nice reporting on results from PET and neuropathology showing that, by both measures, around 25% of people clinically diagnosed with AD do not have high amyloid levels. This is higher in ApoE e4 non-carriers, which is what you’d expect based on conditional probability and clinicians not taking into account ApoE allele status into account when making their diagnosis. In the absence of amyloid, neurodegeneration appears to be fairly slow or absent.

6) Dale Bredesen continues his innovative work in AD, describing here case reports suggesting that there are three types of AD, one inflammatory, one metabolic (e.g., related to insulin resistance), and one related to zinc deficiency.

7) Moran et al. use ADNI data to show that Type 2 Diabetes is associated with CSF tau (explaining 15% of the T2DM-associated cortical thickness loss), but not CSF amyloid, suggesting that T2DM might be related to tau-only AD cases, and/or tau increases that are independent of amyloid.

8) Not AD, but still neurological, in frontotemporal dementia, Ahmed et al. report that fasting blood levels of agouti-related peptide (AgRP) are much higher in patients (~66.5 +/- 85) than in controls (~23 +/- 20). Furher, AgRP levels are correlated with BMI, suggesting that AgRP levels account for the increased eating behavior seen in some variants of FTD. Just interesting to see an example where the effect of hormones on eating behavior could be very strong.

9) Petrovski et al. used WGS data to define an interesting measure of “how tolerant a gene’s regulatory region has been to mutation across evolution.” Specifically, their measure (the “noncoding Residual Variation Intolerance Score”) measures how many common variants a gene has in its regulatory region compared to other genes with a similar mutation rate. They found that higher levels of this measure were significantly associated with genes that are annotated as haploinsufficient, meaning that this is a good way of describing how much cells care about what relative expression levels a gene has.

10) Zheng et al. also used WGS data and found that rare variants in the gene EN1 are significantly associated with the risk of bone fracture. To quantify the effects of rare variants (< 5% MAF) they also used an association test — SKAT — to measure associations of these variants with bone marrow density in windows of 30 bp’s, and found one significant gene with this procedure. Refreshingly they put their code for this analysis online, available here, I haven’t ran it but just want to say +1 to them for putting their wrapper code online. Interestingly, both this paper and the Petrovski paper use the GERP++ score for their evolutionary inference — that seems to be a common tool, check it out here.

11) In influenza news, Lakdawala et al. show that influenza A does a large amount (most?) of its replication in the soft palate, which is the fleshy, soft part in the back of your mouth. Total hindsight bias, but this “makes sense” to me when I think back to the times when I think I had the flu myself — that part of my mouth gets very irritated, and now this makes slightly more sense.

How large of a difference is there between mouse and human inflammatory responses?

Attention conservation notice: Post about a 2014 PNAS paper discussing a 2013 PNAS paper (cited 449 times already!) for a journal club I was required to go to, which you’ve probably already heard of if you’re in the field, and probably don’t care about if you’re not. Also, I am not a geneticist.


The crux of both these papers is the following: let’s compare the genome-wide gene expression of humans in situations in which their inflammatory systems are likely to be going haywire (trauma, burns, sepsis with endotoxins) to the genome-wide gene expression of mice that are also in stressful situations in which their gene expression is likely to be pretty haywire-ish as well.

Here’s the problem: they come to exactly the opposite conclusions: the 2013 paper saying that the correspondence between the human and mice inflammatory signatures isn’t very good, and the 2014 paper saying that it is pretty good. And they use similar data sets.

So why the conflicting results?

1) One of the major differences is that the 2013 paper compared human genes that had been thresholded for significance in the condition of interest to all mice genes, whereas the 2014 paper thresholded both gene sets for significance prior to calculating the correlation. But why threshold for significance and then do correlations at all? It might be nice to threshold for some other characteristic of the genes, such as high variance, that is at least somewhat orthogonal to the actual correlation.

2) The 2013 paper uses Pearson correlations, whereas the 2014 paper uses Spearman rank correlations, making quite a brouhaha about how this is necessary because the data is not normally distributed via a KS test, and the Spearman measure is better in the case of non-normality. If this is so important, why not Kendall’s tau? But I am not convinced that it is: even in the most extreme cases on Wikipedia, the differences between Pearson and Spearman is only ~ 0.15 – 0.2, whereas the two studies found differences of around ~ 0.4 – 0.6. I bet #1 is more key.

Bottom line: The cynical take is that comparing mice to human gene expression patterns poses a large number of analysis quandaries, offering many free parameters for the researchers to draw conclusions of their preference. The more idealistic take is that authors and reviewers must be very careful in ensuring standard, robust methods in the analyses. Overall, I come down in favor of the 2014 paper, because they threshold mouse and human genes in the same way, which is more like comparing apples to apples.