Evidence for the Evolutionary Model

Jonathan M, a contributor to such sites as evolutionnews.org and uncommondescent.com, has recently responded to my formulation of the ERV argument for common ancestry. He announced in this blog post on the Discovery Institute's “Evolution News & Views” site that he intended on addressing each one of the three “layers of ERV evidence,” over the course of additional blog posts. His first post regards layer 1, and his second post regards layers 2 & 3. Here, I have responded to each of his two subsequent posts.

Post #1: Do Shared ERVs Support Common Ancestry?

Layer 1: Sharing and Hierarchical Distribution

He began his response with the following:

Just how target-specific are these ERV integrations? In the portion of the article headed "common creationist responses," we are told that,

...while proviral insertion is not purely random, it is also not locus specific; due to the way it directly attacks the 5' and 3' phosphodiester bonds, with no need to ligate (Skinner et al., 2001). So relative to pure randomness, insertion is non-random, but relative to locus specificity, insertion is highly random.

Really? Let's take a few moments to do what any good student of biology would do -- and briefly survey some of the literature.

I will break my response into three parts. In the first part, I will address the portions that are not relevant to target site selection, and explain why they are not relevant; in the second part, I will address the portions that are relevant to target site selection; and in the third part, I will address the portion regarding the hierarchical distribution of ERVs.

Not Relevant to Target Site Selection

Jonathan M. first brought up the independent acquisition of hominoid syncytin-1/2 and murid syncytin-A/B. It appears that he got the impression from the abstract of Dupressoir et al. (2005)—and its usage of the term “independently acquired”—that these insertions were in identical loci; thus making them relevant in a discussion of target site preference. In actuality, however, all four of these retrovirus-derived genes are at very different loci; they are not even on the same chromosomes. Syncytin-1 is on chromosome 7, syncytin-2 is on chromosome 6, syncytin-A is on chromosome 5, and syncytin-B is on chromosome 14 (Entrez Gene, n.d.). That they were co-opted for the same function is relevant to a discussion of functionality, but is not relevant here. For my response to the argument from functionality refer to the “Common creationist responses” section of my ERV essay.

He then quoted from the abstract of Barbulescu et al. (2001), regarding HERV-K-GC1, before providing the following:

It seems that the most plausible explanation for this is an independent insert in the gorilla and chimpanzee lineages. Notice that the intact preintegration site at the pertinent locus in humans precludes the possibility of the HERV-K provirus having been inserted into the genome of the common ancestor of humans, chimpanzees and gorillas, and subsequently lost from the human genome by processes of genetic recombination. The inserts in the chimpanzee and gorilla lineages must be independent events.

This response not only ignores what I wrote about HERV-K-GC1, but also ignores what Dr. Barbulescu and her colleagues wrote in the body of that very paper. Below, I will restate a portion of what I stated in the “Common creationist responses” section of my ERV essay:

Since [HERV-K-GC1] is an unrepresentative deviation from a pattern, all that needs to be done is identify mechanisms that can account for that deviation. There are at least two known mechanisms to account for it (Barbulescu et al., 2001):

1) The insertion was in an allele that remained heterozygous in the populations across two divergences, given genetic drift and the divergent sub-population size (allelic segregation).

2) The insertion was in a duplicate section of the chromosome that underwent homologous recombination in each respective linage.

Later, he wrote the following:

There is also mounting evidence that, in closely-related species, the same ERV exhibits markedly different patterns of transcription. For example, Yohn et al. (2005) report that,

Based on analysis of finished BAC chimpanzee genome sequence, we characterize a retroviral element (Pan troglodytes endogenous retrovirus 1 [PTERV1]) that has become integrated in the germline of African great ape and Old World monkey species but is absent from humans and Asian ape genomes ... Six out of ten of these genes, for which there are expression data, show significant differences in transcript expression between human and chimpanzee [emphasis added].

First, it is important to note that the PtERVs in question are present in chimpanzees, yet not in humans, as indicated even in the portion of the Yohn et al. (2005) abstract, from which he was quoting. Second, the portion of that abstract that he left out is important in understanding what the authors are saying. Here is a more complete quotation:

Within the chimpanzee, there is a significant integration bias against genes, with only 14 of these insertions mapping within intronic regions. Six out of ten of these genes, for which there are expression data, show significant differences in transcript expression between human and chimpanzee. Our data are consistent with a retroviral infection that bombarded the genomes of chimpanzees and gorillas independently and concurrently, 3–4 million years ago. We speculate on the potential impact of such recent events on the evolution of humans and great apes (Yohn et al., 2005, p.0577).

The authors make it very clear that the differences in gene expression are not between genes with “the same ERV;” but rather, between chimpanzee genes with intronic retroviral insertions and corresponding human genes without retroviral insertions. What this demonstrates is that retroviral insertion can affect gene expression—something I discussed in the “Common creationist responses” section of my ERV essay—but does not relate to target site preference.

Relevant to Target Site Selection

Eventually, he goes on to bring up a number of papers that demonstrate target site preference. And that is exactly what I demonstrated in my ERV essay; that retroviruses insert in various genomic reigns with higher and lower frequencies than would be expected from a random distribution.

For example, most of the discussion by Bushman (2003) used the same data from Schröder et al. (2002) and Wu et al. (2003) that was included in Mitchell et al. (2004); both discussed various areas of higher and lower frequency, such as “chromosomal regions rich in expressed genes... CpG islands... active genes... transcription units... [and areas] near transcription start sites.”

In addition, Sverdlov (2000) reports,

But although this concept of retrovirus selectivity is currently prevailing, practically all genomic regions were reported to be used as primary integration targets, however, with different preferences. There were identified 'hot spots' containing integration sites used up to 280 times more frequently than predicted mathematically. [emphasis added]

Just as with the other examples, rather than demonstrating locus specificity, what is actually being demonstrated is the ability to insert anywhere in the genome, as well as certain areas of higher and lower frequency of insertions, i.e. target site preference.

This is clearly illustrated throughout source the Sverdlov (1998) paper itself (yes, 1998; read the aside) gets the information in the quote from; as the following quotations show:

In this report we extend the PCR-based approach to examine integration within chromosomal targets in vivo. These studies were initiated to test and extend our previous work on the specificity of ALV integration into cell DNA (Shih et al. 1988). For this purpose, we developed a system that detects single integration events in a population of cells at any defined genomic site with single nucleotide resolution. Using this approach we have found that all regions of the genome examined were accessible to retroviral integration. Localized preferences were seen within regions, with some sites being used up to 280 times greater than random (Withers-Ward et al., 1914, p. 1474).

Figure 5 is a graphic presentation summarizing all of the integration events detected within four preselected and two unselected regions. Integrations were detected throughout the sequence in each region with localized hyperreactive sites that were used at a frequency up to 280-fold greater than random. No obvious distinction could be made between sites selected on the basis of previous integration and those chosen at random (Withers-Ward et al., 1914, p. 1477).

[emphasis added]

An even more striking example comes from Wang et al. (2007), where an unprecedented “40,569 unique integration site sequences” were analyzed. Of all those integration events, only 41 “hosted two independent integration events at exactly the same base pair in the human genome (Wang et al., 2007, p. 1188).” Again, shared insertions resulting from parallel integration is rare—not as rare as so many misinformed individuals on the internet that use the ERV argument claim—but rare enough to make it unambiguously clear that retroviral insertion is not locus specific.

Perhaps Jonathan M. misunderstands what I mean by the term “locus specific.” What I mean is the complete dependence on a specific set of recognition sites, such as is observed in restriction endonucleases. The reason I say that locus specificity (or there about) is required by the hypothesis that the shared ERVs among chimpanzees and humans are due to parallel integration is because that is the only way to explain the fact that nearly all of tens of thousands of insertions among chimpanzee and human genomes are shared. And that provides a good transition to the next part of his response:

Out of tens of thousands of ERV elements in the human genome, roughly how many are known to occupy the same sites in humans and chimpanzees? According to this Talk-Origins article, at least seven. Let's call it less than a dozen. Given the sheer number of these retroviruses in our genome (literally tens of thousands), and accounting for the evidence of integration preferences and site biases which I have documented above, what are the odds of finding a handful of ERV elements which have independently inserted themselves into the same locus?

The TalkOrigins article in question states the following:

There are at least seven different known instances of common retrogene insertions between chimps and humans, and this number is sure to grow as both these organism's genomes are sequenced (Bonner et al. 1982; Dangel et al. 1995; Svensson et al. 1995; Kjellman et al. 1999; Lebedev et al. 2000; Sverdlov 2000). Figure 4.4.1 shows a phylogenetic tree of several primates, including humans, from a recent study which identified numerous shared endogenous retroviruses in the genomes of these primates (Lebedev et al. 2000).

That article was written some time ago, and while it may have been true at the time, since then, the entire human and chimpanzee genomes have been sequenced and compared. As I state in my ERV essay, the total indel variation between the chimpanzee and human genomes was found to be ~3%, comprising a maximum of ~45 Mb (~1.5%) in each genome (Chimpanzee Sequencing and Analysis Consortium, 2005), whereas, the total length of all ~200 thousand ERVs is at least ~127 Mb (megabases; million base pairs) (International Human Genome Sequencing Consortium, 2001). Thus, right from the start, we know that the majority of ERV are in orthologous loci. Here is the summary I gave (for more information, refer to the main page):

In summary, indel variation shows that most transposable elements, such as ERVs, cannot be lineage-specific; they must be in identical loci. When the indels are examined, this is corroborated, and less than 0.1% of ERVs are found to be lineage-specific (Polavarapu, Bowen, & McDonald, 2006). Finally, definitive confirmation is obtained by genome-wide comparison, where virtually all ERVs are directly observed to be in identical loci (Chimpanzee Sequencing and Analysis Consortium, 2005).

I want to be very clear; since all indications are that retroviruses began endogenizing tens of millions of years ago (Hughes and Coffin, 2005), common ancestry indicates that the majority of ERVs in the human and chimpanzee genomes must be shared. Only sharing a few, such as 20, or even 200, is dramatically inconsistent with common ancestry. If humans and chimpanzees only shared a few ERVs, then the target site preference demonstrated by myself, and by Jonathan M. would be more than adequate to account for them. It is the fact that we share the vast majority of tens of thousands that makes locus specificity necessary for the model of uncommon ancestry. And such locus specificity is—as previously shown—observed not to be present.

ERV Nested Hierarchy

In the last part of his article Jonathan M. addressed the hierarchical grouping of ERVs among haplorrhines, where he wrote the following:

A Nested Hierarchy?
What about this "nested hierarchy" of which we are told?

We are (incorrectly) told that "There is only one, solitary known deviation of the distributional nested hierarchy; a relatively recently endogenized/fixed ERV called HERV-K-GC1."

This claim, however, is false.

He then quoted Yohn et al. (2005), before writing the following:

As irritating to the evolutionary model as it might be, there are, in fact, a significant number of deviations from the orthodox phylogeny.

Either actually reading Yohn et al. (2005), or at least reading the portion of my “Common creationist responses” section, entitled “Family 1, 2, and 3 CERVs,” it is clear that the authors are not presenting any orthologous ERVs that deviate from the nested hierarchy of distribution. In fact they are not presenting any such orthologous ERVs at all. The following is a portion of what I wrote:

As stated in a publication on PtERV1, by Yohn (2005, p. 578-579) and his colleges, "275 (95.8%) of the insertion sites mapped unambiguously to non-orthologous locations."

They also identified 24 insertions that "could not be definitively resolved as orthologous or non-orthologous" due to the limitations of the "BAC-based end-sequencing mapping approach" used. Part of the results was derived by comparing "three intervals putatively shared between macaque and chimpanzee" to "the available whole-genome shotgun sequences for [the] two genomes."

Upon so doing, there were two instances where they "were able to refine the map location to single basepair resolution... Although the status of the remaining overlapping sites is unknown, [the data resolved] four additional sites as independent insertion events and suggest that the remainder may similarly be non-orthologous."

How much clearer can the authors be? They found most of the insertions they examined to be clearly non-orthologous. When they took a closer look, they found that some of the “ambiguous” ones were also non-orthologous, and simply close together. They then wrote how that indicated the other “ambiguous” ones might also be non-orthologous.

Even in the quotes Jonathan M. provides, the authors state the following:

If these sites were truly orthologous and, thus, ancestral in the human/ape ancestor, it would require that at least six of these sites were deleted in the human lineage (Yohn et al., 2005, p.579).

[...]

[The] PTERV1 phylogenetic tree is inconsistent with the generally accepted species tree for primates, suggesting a horizontal transmission as opposed to a vertical transmission from a common ape ancestor (Yohn et al., 2005, p.579).

The authors are clearly explaining how the PtERVs in question would be inconsistent with the the nested hierarchy of ERV distribution if they were in identical loci. But, as I explained both in this response, and in my ERV essay, they definitively showed that most of the PtERVs in question are in different loci, and that the remaining ones are probably in different loci (for the record, I am giving him the benefit of the doubt, in assuming he isn't instead claiming that common ancestry indicates that ERVs in in different loci should be hierarchically grouped). I must note how the quotes Jonathan M. provided seems to stop right before this is explained in full, and then pick up later.

I must also point out how Jonathan M. did exactly what I lamented in my ERV essay; that creationists often point to deviation from patterns such as the hierarchical distribution of ERVs, without actually providing an explanatory model for why those patterns exist.

I will conclude with these two portions of my ERV essay:

Although patterns are formed and act as powerful evidence of past occurrences and of mechanisms that causes them, the complex nature of the physical world (due to so many simultaneous interactions on so many levels) often causes deviation from those patterns. Thus the patterns act as indicators of what is actually going on, and deviation from them is both expected and rendered likely not indicative of what is actually going on

[…]

It remains true that common ancestry is a powerful predictive model with regards to the placement of ERVs. It predicts that most ERVs should be shared in certain hierarchical sets and not in others, and so far, almost every one has been where predicted. A comprehensive working creationist model that is consistent with uncommon ancestry and incorporates the whole of ERV data would be an important next step, but has thus far gone unformulated.

And with a response to the nested hierarchy of distribution consisting only of pointing out pattern deviation (albeit incorrectly), it is clear that it will remain unformulated.

Post #2: More Points on ERVs

In his second post, Jonathan M. addressed the second and third layers of evidence; the LTR-LTR discontinuity pattern, and the hierarchical distribution of shared mutations among ERVs in identical loci, respectively.

Layer 3: Shared ERV Mutations

He began with the following summary of a quote from Cuevas et al. (2002), before providing said quote:

Shared Mutations?
Regarding shared "mistakes" between primate genomes, this argument again assumes that mutations are random and are unlikely to occur convergently. Cuevas et al. (2002), however, have documented, in retroviruses, the occurrence of molecular convergenes in 12 variable sites in independent lineages. Some of these convergent mutations even took place in intergenic regions (changes in which are normally thought to be selectively neutral) and also in synonymous sites. The authors also note that this observation is fairly widespread among HIV-1 virus clones in humans and in SHV strains isolated from macaques, monkeys and humans.

He then cited Bull et al. (1997) as a second example. Unfortunately, he was only discussing the sharing of mutations among ERVs in identical loci—not their hierarchical distribution. Nonetheless, I will discuss both, below.

As explained in the two publications he referenced, some of the ways that accumulated proviral mutations could deviate from a random distribution are directional selection, coincidence with respect to accumulated random mutations, and non-randomness in the mutations themselves.

Coincidence is fairly straight forward, and directional selection could account for some shared mutations that are particularly efficient at deactivating proviruses that confer detriment to the host organism with the provirus in its genome (Cuevas et al., 2002, p. 540). Non-randomness of mutation could also account for some such sharing, but it is important to note that not only is this not directly observed in these studies, but its inference is quite speculative, in light of the myriad of other explanations that could account for the observed distribution of fixed mutations (Bull et al., 1997, p.1505).

Given that Jonathan M. only claimed that such convergence could account for shared ERV mutations, the studies he referenced are relevant, and I agree that convergent events may muddy the waters a bit, when attempting to determine phylogeny. But the fatal flaws in attempting to address the third layer of ERV evidence for common ancestry with these deviant accumulations of mutations are as follows:

1) Since the same proviruses are in the same loci, tendencies towards forming and fixing the same mutations in identically positioned ERVs would be quite similar in all of them (Cuevas et al., 2002, p. 540; Bull et al., 1997, p.1505). This would make it difficult to form a pattern, such as a nested hierarchy.

2) As expected from point #1, the convergent events among interbreeding populations displayed virtually no hierarchical grouping (Bull et al., 1997, p.1500-1501)—that is to say; virtually no nested hierarchy formed among organisms known not to share common ancestry.

To expand on point #2; a phylogenetic tree of initial lineages was constructed in Bull et al. (1997, p. 1503), but examining the data it was constructed from shows how truly weak the pattern is. Not only did more convergent mutations deviate from the pattern than those that comprised it, but the only group within a group, ((S1,S3)S2), formed only by two convergences in S1 and S3, to the one convergence in S1 and S2. Yet looking at each of the 16 convergent events in question, showed that three of them involved S1, S2, and a C lineage, but not S3, whereas only one event involved S1, S3, and a C lineage, but not S2. For the grouping of the C lineages, only one event involved C1 and C2, and no other, whereas an additional four events involved C1 and C2, and S lineage(s) (Bull et al., 1997, p. 1501-1502).

It should be very clear from examining this data that even when the volume of convergence is significant, there is little hierarchy in its distribution. But uncommon ancestry not only necessitates that these convergent events ubiquitous among the diversity of life, but that they also form nested hierarchies as robust as those observed. That simply isn’t indicated in studies, such as these.

Layer 2: LTR-LTR discontinuity

Jonathan M. then moved on to the second layer, where he wrote the following:

LTRs And Phylogeny
The other argument offered by the article pertains to primate phylogenies in relation to long terminal repeat (LTR) sequences. Because LTRs are identical at the time of integration, it is argued, if the 5' and 3' LTR sequences are very different with respect to one another, this should correspond with an older insertion. The problem is that the pattern is nothing like as neat and tidy as many Darwinists would like us to think.

He followed the above statement with a quote from Kijima and Innan (2010) that explained how gene conversation causes minor pattern deviation among primate genomes. That is all he provided in is response to the second layer of evidence. I am genuinely perplexed as to why he did this, since I clearly state in my ERV essay, which he is supposedly responding to, that this is the case:

There is deviation from the pattern [of LTR-LTR discontinuity]—likely caused by viral transfer and interelement recombination/conversion (Hughes & Coffin, 2005) and viral transfer (Belshaw et al., 2004)—but the pattern is holds for many full-length ERVs and is explainable only by decent with modification from a specific series of common ancestral species.

As shown above, I even mention sources of deviation that Jonathan M. doesn’t. What I emphasize is the existence of the pattern itself, and how that provides evidence for common ancestry.

Summary/Conclusion

When Jonathan M. announced that he was going to scrutinize and respond to my essay, I was interested to see what he had to say, and looked forward to the challenge. What he ended up providing, however, was mostly disappointing.

With respect to layer 1, he brought up syncytin, witch wasn't directly relevant; then restated the HERV-K-GC1 argument I refuted in the essay, without even mentioning what I said, let alone responding to it. He misrepresented observed ERV modification of gene expression, and when he finally discussed research on target site preference, what he demonstrated about it was what I already demonstrated in the essay. He then used and old, outdated TalkOrigins article that grossly underrepresented the actual number of primate ERVs in identical loci. Finally, he addressed the hierarchical distribution by restating the PtERV1 argument that I also refuted in the essay; again, without mentioning or responding to what I said. With respect to layer 2, he merely pointed out deviation that I already pointed out in the essay. The only partially valid point he made was in response to layer 3, where he presented an explanation that was moderately supported by the sources he cited.

But what Jonathan M. didn't even attempt to explain is why the hierarchical patterns exist in the first place, or why they corroborate one another. Despite all this, he was still bold enough to conclude with; "Unfortunately for Darwinists, however, the evidence for common ancestry is paper thin on the ground."

I think the most appropriate conclusion to this rebuttal is to restate the conclusion I wrote for the ERV essay Jonathan M. attempted to respond to:

I can only hope that if anyone else responds to my formulation of the ERV argument for common ancestry, they do not provide non-responses, and do provide a "comprehensive working creationist model that is consistent with uncommon ancestry and incorporates the whole of ERV data."

Update (6-7-11): Mistakes, Revisions, and Copy/Pasting — Just How Canned is Jonathan M’s Response?

As stated in the above aside, Jonathan M. merely copy/pasted a Sverdlov (1998) quote (replete with misattribution) from an essay by Sean Pitman, although he has since corrected it (Google Cache, 3 June 2011). There, I had expressed my curiosity as to how many other quotes he provided without reading the research publications they came from. Well, I finally got around to doing some more digging, and the results are enlightening. Here, I will go through his response in greater detail that in the above summary, and will point out the many copy/pasted quotes, uncited sources, and later revisions by Jonathan M.

In his first post, after quoting my claim about target site preference, and indicating that what follows would be the results of “[surveying] some of the literature” on such preference, Jonathan M. wrote a bit about syncytin, and provided a supporting quote. Simply mentioning functionality would have been a non sequitur, were it not for the fact that he insinuated that the derivative proviruses independently inserted in identical loci (which I showed to be untrue). He provided a summarizing sentence, and then immediately launched into a quote from Barbulescu et al., 2001 about HERV-K-GC1, followed by an argument that I had already addressed in my essay.

So why the misrepresentation of syncytin and the non-response of repeating the already addressed HERV-K-GC1 argument? The answer is that they are both reused. Not only has Jonathan M. already used those same two quotes, in the same order, with the same added emphases in a previous article he wrote for uncommondecent.org, but they are not even his quotes.

In November of 2006, Dan Reynolds posted an article entitled “Does molecular evidence prove common ancestry ‘fact’?” to the TASC website. There, Reynolds presents the same two quotes, in the same order, and one even has the same added emphasis. The abrupt transition from syncytin to HERV-K-GC1 is also the same.

Below is a composite image of three side-by-side screen shots; Jonathan M’s response to my essay (left), Reynolds’ 2006 article (middle), and Jonathan M’s previous UD post (right). It is clear that he heavily used Reynolds’ 2006 article as a source and template—practically a paraphrase with some addition—yet he did not cite Reynolds’ article as a source in either of his posts.

After providing the Sverdlov (1998) quote, discussed in the above aside, Jonathan M. provided two quotes from Yohn et al. (2005), which he then misrepresented as indicating that “the same ERV exhibits markedly different patterns of transcription,” when the authors were actually talking about the expression difference between genes with intronic insertions and those without such insertions. He has since removed one of the quotes, and removed his explicit misrepresentation, rendering the remaining quote nothing more than a redundant repetition of the PtERV1 argument he later repeated in response to the nested hierarchy of ERV distribution (Google Cache, 3 June 2011).

Next, Jonathan M. provided a quote from Michael Lynch, which has been on the Creation Wiki article on Pseudogenes since July of 2010. Immediately afterwards, he provided a quote from Daniels and Deininger (1985), which he got from a UD article he wrote—the same one he got the syncytin/HERV-K-GC1 quotes from.

Jonathan M. then used the TalkOrigins article by Douglas Theobald that Ashby Camp’s 2001 TrueOrigins article was a response to—the same article that Reynolds claimed his 2006 article was inspired by.

Finally, Jonathan provided two quotes from Yohn et al. (2005), followed by a sentence about deviation from the nested hierarchy of distribution (implying that PtERV1 insertions constitute said divergence, which I refuted in my essay). Interestingly enough, he copy/passed both quotes from the 2001 Pitman essay—the same essay he got the misattributed Sverdlov (1998) quote from. The Pitman essay even follows those two quotes with a sentence about deviation from the nested hierarchy of distribution. Below is a composite image of two side-by-side screen shots; Jonathan M’s response to my essay (left), and Sean Pitman's 2001 essay (right):

In his second post, Jonathan M. first responds to layer 3 of ERV evidence for common ancestry, by quoting Cuevas et al. (2002) and mentioning Bull et al. (1997). Guess what also provides the same quote from Cuevas et al. (2002) and mentions Bull et al. (1997); the 2001 Pitman essay.

Finally, he responds to layer 2, by mentioning pattern deviation caused by interelement conversion, and by quoting from Kijima and Innan (2010). Oddly enough he does so after switching to general terms by saying that “the pattern is nothing like as neat and tidy as many Darwinists would like us to think,” despite that I make it clear in my essay (the one he is purportedly responding to, here) that such deviation does exist, and that it is caused by the same process he describes (interelement conversion), as well as interelement recombination and viral transfer.

After discovering all this, it is no wonder that Jonathan M’s response to my ERV essay was so scarily relevant; he was mostly repeating arguments from others, without citing them as sources—misrepresentations, misattributions, misunderstandings, and all.