Evolution Conference 2013 - Sunday

I saw a long-tailed weasel near the main tent later in the day which was pretty exciting and Matt saw a moose in the parking lot that Kris and I ran down to try and find but couldn’t.

Talks-

Amy Dapper just presented some theory that the pattern of evolution of male reproductive genes is more consistent with a null of relaxed constraint rather than positive selection. It was quite interesting and based on a couple observations such as that sex-specific gene expression causes the selection coefficient to be reduced by half and that genes that select for highly competitive sperm are only actually beneficial when the female mates multiply (and with males who have different alleles at those loci). This causes the strength of selection to be scaled down even further by the harmonic mean of the number of mates. Her main result was to show that recently published dN/dS ratios that seem so high in male reproductive genes fall exactly on the null expectation of dN/dS that she calculated. It was interesting and she said that it should be published by the end of the summer. It may make a fun lab meeting paper to discuss once it's out.

David Gokhman presented new data on the epigenome (methylome specifically) of the neandertal and denisovan. He used the natural degradation of C->U and Cmeth->T to determine which bases in ancient dna was methylated. found that of the regions differentially methylated between humans and neandertals (Hox8,9,10) all control bone growth and expression correlates with patterns of neandertal morphology and known human pathology. Really cool.

Went to Zach Gompert’s talk - way over my head. I followed the selection equation parts and then once he actually got into the model I lost it.

Also attended Joan Roughgarden’s talk. Was not impressed. 

Matt Hahn discussed genomic islands of speciation and described how the divergence in these islands - when measured in an absolute manner, is incredibly low. The point is that while there is likely something - selection possibly -  driving the patterns of Fst that we identify as “islands” it does not mean that there is speciation with gene flow occurring. 

Dolph Schulter’s ASN presidential talk was interesting. He talked about the latitude-speciation correlation and if it exists or not, decided that ecological opportunity (which we can’t really measure) is really what drives everything. Talked about how the stickleback fish species are the result of repeated fixation of standing variation and how that doesn’t align with the standard thoughts about DMI’s - he has wide-scale additivity across many genes that seems to isolate species. Claimed that it was much different than the classic DMI model.

Evolution Conference 2013 - Saturday

Kris and I arrived late Friday night. It’s good that we had Siri to navigate for us as we would have been completely lost otherwise. Turns out that Brice got one hell of an awesome mansion for us to live in for the next couple days.  

Driving in in the morning, we saw a coyote - great way to start a biology conference. 

We got a bit lost on the way in - it was not immediately obvious where the conference actually was, but we finally found it and got registered.

Here’s some talks that I went to that really stood out:

Daniel Matute - great, talked fast and I am not 100% familiar with all the molecular techniques that you can use in drosophila, so I got a little lost in the end. He is using deficiency mapping technique to find the genes that isolate melanogaster/yakuba (I think it was yakuba, but now that I write it I’m not really sure). Inviability depends on x-x interactions as well as a 3rd locus somewhere else. Should have typed this yesterday directly after the talk as I can’t remember the specific details. I talked with him later and it turns out that he’ll be starting as a professor soon and is looking for grad students.

Robin Hopkins discussed reinforcement among flowers in texas: blue in most of the range but red in the area of overlap. She had me pretty convinced that it was reinforcement: hybrids show really low fitness, but are common when flower color is the same between the two species.  Also got into the mechanism of reinforcement: the butterfly pollinators have a preference for flowers of the same color - some like red, some like blue, but the ones that like red always go to red and the ones that like blue always go to blue. The one thing that I was confused about is that the preferences I just described operate in the area of sympatry, but not in the area of allopatry where pollinators have a distinct preference for blue. I’m not sure why it would be different as it’s the same species of butterfly that does the all pollinating. She also got into the genetics of flower color change from light blue to dark red, it’s a 2-locus, each with complete dominance (the standard 9:3:3:1 ratio). One loci controls brightness, one does red/blue. Yaniv asked afterward it it may be collapse rather than reinforcement though, but I had to run to the next talk and didn’t hear what she said. I’ll try to talk to Yaniv about it.

Matt Jones’ talk was great, it will be fantastic to have him in the lab.

Amanda Moehring claimed to have found the first speciation gene for a behavioral trait in melanogaster but she didn't actually present any data, only asserted that it had an effect. I found it really frustrating  as she showed images of her deletion mapping, but not any actual data about what the region she found actually does. Many of the speciation people were there (Matute, Noor, Yasir) just shaking their heads at her the whole time. It was one of these  extraordinary  claims that have no evidence supporting it. I talked to both Daniel Matute and Yasir about it later and neither were impressed. I’ll be looking up the papers and possibly writing more about it later.

I realized that I printed the wrong version of my poster and had to skip the poster session to re-print it, but I ended up getting together with Yasir and Erica and Ryan and Matt for some beers later.

Great first day.

Diluting and pooling - last step

I have now prepared and amplified all my samples. I used 12 cycles and 8ul of product so that I wouldn't blow through the entire 20ul of library if something went wrong. Most of these had plenty of product, the couple that didn't I re-amplified for 14 cycles. For the dilution, I used the values from the nanodrop rather than the bioanalyzer as Sara has had some issues with pooling based on bioanalyzer values.

Here are the bioanalyzer traces of the completed pools:

The final concentration is a little low, but as the sequencing facility only needs ~10ul of 10nmol/L it should be fine. I actually ended up making 4 pools as the first two didn't stack up well on the qPCR. I think this was because I was pipetting such small volumes that the liquid was evaporating and changing my concentration. For the second set (Pools 3 and 4, seen above) I made sure that my sample volumes were never less than ~10ul. It seemed to work much better. Here's the qPCR trace of 5 of my samples. They should all be right on top of each other (except for the light blue one which is the no-template-control:

There is only about 1 cycle spread from the first one to the last, whereas the pools 1 and 2 had at least a 4 cycle spread. 

S~ is now setting up the account at Utah and I'll be sending the samples as soon as all the paperwork goes through. (We've had some issues with sequencing at Berkeley and so we're gonna try a new facility.)

Epigenetics and a new direction

I have been keeping an eye on a couple other science blogs - homologus and Judge Starling among them, and I have decided that I should include among my descriptions of experiments an occasional review of some bit of literature that I find interesting. And so:

PNAS recently published two papers: Epigenetics, a "Core Concepts" piece by Sarah Williams (2013) and Epigenetics: Core Misconcept by Mark Ptashne (2013) where he argues that Williams is perpetuating some of the main misconceptions about epigenetics. In order to analyze this little spat, I will first define epigenetics and then analyze what each of the authors say and whether Williams is indeed perpetuating misconceptions. This disagreement boils down to another fundamental argument over whether DNA methylation and/or histone placement should be considered "epigenetic mechanisms" and I will conclude by pitching in my two cents in this debate.

Here are a couple of current definitions of epigenetics from the literature:

From Bird (2007): "... epigenetic events... [are] the structural adaptation of chromosomal regions so as to register, signal or perpetuate altered activity states."

From Eccleston et al (2007): "Epigenetics is typically defined as the study of heritable changes in gene expression that are not due to changes in DNA sequence."

From Sasaki et al (2008): "Epigenetics refers to a collection of mechanisms and phenomena that define the phenotype of a cell without affecting the genotype"

From Bonasio et al (2010): "Epigenetic signals are responsible for the establishment, maintenance, and reversal of metastable transcriptional states that are fundamental for the cell’s ability to “remember” past events, such as changes in the external environment or developmental cues."

These definitions all agree that epigenetics does not include changes to the base sequence of DNA, but rather changes in the expression level of different genes. Said another way, epigenetic changes must be reversible: coding changes are not reversible (a back mutation is quite rare) while expression changes are (it's relatively easy to turn genes on and off). These definitions also all agree that these expression changes must be heritable from one generation to the next (it should be noted that this is most often cell generations i.e.: heritable through mitosis, rather than organism generations, though both are applicable).

One other condition is also commonly found in the literature and is necessary for an epigenetic system: it must be self-perpetuating (Riddihough and Zahn 2010; Bonasio et al 2010). Self-perpetuating means that there is a positive feedback loop somewhere such that once it is initiated and even in the absence of the initial signal, the new expression level will persist.

In summary: there are three characteristics that are necessary for a system to be labeled "epigenetic:"
1.) It must be reversible by dealing with changes in expression level not sequence identity.
2.) It must be heritable, at least through mitosis, possibly through meiosis and fertilization.
3.) It must be self-perpetuating in the absence of the initial signal.

The predominate biological process that this refers to is development where a single cell divides and differentiates into a large number of distinct tissue types. At the end of the process, each cell in every tissue has the complete genome of the organism, but each cell type has it's own specific set of genes that is expressed. This is one of the most incredible - but rarely-recognized - cases of plasticity: a single set of genes gives rise to a vast number of different cell types and functions. Epigenetics is concerned with how, for example, a liver cells divides into two new liver cells. Certain genes are on in each liver cell that, when present in the daughter cells, cue those to also express the suite of "liver genes." Thus, this aspect of development complies with the heritable and self-perpetuating requirements of epigenetics. It is also reversible as no coding changes have occurred - only expression levels change, not the DNA sequence. We know this is true because the entire genome is present in every cell - it is certainly NOT the case the all of the "non-liver genes" are destroyed or excluded from future liver cells. This is depicted in the following diagram (from Ptashne 2013):

The blue circle protein (a transcription factor) activates the gene for the red oval protein (also a transcription factor) by binding to a cis-regulatory element (blue box). The red protein activates itself by binding to the red cis-regulatory element and also activates a suite of other green genes by binding to their cis-regulatory elements (red boxes). 

So what are some concrete examples of epigenetics? The prime example is a transcription factor that activates the expression of itself (self-perpetuating) and also activates the expression (reversible) of a suite of other genes as in the above figure. When this transcription factor is in the cytoplasm of a cell undergoing the cytokinesis phase of mitosis, it will be allocated into the two daughter cells (heritable) and will thus cause them to express a similar suite of genes as the parent cell. Another common example is prions. Prions are poteins with a specific tertiary conformation (reversible) that cause other proteins (self-perpetuating) to assume the same confirmation (heritable). Though prions seem to me to be in a slightly different class of epigenetics than transcription factors as prions don't involve DNA at all. This is not to say that prions are not epigenetic - they most certainly are - but rather that the term epigenetic is quite broad. I will from here on concentrate on the subset of epigenetics that deals directly with gene expression.

Moving on to Williams' (2013) Core Concept article, I found it to be a bit brief but not horribly off the mark. Williams focused on only two types of epigenetic inheritance: DNA methylation and histone placement and she did not go into detail about the mechanism by which these epigenetic changes occur. Her argument mainly concerned the reversible aspect of epigenetics: the fact that differential methylation (or differential histone placement) results in differential gene expression. She also asserted that these 'flags' on a bit of DNA are copied along with the rest of the genome during the S phase of the cell cycle and are therefore heritable. She did not directly address whether methylation or histone modifications are self-perpetuating, though the argument is that due to the action of methyltransferases the methylated sites are indeed perpetuated, and hence the expression profile of that site is maintained.  Specifically, De Novo Methyltrasnferase 1 (DNMT1) finds DNA where a methyl group is on one strand but not the other (as woud occur durring DNA synthesis) and it methylates the second strand. This results in two daughter cells each with accurately-reproduced methylation patterns and hence similar expression profiles as the parent cell.

Ptashne (2013) takes issue however with Williams' (2013) description of epigenetics. In doing so he puts forth two criteria which together are both necessary and sufficient to call something "epigenetic": these are memory and specificity. Ptashne's "memory" is a combination of the earlier "self-perpetuating" and "heritable" - a distinction that often becomes blurred (for instance: see the preceding paragraph - my argument for heritable and self-perpetuating are actually the same). As such I agree with Ptashne (2013) that the more general term of "memory" is useful. Technically, memory is the continuation of a specific expression profile of a given gene. Under this criterion DNA methylation is an epigenetic mechanism as DNMT1 faithfully transmits the signal to new molecules. On the other hand, histone placements are not known to be heritable or self-perpetuate, even though they do seem to correlate with gene expression, and so should not be considered "epigenetic" according to Ptashne (2007, 2013).

Both DNA methylation and histone placement also fails Ptashne's (2013) second test of specificity. Specificity has to do with which genes a give transcription factor acts upon: for instance in the earlier figure, the red circle only acts on the blue oval's regulatory region, NOT on any of the geen genes' regulatory regions. The opposite of something being specific is for it to be general in which case it acts ubiquitously across the genome. This is the sense in which DNMT1 fails the test: any time it finds a bit of hemi-methylated DNA, it methylates the unmethlyated strand regardless of location. Histone placement also fails this test, as according to Ptashne (2007) we have no evidence that histones are placed specifically. Ptashne states: "... it is said [that] chemical modifications to DNA... drive gene regulation. This obviously cannot be true because the enzymes that impose such modifications lack the essential specificity: All nucleosomes, for example, “look alike,” and so these enzymes would have no way, on their own, of specifying which genes to regulate under any given set of conditions." (Ptashne 2013). Clearly Ptashne takes issue with Williams' claim that DNA methylation and histone modifications are epigenetic mechanisms.

I disagree with Ptashne, and think that histone modifications and DNA methylation should be considered epigenetic. I will discuss each of these in turn, first DNA methylation. To be fair, Ptashne (2013) does opt-out of making a call on the methylation side of things, citing that its role in development remains unclear because it doesn't occur in the flies or worms - which he seems to consider the two "real" models for development. I can only conclude that Ptashne is slightly ignorant of the current findings in mammal models (yes mammal models for development do exist, even if Ptashne doesn't consider them "real").  Sasaki et al published an entire review in Nature Reviews Genetics five years ago (Sasaki 2008) that describes most of how trans-generational epigenetic events are established in the germ-line. One of the main things that happens during germ cell development is the erasure and reestablishment of the parental imprint. Imprinting is regulated by DNA methylation in the imprinting control region and fits both of Ptashne's requirements: memory because the methylation established at this point in development regulates gene expression throughout life, and specificity because there are only a subset of genes that are imprinted - it's not a ubiquitous genome-wide process. So contrary to what Ptashne (2013) claims, we do know plenty about DNA methylation in mammalian development and furthermore it does fit both of his requirements. That being said, we still do not have a solid understanding of exactly how the specificity is achieved - there is not a DNMT specific for each region that needs to be methylated (indeed there are only 3 DNMTs that actively methylate DNA). Presumably there is some type of factor that is specific to certain regions and recruits the DNMTs to specific areas. DNMT3L may fill this role as it does not actually methylate DNA, but instead complexes with DNMT3a and facilitates proper methylation (Sasaki 2008). This DNMT3a-DNMT3L complex also relies upon a specific histone (H3K4 allows methylation while H3K4me does not)(Sasaki 2008), and so specificity may be achieved due to an interaction between local histone type and the DNMT complex. In short DNA methylation certainly counts as an epigenetic mechanism of transcription control.

Ptashne (2007, 2013) also claims that we have no evidence that the pattern of histones is heritable and so should not be considered epigenetic. On the contrary, it is true that allele-specific histone patterns are conserved within tissues and so must be heritable. For example, histones (not DNA methylation in this case) are the primary drivers of imprinted expression of the Kcnq1 region of Chromosome 7 in mice (Umlauf et al 2004; Lewis et al 2004). This shows that histones do fit the specificity criterion - they are allele specific, as well as the memory criterion - imprinting is tissue-specific (and *naturally* allele-specific). As far as I can tell, we do not know how this specificity is achieved, but just because we don't  know how it's achieved, doesn't mean that it is not indeed specific.

In conclusion, I consider both differential DNA methylation and differential histone placement to be true epigenetic mechanisms of gene regulation as both of these processes are heritable, reversible, self-perpetuating, and specific.

----------------------------------------------------------------------------------------------------------------------------------
Bird, A. 2007. Perceptions of epigenetics. Nature 447:396–398.

Bonasio, R., S. Tu, and D. Reinberg. 2010. Molecular Signals of Epigenetic States. Science 330:612–616.

Eccleston, A., N. DeWitt, C. Gunter, B. Marte, and D. Nath. 2007. Epigenetics. Nature 447:395.

Lewis, A., K. Mitsuya, D. Umlauf, P. Smith, W. Dean, J. Walter, M. Higgins, R. Feil, and W. Reik. 2004. Imprinting on distal chromosome 7 in the placenta involves repressive histone methylation independent of DNA methylation. Nature Genetics 36:1291–1295.

Ptashne, M. 2007. On the use of the word “epigenetic.” Current Biology 17:R233–R236.

Ptashne, M. 2013. Epigenetics: Core misconcept. Proceedings of the National Academy of Sciences 110:7101–7103.

Riddihough, G., and L. M. Zahn. 2010. What is epigenetics? Science 330:611.

Sasaki, H., and Y. Matsui. 2008. Epigenetic events in mammalian germ-cell development: reprogramming and beyond. Nature Reviews Genetics 2008:129–140.

Umlauf, D., Y. Goto, R. Cao, F. Cerqueira, A. Wagschal, Y. Zhang, and R. Feil. 2004. Imprinting along the Kcnq1 domain on mouse chromosome 7 involves repressive histone methylation and recruitment of Polycomb group complexes. Nature Genetics 36:1296–1300.

Williams, S. 2013. Epigenetics. Proceedings of the National Academy of Sciences 110:3209.

Placental Histology

Another experiment that I'm pursuing in collaboration with L~ (an undergraduate here at UM) is to analyze placental histology of hybrid hamsters. She found that there seems to be some striking abnormalities in the layering of the hybrids' placentas. Placentas are very complex and formed from three main embryonic tissue layers (labyrinthine trophoblast, spongiotrophoblast, and trophoblast giant cells) as well as a maternal layer (decidua).

***This figure is taken from Wagschal and Feil (2006) - a great paper on imprinting and the placenta***

So far L~ has found that there seems to be a sex-specific placental abnormalities in the hybrids - only the female hybrids show major disruptions of the boundaries between the layers, especially between the labyrinthine trophoblast and spongiotrophoblast. This is quite exciting as I can find nowhere in the published literature that reports sex-specific hybrid phenotypes. It also complicates the over/under-growth story because overgrowth assorts by cross type, not sex while undergrowth is only found in males, not females. If we assume that abnormal layering of the placenta results in abnormal nutrient transfer, then L~'s defects do not explain the patterns of over/under-growth.

However, the issue we have is that the sample size is quite low - 7 males and 2 females total from both hybrid types. We are now curious if the pattern that both females show abnormal morphology is true or only an artifact of small sample size. To address this, I have just set up 10 more crosses (expectation is ~50 offspring/placentas) that L~ and I will be dissecting and doing histology to next week.

I hope to be able to present the findings at the Evolution Conference in Salt Lake in a couple months.

6-4-13 update: So far I have dissected 5 Sun x Cam (overgrown) crosses and 1 Cam x Sun (undergrown) cross resulting in 10 placentas (7 Sun x Cam and 3 Cam x Sun). I fixed these in 4%PFA overnight and used the histology lab's tissue processer to embed them in parafin. Yesterday I sectioned them and put them on slides. Now I need to stain them and then measure them.  I had forgotten how difficult and touchy sectioning is. Major props to L~ for all the sectioning she did on the earlier samples.

6-18-13 update: I have dissected everything, embedded them, sectioned them, stained them, and imaged them. I'm going to present the results at Evolution this weekend and I'll type up a post afterwards.

----------------------------------------------------------------------------------------------------------------------------------
Wagschal, A. and R. Feil. 2006. Genomic imprinting in the placenta. Cytogenetic Genome Research 113:90–98.

PCR amplification and the number of cycles

I had earlier decided that amplifying 8ul for 9 cycles would be good for my libraries. It turned out that 9 cycles was much to few and I increased the cycle number to 12 as per J~'s suggestion. I have now amplified all 40 samples at 12 cycles (still 8ul) and have just finished bioanalyzing the results. It seems that 12 has worked great for many of my samples, but has under-amplified a few (see below).

Of these samples, the first 4 are under-amplified, the middle 4 are fine, and the last 4 are perfect. None are over-amplified which is great. 

Next I need to pool all of these samples into two lanes as per my earlier plans. The pooling will be done so that there are equal molar ratios of each sample in the final mix. The issue I am having is that J~ and I had talked earlier and said that I should amplify each sample the same number of cycles (12) but at this point, 12 cycles has under-amplified some of them and not others. One option is to amplify the low samples a couple times more to bring them up to par with the rest. I am unsure of whether treating some samples differently will cause any catastrophic failures of my experiment. It seems like it will not as the expression analysis will be standardized by the relative expression of a number of housekeeping genes. I would guess that this should remove any effects of cycle number, but I'm not sure. I'll talk to J~ and report back.

Below is a plot of all the samples spread along the X-axis and the final concentration (ng/ul) that the bioanalyzer reported on the Y. The different colors represent different batches of library prep (batch size = 8) and the different shapes represent the different RNA extraction batches (batch size = 4). The horizontal line is at 5ng/ul.

It looks like there are no batch effects for the most part. The one main exception is in the red batch where the squares (males) are lower than the circles (females). Earlier, I had started with unequal amounts of RNA for that batch and I ended up re-doing the males. It seems that something went wrong with the 2nd prep of those, and I need to re-prep them again. As long as I am at it I will also re-prep some of the others that are below 5ng/ul. 

Cycle number in the TruSeq protocol

The qPCR results showed that 13-14 cycles would be excellent for my libraries (see this post). However, the amount of library that I started with was 1 40th of the total library (i.e. 0.5uL). If 0.5ul achieves the exponential phase in 13 cycles, then 8uL should achieve exponential phase in 9 cycles:
starting      cycles to
   ul         exponential
0.5              13
1                 12
2                 11
4                 10
8                  9
16                8

Thus I used 8uL of my libraries and amplified them for 9 cycles. This resulted in some libraries that were perfectly amplified and some that were not amplified enough. The top sample is great, a small even peak (and no secondary hump under the 1500 marker). The second sample does have a peak in the right area, but hasn't been amplified enough - there is not enough library there to sequence.

Fortunately, I only used 8uL of my pre-amplification library, not the entire 20uL just in case this issue arose. So now, instead of going back and re-preparing the entire library, I can just take and amplify the remaining parts. 

I'll need to talk to J~, as there are some serious batch effects (the first 8 all look like the top image, while the second 8 all look like the bottom image) but it looks like 8ul at 11 or 12 cycles will work fine for most everything without over-amplifying the library. As I will be comparing transcript levels, it is important that all the libraries are amplified the same number of cycles. 

Over-amplification is also bad because at that point in the PCR, the DNA molecules are being heated and cooled without being replicated. Thus they will fall apart and re-anneal repeatedly. They will often re-anneal with a DNA of a different sequence. This happens because the adapters have the same sequence, while the center sequences vary, two different molecules can meet eachother and the adapters on the ends will anneal while the middle region will for a big loop. This will mess up the sequencing machine and you'll get bad results. You know if you've over-amplified a library because there will be a low hump that around the 1500 marker. This is the weird hybridized molecules which travel slowly through a gel and so look quite large. In fact they are not any larger than the rest of the library and so can't be selectively removed via their size. The only thing that remains at that point is to re-amplify the library.

In short, I need my samples amplified more, but not too much more, so a cycle or two should be fine. J~ will be able to help me out with that.