Another diagnosis for a hobbit

Israel Hershkovitz, Liora Kornreich, and Zvi Laron think they know the problem with Liang Bua 1. Almost 40 years ago, Laron began studying patients with a congenital deficiency of IGF-I (insulin-like growth factor, I). This deficiency occurs because of a defect to the growth hormone receptor, which then does not respond to growth hormone (GH). Hence, patients have a high circulating level of GH, but a low level of IGF-I. After Laron's description, this type of dwarfism was called Laron syndrome, or "Laron-type dwarfism". Since 1970, the disorder has been identified in families throughout the world, caused by a large variety of mutatations to the GHR gene. Much of this is reviewed in OMIM.

In the last few decades, a large number of clinical cases of Laron syndrome have been compiled. Hershkovitz, Kornreich, and Laron (2007) review the characteristics of the LS sample. Patients were dwarfed -- significantly short in stature for their age -- by more than 4 standard deviations (SD) below the average for their population. Moreover, they had small endocranial volumes, as much as 5 SD below the average for their population.

Here, I have reproduced Table 1 of the paper, including the list of similarities between Laron syndrome patients and the LB 1 skeleton:

There are two notable features of this list, besides its sheer length. First, it includes characters from around the skeleton. This is the first substantial examination we have seen of the LB 1 features that compare the full body to the effects of any kind of human dwarfism. Evidence from the postcrania are especially important, because they form a constellation that may be the result of a common developmental cause. Second, the list includes a broad range of features that are not "outside the range" of modern human variability -- the kinds of rare features that a clinician would recognize as symptomatic in combination with other features, but that by themselves may be found within otherwise normal humans.

If you've been following closely, you may remember that Richards (2006) also proposed that the features of LB 1 might be explained by a mutation to the IGF-I pathway, possibly in combination with other changes affecting brain size. Richards pointed out that pituitary dwarfism, including Laron syndrome, may alter the proportions of the limbs in a way similar to LB 1, and I view that as an important conclusion of the current paper (Herskovitz et al. 2007) as well. In fact, Hershkovitz and colleagues argue that many of the purportedly "unusual" features of the skeleton are straightforward consequences of its small size. This includes not only the proportions of the limb bones, but other details such as their slight muscle markings.

Interestingly, the low humeral torsion of LB 1 also figures into the LS diagnosis, and they spend nearly a page reviewing this feature. The torsion increases with age up to around 16, and developmental abnormalities including LS may cause it to fall below the general adult range. But this has become a very equivocal feature. Larson and colleagues (2007) reported that the humeral torsion exhibited by LB 1 was within the range of contemporary Australians. There's a huge range of torsion included within normal human populations, now -- extending as low as macaque values. The more comparisons are included, the more the LB 1 specimen seems to fall in the human range. This is not too surprising; if every unusual skeleton could be diagnosed by comparison with a small number of specimens, there would be no need for pathologists!

Brain size

Richards (2006) considered Laron syndrome briefly, but concluded that Laron syndrome patients have a cranium that is "near-normal in size." In the present paper, Hershkovitz et al. claim that the brain size is reduced by "up to 5 SD" in Laron syndrome. What gives?

Here is the relevant text from Hershkovitz et al.:

There is no doubt that the most striking characteristic of LB1 is not small stature but rather the minute cranial capacity. Despite the fact that the cranial volume in patients with LS is usually not decreased to the same degree as observed in LB1, three points should be mentioned: a) skulls of LS patients manifest most of the unique LB1 cranial features, b) a small head is a major characteristic of LS patients (up to 5 SD below the norm) and in IGF-I gene deletion (Woods et al., 1996). Jacob et al. (2006) reported that the LB1 cranial volume falls 5.5 SD below the combined sex Rampasasa mean, similar to what has been reported for LS patients, and c) there is a high degree of association between microcephaly and growth failure in general (OConnell et al., 1965; Pryor and Thelander, 1968), GH deficiency (Dacuo-Voutetakis et al., 1974), and congenital IGF-I deficiency (Laron et al., 1968; Woods et al., 1996) in particular.
Additionally, many of the unique anatomical landmarks left by the brain of LB1 on the endocranial bony surface (Falk et al., 2005), are seen also in LS patients, and derived from the reorganization of the brain to fit into a small cranial space... (Hershkovitz et al. 2007:7).

Additionally, they point out that the genetic background of their sample of LS patients is different from that of recent and archaeological Southeast Asian islanders, which may also produce differences in the manifestation of growth deficiencies.

Is this fully convincing? The radiographs in the paper do not show skulls as reduced in cranial volume as LB 1. As far as I know (they do not present a range) there are none. Perhaps Richards (2006) is correct that a second explanation is necessary besides GH/IGR-I to explain the small brain, or perhaps the manifestation of such disorders in this population really is different. Plausibly, an archaeological specimen from anywhere is simply not comparable to the development of modern agricultural populations. I think the brain size remains a big hole in the hypothesis.

The hypothesis is testable!

The best thing about the LS hypothesis is that it is testable. There are other features of the skeleton that reflect LS that have not yet been reported for the LB 1 skeleton, but that ought to be observable.

Hershkovitz et al. (2007) point to the pneumatization of the mastoid region as possibly the most important test. LS patients have minimal or no pneumatization of this part of the cranial base; meaning that instead of spongy bone and open sinuses, they have dense compact bone:

Unfortunately, no radiographs of LB1's skull are as yet available and therefore appreciation of the extent of pneumatization in the LB1 skull is impossible. Non-pneumatized (acellular) mastoid process (Fig. 4), lack of (or minimal) frontal sinus (Fig. 2), and small paranasal sinuses are characteristic of LS (Kornreich et al., 2002) (Hershkovitz et al. 2007:3).

CT scans of LB 1 do exist, and they should be easy to check. Very easy. As in, somebody already knows the answer. That somebody just isn't me.

But is it a species?

What would it tell you if the hypothesis were true -- if LB 1 actually does have a mutation inducing a GH/IGR-I defect and this explains its stature, morphology, and brain size? For instance, does it represent a real ancient hominid species or just a pathological member of our own?

Hershkovitz, Kornreich, and Laron agree with Jacob et al. (2006), that many of the "unusual" characteristics of the skeleton actually are normal or reasonably common within the regional population of modern humans. For that reason, they find that the skeleton possesses no features that preclude it from membership in our species. So the short answer is, they think H. floresiensis is sunk.

But their longer answer is quite interesting as a defense of taxonomic conservatism, and is worth reading closely:

It is not the numerous conundrums that have been located by us and other researchers (Jacob et al., 2006; Martin et al., 2006a,b) in the Homo floresiensis publications which refute its status as a new species, but rather the wrong arguments brought to support it.
The combination of "modern" and "primitive" morphological characteristics is one of the major arguments raised by Brown et al. (2004) to differentiate LB1 from Homo sapiens. Nobody would argue, however, that LS patients who also manifest a similar combination (e.g., an extremely oval-shaped pelvic inlet, or a "bell-shaped" form of the thoracic cage), are direct descendents of Homo erectus (an idea advocated strongly for LB1 in the first paper) nor of the australopithecines (a notion which appears in the second publication). Based on morphological comparison between LS patients and normal short children, it is clearly evident that many of the "unique" primitive morphological traits seen in LB1 are due to her small stature (Takano et al., 1986). This also explains why LB1 shares most of her features, including the most "unique" ones (e.g., the deep fissure separating the mastoid process from the petrous crest of the tympanic bone; the absence of a true chin etc.) with local pygmoid populations (Jacob et al., 2006). Ignoring the possibility that LB1 is derived from a small stature population (Rampasasa pygmies are good candidates, as suggested by Jacob et al. in 2006) with its own distinct morphological features may lead to erroneous conclusions. For example, recently Larson et al. (2006) reported on a clavicle (short relative to humeral length) and scapula (normal) of LB1 and suggested that "A short clavicle may indicate a more protracted scapular position, raising the possibility of a previously unsuspected transitional stage in the course of hominin pectoral girdle evolution" (p A21). However, the length of the clavicle is mainly dictated by the shape and diameter of the upper thoracic cage. This is why both LS patients and KNM-WT 15000 H. erectus (both manifesting a very similar fan-shaped thorax) have a relatively short clavicle.
In contrast to Morwood's statement (2005) that LB1 manifests a combination of primitive and derived features that dictate exclusion from the species sapiens, we have herein offered evidence to suggest that LB1 is but a local individual in a highly inbred, probably pygmy-like population (of Homo sapiens) in whom a mutation of the GH receptor had occurred. (Hershkovitz et al. 2007:9).

In short, the persuasiveness of any combination of features as evidence depends on their correlation with each other. If they are all strongly correlated -- for instance, if they are effects of a common cause -- then the combination of features is best interpreted as evidence for that cause, rather than as multiple instances of evidence for some other hypothesis. In this case, Hershkovitz et al. argue that the common cause explaining the data does not require a species interpretation. Instead, they argue (following Jacob et al. 2006) that LB 1 and other specimens share many features with recent local people. So, the hypothesis that the LB hominids are Homo sapiens is well supported.

Now, what could contradict that hypothesis? In other words, what would be the right argument to support a new species?

Here, the morphology of the other specimens besides LB 1 come into play. It seems very unlikely that multiple archaeological individuals over many thousands of years would have had the same rare mutation(s) of the GH/IGR-I axis unless that mutation were very common in the local population. Richards (2006) accepted at face value the argument that these archaeological individuals were in fact of the same short stature and small size as LB 1, and suggested that the ancient Flores population of H. sapiens simply had a high frequency of this variant (in his view, possibly along with another variant affecting brain size). Hershkovitz and colleagues appear willing to accept this hypothesis, pointing out that LS patients have normal reproductive potential and are relatively more common in some populations:

As LB1 replicates most of the diagnostic features of LS patients (Table 1), as well as those of pygmoid Australomelanesians (Jacob et al., 2006), it can be assumed that the findings from the island of Flores represent a local, highly inbred, low stature Homo sapiens population in whom a mutation in the GH receptor had occurred. The long time presence of LB1-type humans on the island of Flores is not surprising considering that LS patients, and derived dwarfed populations with GHRH-R defect, reproduce normally (Laron, 2004) (Hershkovitz et al. 2007:9).

But it is not necessary to take this view of a long-term population with a variant GH/IGR-I allele, if the other specimens are not actually unusual for modern humans. That is the argument put forward by Jacob et al. (2006), and it doesn't yet seem to have been contradicted. The most persuasive commonalities among this collection of fragments are (1) that they are all small, and (2) that the second mandible LB 6/1 shares several features with the first. But Jacob et al. (2006) claim (1) that the local population was small anyway, and (2) that these features are regionally common and not persuasive as evidence for a distinct lineage.

An alternative claim might be that H. floresiensis was a genuine evolutionary species on Flores (and possibly other islands), and that local people today retain features from this ancient species due to local introgression. But of course, local ancestry of some features might occur whether the ancient Flores population was another species or not. We call the latter hypothesis "multiregional evolution." So any distinctiveness of the local people is in no respect evidence that ancient people on Flores were a different species; if anything, the long-term retention of local features into living populations is a refutation that they were a different species. There is nothing impossible about introgression -- as I've said many times -- but it actually is a bit easier if speciation has not occurred!

Picky details

As in many clinical descriptions of dwarfism, there is a lot of "SD" talk in this paper. That substitutes an absolute measure (e.g., meters) for a relative one (compared to the population variability). And in some ways, that confounds two different kinds of change. For example, after a very good discussion of the problems estimating proportions and stature of LB 1, the paper includes this:

Finally, Jacob et al. (2006) estimated that the stature of LB1 falls 3.3 SD below the local Rampasasa pygmy average stature of 1.46 m, within the range of the deviation in stature reported in some of the Israeli LS patients (Laron, 2004).

This is not really a valid comparison. If pygmy populations of humans already have a variant of the GH/IGF-I axis that results in reduced stature, then a further mutation on that axis should not exert the same proportional effect. We ought to expect a dwarf in a population of pygmies to be close to the stature of dwarfs elsewhere.

Instead, the important comparison is the stature itself, not the number of standard deviations below mean. Hershkovitz et al. (2007) report that the stature of female Laron syndrome patients in their sample ranges as low as 95 cm, which is smaller than the minimum stature estimate of 106 cm for LB 1. Hence, it is consistent with the diagnosis.

Also, the genetic heterogeneity of LS means that there can be substantial variations among people with different mutations:

So far 57 mutations have been described in LS patients residing in various parts of the world including South Asia (Rosenfeld et al., 1994; Rosenbloom and Guevara-Aguirre, 1998; Laron, 1999; Shevah et al., 2005). These numerous molecular defects on the GH receptor gene or the postreceptor cascade (Elders et al., 1973; Godowski et al., 1989; Laron et al., 1992; Rosenbloom et al., 1999; Laron, 2004; Woods and Savage, 2004) produce a large variety of short stature phenotypes and a wide spectrum of intellectual abilities and deficits (Shevah et al., 2005), which may also explain the differences between the LS patients and LB1 (Hershkovitz et al. 2007:9).

This is the kind of quote that can drive a person crazy. The disorder is genetically heterogeneous. As reflected in OMIM, it may even include individuals with normal GHR function, but with other downstream problems that decrease IGF-I. But it is unsatisfying because it means that no comparison can necessarily capture the effects of the disorder. So for something like the exceptionally small brain size of LB 1, it is quite possible to say, "Well, there are at least 57 different ways to have this disorder, and maybe the 58th will be manifested with even smaller brain size.

On the other hand, with 57 different varieties (hmm....) we can probably say that the sample space of genetic mutations is now very large, so we are seeing possibly a good representation of the possible phenotypic effects of changes to this axis. At least, that's my optimistic answer.

Summary

This is a powerful paper. The overlap between the morphology of LB 1 and Laron syndrome symptoms is very extensive.

To my mind, much of the credibility of the species hypothesis -- that H. floresiensis really existed on Flores for a long time and evolved a mean phenotype including derived features absent in other populations -- depends on finding more specimens from earlier time intervals. If the archaeology of the island could be extended into the period after 500,000 years, it would document the long-term persistence of some hominid population across the interval from 700,000 years ago to 90,000. At 90,000 years, given ambiguities in dating, it is entirely possible that remains may be attributed to modern humans. So documenting a persistence in between those dates is important.

Likewise, the anatomical evolution of those populations would be a key piece of evidence. Were they, as Jacob et al. (2006) suggest, connected by gene flow to the Asian landmass by recurrent connections? Or were they really isolated on Flores or possibly other islands? Only a trace of the evolutionary history, through morphology or DNA, can provide evidence of this isolation.

I don't view any of this as impossible, but naturally it remains to be demonstrated. Likewise there is nothing impossible about such a population having a unique GH/IGR-I variant, either by drift or as an adaptation to the island. But we are waiting for the evidence that they were there throughout that time.

References:

Hershkovitz I, Kornreich L, Laron Z. 2007. Comparative skeletal features between Homo floresiensis and patients with primary growth hormone insensitivity (Laron Syndrome). Am J Phys Anthropol (early) doi:10.1002/ajpa.20655

Richards GD. 2006. Genetic, physiologic and ecogeographic factors contributing to variation in Homo sapiens: Homo floresiensis reconsidered. J Evol Biol 19:1744-1767. doi:10.1111/j.1420-9101.2006.01179.x

Jacob T, Indriati E, Soejono RP, Hsü K, Frayer DW, Eckhardt RB, Kuperavage AJ, Thorne A, and Henneberg M. 2006. Pygmoid Australomelanesian Homo sapiens skeletal remains from Liang Bua, Flores: Population affinities and pathological abnormalities. Proc Nat Acad Sci USA. 103:13421-13426. DOI link