Summary of Studies on Primate Phylogeny and "Flying Primate" Hypothesis:

Molecular Phylogeny of Bats in Disarray:

Numerous papers have appeared recently  that focus upon, or touch uon, the bat problem using molecular techniques. One of the recurring features of these papers is that microbats are not a monophyletic group. Instead, these papers propose that some microbats share a common ancestor with megabats not shared with the remaining group of microbats (see below for a diagram of this arrangement, called microbat polyphyly... or paraphyly).

The problem with these studies is not the new, heterodox proposal to split the microbats. Since I have been lambasted for trying to split bats into two groups, I am quite gleeful at the prospect of the opposition trying to defend a three-way split!

The problem is that different families of microbats are linked to the megabats in different studies.

In seven papers, (15, 20, 21, 24, 25, 29 in list below) four different microbat families or super families are each linked to megabats, in separate phylogenies. (Phyllostomidae, Emballonuridae, Rhinolophoidea, Vespertilionidae).

I think that it should be obvious that these four different examples of microbat polyphyly are all incompatible with each other. They cannot therefore provide collective evidence against other bat phylogenies, such as the flying primate scenario I proposed. I provide more details of this point below, where one can see that only a very superficial analysis would conclude that we can presently exclude any of the scenarios proposed for the evolution of bats.

The four or five different scenarios involving microbat polyphyly cannot all be right. Perhaps they are all wrong!......for reasons that I explore in detail below;.

In the present state of confusion, I would like to propose some criteria that I think should be satisfied by molecular studies before one accepts at face value the incongruent phylogenies they seem to imply:-

    1. Walk before Run before Fly:
    The molecular data and analysis methods used should have demonstrated their validity in resolving an uncontroversial, deep mammalian divergence before tackling a more controversial deep divergence such as the bats using the same data set. I illustrate this below, with the tarsier, which is indubitably a primate, but which ends up all over the mammalian tree in different molecular phylogenies that make strong claims about bats. For similar reasons, I would be deeply suspicious of a molecular study that splits taxa that are universally aligned by all other studies (e.g. emballonurids and rhinopomatids).

2. Consider Base Compositional Biases on a Site-Specific Basis:
    As shown in the example below, it is not possible to rule out a distorting contribution to the constructed phylogeny from the overall GC  content of the sequence alone. It is essential to estimate the contribution of GC/AT ratio to the informative subsitutions. When this is done for bats, the majority of subsitutions that are responsible for the bizarre new proposals (e.g. splitting the microbats) are found to be AT, not GC.

3. Take due consideration of the abundant morphological and ecological data available about the bats under study.

"...molecular phylogenetic analyses do not free systematists from a thorough inclusion of morphological and ecological data" Keifer et al 2002 molecular Phylogenetics and Evolution 25: 557-566

A little megabat, Nyctimene robinsoni.                    A large microbat, Macroderma gigas


Recent studies have appeared in support of two of my controversial predictions about primates and megabats:-

1. The heterodox, basal position of Tarsius within the primates (Jaworski 1995, McNIff and Allard 1998).
2. The phylogenetic separation of megabats and microbats, with megabats close to primates (Schreiber et al 1994, Jaworski 1995, McNiff and Allard 1998).

If one did a count of hands, the number of molecular studies that have come out against "flying primates" would be greater than those in support. The new supportive studies are in the minority, but are notable for:


The following is a summary of the relevant studies to date, with a commentary that helps address the major points of conflict.

To facilitate comparison between the numerous studies, I have divided the studies into 4 categories according to which of the following possibilities they support:-

A. Flying Primates: Megabats and colugos are sister taxa to the primates, with microbats well-separated and more basal on the tree.

B. Bat Diphyly: Megabats and microbats do not share a common flying ancestor (as in A), but megabats are not the closest sister taxon of primates.

C. Microbat Polyphyly: Microbats are not a monophyletic group but are split by other taxa, especially by megabats.

D. Monophyly of bats: Megabats and microbats share a single common (flying) ancestor.



References for each numbered row are given below the figure illustrating the growth of DNA sequence data on the bat problem.
         4 hypotheses                 A 
    Flying Primates
      Bat Diphyly 
Microbat Polyphyly
    Bat Monophyly
# Taxa (kB sequence data in brackets)

Neural Systems

1. Retinal ganglion cells

2.Retino-tectal  +

3.Retino-tectal  +


5. LGN  +

6. Visual Cortex  +

7.Hippocampus  +

8.Motor  +


9. Include flight apparatus

10. Exclude flight apparatus +



11. Serum protein epitopes +

12.  Transferrin


Protein sequence data

13 Haemoglobin

14. alpha Crystallin

DNA sequence data

15. IRBP gene

13  (16)
16. COII mtDNA

+ 21  (2.8)
17. 12SrDNA (a)

+ 11  (3.3)
18. 12SrDNA (b)

+ 5  (3.5)
19. Mhc-DRB intron

+ 26  (~50)
20.epsilonglobin intron

17  (20)
21. vWFactor gene

27  (35)
22. alpha-crystallin

14 (15)
23.12SrDNA (c) [Note conflict with a&b]
+ (+++)

88 (88)
24. mtDNA+nDNA

25.SRY gene HMG  ?

? 14(3.4)
DNA-DNA Hybridisation

26. Corrected for A+T 


27. Uncorr. for A+T


8 support  A 6 (or 9) support  B 6  support C 6 support D

Note on the 4 hypotheses: In view of the disagreements about most deep mammalian branches at the interordinal level, even at the interfamilial level (for microbats)  , I think that it is important to note partial support that may lie somewhere between the two extreme positions in the debate (i.e. Bat Monophyly vs. Flying Primates). Hence, A and B both question the widely-held view that bats are monophyletic, even though B does not link the split megabat branch strongly to primates, as in A.  Similary, version C does not provide unequivocal support for bat monophyly (even though some authors would have it thus), because the microbats are split by the megabats. While this may be neither here nor there for those with a primary interest in refuting the flying primate hypothesis, splitting the microbats is probably as heterodox as splitting the bats as a whole. Certainly the implications of variant C for bat taxonomy are serious. Studies supporting this scenario should not be buried with those in D, as if they provide unequivocal support for monophyly. If microbats are polyphyletic, then some more hard work has to be done to define their major groupings before anything sensible can be said about the relations of microbats to megabats.

   To be updated shortly (JDP 15.02.02)

Support for Different Hypotheses of Bat Evolution from DNA Sequence Data:
It can be seen that recent molecular work, with a greatly increased sample size compared with earlier studies, is not providing compelling support for the monophyly of bats. The sample sizes and proportion of the genome available for each taxon are still small in bat studies compared with other studies of mammalian which use whole mitochondrial genomes, for example.
For scale, note that the whole mitochondrial genome is about 140 Kb. Many(?easier) problems of more recent mammalian evolution than those involving bats are unresolved, despite whole mitochondrial genomes being available for each taxon (e.g. marine mammals).

Sequence Data: Numbers refer to Table above:

1. Dann JF and Buhl EH 1990 Morphology of retinal ganglion cells in the flying fox (Pteropus scapulatus): A LUcifer Yellow investigation Journal of Comparative Neurology 301:401-416

  1a  Pettigrew, J.D., Dreher, B., Hopkins, C.S, McCall, M. J. and Brown, M. (1988)  Peak density and distribution of ganglion cells in the retinae of microchiropteran bats: Implications for visual acuity.   Brain Behav. Evol32: 39-56

2. Rosa MGP and Schmid LM 1994 Retinal topography and visual field representation in the superior colliculus  of the megachiropteran, Pteropus. Visual Neuroscience 11:1037-1057

3. Pettigrew JD 1986 Flying primates? Megabats have the advanced pathway from eye to midbrain. Science 231:1304-1306

4. Thiele AM Vogelsang M and Hoffmann K-P 1991 Pattern of retinotectal projection in the megachiropteran bat Rousettus aegyptiacus. Journal of Comparative Neurology 314:671-683

5. Pettigrew, J.D., Jamieson, B.G.M., Hall, L.S., Robson, S. R.,  McAnally, K.I. and Cooper, H.M. 1989  Phylogenetic relations   between    microbats, megabats and primates (Mammalia:  Chiroptera and Primates) Phil. Trans. Roy. Soc. B. 325: 489-559

6. Rosa, M.G.P, Schmid, L.M., Krubitzer, L.A. and Pettigrew, J.D. (1993) Retinotopic organisation of the primary visual cortex of flying foxes (Pteropus poliocephalus and Pteropus scapulatus). J. Comp. Neurol. 331:1-18

7. Buhl EH and Dann JF 1991 Cytoarchitecture, neuronal composition, and entorhinal afferents of the flying fox hippocampus Hippocampus1:131-152

8. Kennedy W 1991 Origins of the corticospinal tract of the flying fox: correlation with cytoarchitecture and electrophysiology. MSc Thesis: University of Queensland, Brisbane.

9. Simmons NB 1994 The case for chiropteran monophly American Museum Novitates 3103:1-54

10. Hill JE and Smith JD 1984 Bats: A natural history . pp 243 British Museum (Natural History)
       Pettigrew, J.D., Jamieson, B.G.M., Hall, L.S., Robson, S. R.,  McAnally, K.I. and Cooper, H.M. 1989  Phylogenetic relations   between    microbats, megabats and primates (Mammalia:  Chiroptera and Primates) Phil. Trans. Roy. Soc. B. 325: 489-559

11. Schreiber A Bauer D and Bauer K 1994 Mammalian evolution from serum protein epitopes Biological Journal of  theLinnaean Society51:359-376

12. Pierson, ED. 1986 Molecular Systematics of the Microchiroptera: Higher taxon realtions and biogeography. PhD Thesis. UC Berkeley.

13. Pettigrew, J.D., Jamieson, B.G.M., Hall, L.S., Robson, S. R.,  McAnally, K.I. and Cooper, H.M. 1989  Phylogenetic relations between microbats, megabats and primates (Mammalia:  Chiroptera and Primates) Phil. Trans. Roy. Soc. B. 325: 489-559

14. Jaworski CJ 1995 J Mol Evol 41:901-908 A reassessment of mammalian alphaA-Crystallin sequences using DNA sequencing: Implications for anthropoid affinities of tarsier

15. Stanhope MJ, Czelusniak J, Si J-S, Nickerson J Goodman M 1992 A molecular perspective on mammalian from gene encoding Retinoid Interreceptor Binding Protein with convincing evidence for bat monophyly. Molecular Phylogenetics and Evolution 1:148-160

Springer MS Burk A Kavanagh JR Waddell VG Stanhope MJ 1997 Proc Natl Acad Sci 94:13754-13759
The interphotoreceptor retinoid binding protein gene in therian mammals: Implications for higher level relationships and evidence for loss of function in the marsupial mole.

16. Adkins RM and Honeycutt RL 1991 Molecular phylogeny of the superorder Archonta. Proceedings of the National Academy USA 88:10317-10321

17. Mindell DP, Dick CW and Baker RJ 1991 Phylogenetic relationships among megabats, microbats and primates. Proceedings of the National Academy of Sciences USA 88:10322-10326

18. Ammerman LK and Hillis DM 1992 A molecular test of bat relationships: Monophyly or diphyly? Systematic Biology 41:222-232

19. Kupfermann H Satta Y Takahata N Tichy Klein 1999   Evolution of Mhc-DRB introns:Implications for the origin of primates. Journal of Molecular Evolution 48:663-674

20. Bailey WJ, Slightom JL and Goodman M 1992 Rejection of the "flying primate" hypothesis by phylogenetic evidence from the epsilon globin gene. Science256:86-89

21. Porter CA Goodman M Stanhope MJ 1996 Evidence on mammalian phylogeny from sequences of Exon 28 of the von Willebrand Factor gene. Molecular Phylogenetics and Evolution 5: 89-101

22. Jaworski CJ 1995 A reassessment of mammalian alphaA-Crystallin sequences using DNA sequencing: Implications for anthropoid affinities of tarsier. J Mol Evol 41:901-908

23. McNiff BE and Allard MW 1998 A test of Archonta monophyly and the phylogenetic utility of the mitochondrial gene 12S rRNA
American Journalof Physical Anthopology 107:225-241

24. Teeling EC, Scally M, Kao DJ, Romagnoli ML, Springer MS, Stanhope MJ.2000 Molecular evidence regarding the origin of echolocation and flight in bats. Nature. 403(6766):188-92

25.Bullejos M, Sanchez A, Burgos M, Jiminez R, Diaz de la Guardia 2000 The SRY gene HMG-box in micro- and megabats. CytogenetCell Genet 88:30-34

26. J Pettigrew, J.D. and Kirsch, J.A.W. (1998)  The bat problem:   I.DNA-hybridisation melting curves based on DNA enriched for AT- or GC-content. Phil. Trans. Roy. Soc. B, 353: 369-379.

27. Kirsch JAW Flannery TF Springer MS LaPointe F-J 1995 Phylogeny of the Pteropodidae (Mammalia: Chiroptera) based on DNA hybridisation, with evidence for bat monophyly. Australian Journal of Zoology 43:395-428

28. Hutcheon JM Kirsch JAW Pettigrew JD 1998 Base-compositional bias and the bat problem. III The question of microchiropteran monophyly. Philosophical Transactions of the Royal Society B 353:607-617

29. Liu, F-G, Miyamoto, MM, Freirre NP, Ong PQ, Tennant MR, Young TS, Gugel KF 2001 Molecular and morphological supertrees for Eutherian (Placental) mammals. Science 291: 1786-1789

30. Teeling EC, Madsen O, Van Den Bussche, de Jong W., Stanhope MJ, Springer MS 2002 Microbat paraphyly and convergent evolution of a key innovation in Old World rhinoophoid microbats. PNAS 99: 1431-1436


1. Two studies that support both the splitting of meagabats and microbats as well as the basal position of Tarsius:

Jaworski CJ 1995 J Mol Evol 41:901-908
A reassessment of mammalian alphaA-Crystallin sequences using DNA sequencing: Implications for anthropoid affinities of tarsier.

The addition of sequence from a lemur (the sifaka) revelas that it has a substitution thought previously to be a derived feature defining anthropoids. The new tree removes the tarsier from its position with anthropoids. Pteropus (a megabat) and Tonatia (a phylostomid microbat) are NOT joined by this data set.

McNiff BE and Allard MW 1998 American Journalof Physical Anthopology 107:225-241
A test of Archonta monophyly and the phylogenetic utility of the mitochondrial gene 12S rRNA

This large data set is almost equal to the sum of all previous molecular data sets on the bat problem (but note that there is some way to go, since many problems in mammalian phylogenetics outside the bats have been tackled by sequencing the whole mitochondrial genome (around 140kB) in each species under study. ...It is also perhaps significant that the thorny problem of bat phylogeny seems to be recent conference on mammalian phylogeny in Japan had dozens of phylogenetic trees presented, but none which included both microbats and megabats on the same tree so that one could see if they were joined or not!)

Despite their huge data set, the authors are quite modest about their results and about the fact that more data from the same gene now contradicts earlier conclusions. (e.g. Three earlier studies of 12S rDNA sequences concluded that bats are monophyletic. Does one give those studies 3 votes in favour of bat monophyly against the McNiff and Allard's cogent 1 against? Or does one give more weight to this much larger data set and modest claims?).

I have considerable confidence in the results, not so much because my position is supported by the bats being split, but because Tarsius occupies a basal position in the primate clade, with the colugo, Cynocephalus, exactly as the brain data dictate. The congruence between 12S rDNA sequence data and brain data can be regarded as a validation of the ability of these sequence data to "see" deep divergences in the same way that neural data are known to be able to do. If this study got Tarsius right, as I believe it has (Jaworski's data concur) then perhaps its failure to find any support for bat monophyly is right too! On the subject of megabat-primate affinity, I can wait patiently for more data and studies of this kind. After all, the amount of data is still small compared to that devoted to other phylogenetic problems and the last word has yet to be written. These authors conclude in the following way :_

"Our study demonstrates the need for caution and skepticism of studies which claim great strengths for a particular genetic marker, particularly when the evidence is based on limited sample sizes".


2. Examination of the Effects of A+T Bias on Phylogenetic Reconstruction:

van den Bussche RA Baker RJ Huelsenbeck JP Hillis DM 1998 Molecular Phyogenetics and Evolution 13:408-416
Base compositional bias and phylogenetic analyses: A test of the "Flying DNA" hypothesis.

This study was provoked by an article in which I pointed out the large A+T bias (4/1) in the substitutions claimed to support bat monophyly in each of four DNA sequencing studies. Since all megabats and some microbats were known to have A+T bias along with a reduced genomic size, it seemed to me that the sequencing studies might merely be confirming the bias already known to exist. There was no independent verification that the A+T bias was inherited from a common flying ancestor, rather than being derived independently in the two lines of bats. The latter possibility, quite apart from the common coccurrence of A+T biases in many unrelated taxa from slime moulds to malarial parasites, was suggested by the fact that many microbats (e.g. verspertilionids) do not have the bias.

van den Bussche et al examine this issue in a simulation analysis and argue that A+T bias effects are too small to bring the two kinds of bats together artefactually as I suggested.

Problems with van den Bussche et al:

In summary, despite the authors ' attempts to deny or disregard an A+T effect in bat phylogeny, there is a prima facie case for this effect in the very substitutions that they claim in support of a special microbat-megabat link. When site-specific effects are taken into account, the analysis by van den Bussche et al supports the role played by base compositional biases in erroneous associations. Finally, the details of the evolution of such biases are still being worked out and may be even further from the model they chose for analysis. To take just one example, from work on large A+T bias of the Plasmodium genome,  translational selection may result in biassed codons (e.g. an excess of C in the third base position) in highly expressed genes that compound the A+T bias at less expressed genes (Musto et al 1999).

MORE MISLEADING DNA DATA: Here is a quote from a study showing how misleading DNA sequence data can be, if it is taken at face value without careful consideration of the various biases. Using sequence data alone, Naylor and Brown (1998) were unable to confirm the undisputable phylogenetic position of Amphioxuswithin the cephalochordates!

Naylor GJP, Brown WM SYSTEMATIC BIOLOGY  47: (1) 61-76 MAR 1998
Amphioxus mitochondrial DNA, chordate phylogeny, and the limits of inference based on comparisons of sequences
Analyses of both the nucleotide and amino acid sequences derived from all 13 mitochondrial protein-encoding genes (12,234 bp) of 19 metazoan species, including that of the lancelet Branchiostoma floridae ("amphioxus"), fail to yield the widely accepted phylogeny for chordates and, within chordates, for vertebrates. Given the breadth and the compelling nature of the data supporting that phylogeny, relationships supported by the mitochondrial sequence comparisons are almost certainly incorrect, despite their being supported by equally weighted parsimony, distance, and maximum-likelihood analyses. The incorrect groupings probably result in part from convergent base-compositional similarities among some of the taxa, similarities that are strong enough to overwhelm the historical signal. Comparisons among very distantly related taxa are likely to be particularly susceptible to such artifacts, because the historical signal is already greatly attenuated. Empirical results underscore the need for approaches to phylogenetic inference that go beyond simple site-by-site comparison of aligned sequences. This study and others indicate that, once a sequence sample of reasonable size has been obtained, accurate phylogenetic estimation may be better served by incorporating knowledge of molecular structures and processes into inference models and by seeking additional higher order characters embedded in those sequences, than by gathering ever larger sequence samples from the same organisms in the hope that the historical signal will eventually prevail.

In 1991, an editorial in the prestigious journal Science lambasted me, complete with a cartoon,  for suggesting that DNA of the two kinds of bats might have convergent similarities. Michael Lee summarises the recent examples where DNA similarities seem to be related to functional constraints rather than ancestry. Lee avoids the mine-field of bat phylogeny, but uses a number of other examples to make a good case for the obfuscatory effects of adaptive changes in DNA that are interpreted by the unwary as resulting from shared ancestry.

Lee, M. 1999 Trrends in Ecology and Evolution (TREE) 14:177-178 Molecular phylogenies become functional.

Mark Pagel recently reviewed this area, citing the new statistical methods that are being used to infer phylogeny from molecular data in place of the standard approaches such as parsimony (Pagel M 1999 Nature 401: 877-884 Inferring the historical patterns of biological evolution).
Pagel spends some time on the base compositional bias problem (Pagel's approach would be concordant with my view that A+T bias has evolved separately in all megabats and some microbats) and cites a number of examples of molecular convergence of sequence data:-

Yeager M & Hughes AL Immunol. Rev. 167:45-58 1999 Evolution of the mammalian MHC: natural selection, recombination, and convergent evolution.
Bull JJ et al  Genetics 147:1497-1507 1997 Exceptional convergent evolution in a virus
Roux KH et al PNAS 95:11804-11809 1998 Structural analysis of the nurse shark (new) antigen receptor(NAR): Molecular convergence of NAR and unusual mammalian immunoglobulins
Crandall KA Kelsey CR Imamichi H Lane HC and Salzman NP  Mol. Biol. Evol. 16:372-382 Parallel evolution of drug resistance in HIV: Failure of nonsynonymous/synonymous substitution rate ration to detect selection.


3. Sequence Studies Supporting Microbat Polyphyly.

It is a surprisingly common result to find that that one microbat group is united with megabats to the exclusion of other microbats. This finding is hard to take seriously!
    First, there is a big swag of compelling derived characters shared by microbats. These include the extraordinary cochlear apparatus that is proportionately larger in all microbats than in any other animal. It is so large that it is easily identified in X-rays of 50 Myr old fossil microbats, enabling a firm conclusion that echolocation was a stem, or very early, characeristic of this group (Novacek 1985). My recent work with Pete Casseday and Ellen Covey on the brains of microbats (MS in preparation) likewise makes it clear that microbats form a monophyletic group.
    Second, and perhaps more devastating for microbat polyphyly, different studies get different microbat splits!

For example:-
a. immunological data on transferrins placed emballonurid microbats as the nearest sister group to megabats;
b. DNA  sequences have alternately linked the phyllostomids or the megadermatids to the megabats to the exclusion of the other microbats;
c. DNA hybridisation links rhinolophoids and megabats to the exclusion of other microbats.

Adherents of bat monophyly tend to lump such results together with sparser results indicating bat monophyly, despite the fact that splitting the microbats is the "thin end of the wedge" for incursions by the opposing camp with doubts about monophyly.

 If a consensus on microbat phylogeny cannot be reached, then it is asking a lot of the same techniques to decide whether and where megabats  should be joined to the microbats.

?Resolution: The conflict inherent in "microbat polyphyly trtees" can be readily resolved once one realises that the microbat taxa that tend to align with flying foxes in the different studies are all taxa with marked genomic modifications such as reduced genome size and increased A+T content. Since flying foxes share these modifications to an even greater degree, there are coincidental A & T substitutions in microbat and flying fox DNA (Pettigrew 1994). Microbat taxa without marked genomic modifications, such as the vespertilionid genus, Myotis, are well-separated from flying foxes in recent molecular phylogenies.

A. Porter CA Goodman M Stanhope MJ 1996 Molecular Phylogenetics and Evolution 5: 89-101
Evidence on mammalian phylogeny from sequences of Exon 28 of the von Willebrand Factor gene.

In other cases where there is a considerable degree of genomic modification (A+T bias, genome size), translating codons into amino acids has helped to reduce errors in phylogentic reconstructions (Amphioxus, Dictyostelium, Plasmodium). In the case of IRBP the authors could check to see how robust the DNA findings are when converted into amino acid sequence.


B. Stanhope MJ Czelusniak J Si JS Nickerson J Goodman M Molecular Phylogenetics and Evolution 1: 148-160
A molecular perpective on mammalian evolution from the gene encoding interphotoreceptor retinoid binding protein with convincing evidence for bat monophyly.

A note on the over-cooked title: Normally one leaves it to the recipients of the wisdom, rather than the donors,  to decide whether the wisdom is "convincing" enough.
My own test of a convincing phylogenetic study is to see how it handles the contentious deep branches. I did not find this study compelling for a number of reasons that include the use of an artiodactyl as the out group (an odd choice, given all the evidence for Cetungulata that would make an artiodactyl part of the in group).

I think that one can make a case that a study like this should first demonstrate its bona fides by producing a sensible outcome on one of the difficult deep mammalian branches before launching blithely at the flying primate problem as if that can easily be solved with a few kB of uncritically-examined DNA data.There are a number reasons for microbats and megabats to associate in a molecular analysis that have nothing to do with their remote shared ancestry (e.g. base compositional bias, adaptive evolution). There should therefore be some test of the validity of associations produced by the analysis. Take the rabbit as an example. The sequences of some of its moleculaes (e.g. globins) associate rabbit with primates to an extent that is still not well explained, but which lead to many investigators sneakily leaving the rabbit off their data matrix because of the spurious association with higher primates.  Why is the microbat-megabat association treated so seriously when the primate-rabbit one is not? (?Perhaps because no one has published a "flying rabbits" hypothesis in Science).

4. Nature paper proposing microbat polyphyly.

Teeling EC, Scally M, Kao DJ, Romagnoli ML, Springer MS, Stanhope MJ.2000 Molecular evidence regarding the origin of echolocation and flight in bats. Nature. 403(6766):188-92

I have provided a lenghty critique of this paper because Nature is such a high impact journal. It would therefore be easy to be misled into thinking that this paper is much better than it is. In fact, all the familiar problems are present, such as fragmentary data and inadequate treatment of the base compositional problem.

This paper has some authors who were ealier dismissive of my thesis that bats should be split into two lines.  Now they advocate that bats should be split into three lineages!

I am happy to discuss any arrangement that calls bat monophyly into question, so this paper is grist for the mill. The new phylogenetic arrangement suggested is so radical that it is surprising that there has been so little comment. The authorsí extremely off-beat suggestions about the evolution of echolocation and flight would seem to have deserved a News and Views in the same issue of Nature.

The new phylogenetic suggestion, from mtDNA and nDNA sequence data, is that microbats form two groups, a Palaetropical rhinolophoid group of 4 or 5 families, and the remaining 13 or 14 familes of the microbats, and that the rhinolophoid group is aligned with megabats rather than with the other microbats.

This is not a new result, since a rhinolophoid-megabat association was already noticed in molecular studies and dismissed as an artefact of base-compositional bias. This bias may have arisen because rhinolophoids and megabats share the highest A+T content of any group of mammals, presumably as a result of very high metabolism unbuffered by long periods of torpor. Torpor is largely absent in both rhinolophoids and megabats, in contrast to most other microbats families, so the lifetime impact of high aerobic metabolism would be expected to be greater in these two groups than in other .

The new study claims to rule out A+T bias as an explanation. This claim appears only to be lip service to the thorny issue of base compositional bias, since they have not carried out any of the appropriate controls. For example, A+T bias is site specific and we are not given any information about the A+T content of the substitutions claimed in support of the new hypothesis, despite the fact that strong A+T bias has been observed at such sites in bats. Nor is there any attempt to translate codons into amino acid sequence, a procedure that breaks the A+T bias in two examples of aberrant phylogenies produced by A+T bias (slime mould and lanceletÖsee below).

I tried to carry out some of these controls for A+T bias on the GenBank sequence data for bats used in the study and was very disappointed with the quality of the data available. There is barely enough data to translate into amino acids. It is not possible to get the phylogenetic reconstructions (e.g. PAUP,  PHYLIP) to converge using these amino acid data. The available data, converted to amino acids, certainly do not support the new hypothesis put forward in the paper. Particularly disappointing was the fragmentary nature of the nuclear DNA sequence data. Many of us are skeptical about the usefulness of mtDNA data in resolving such a distant event at the origin of bats, at least with the small data sets represented here, so we are very keen to get our teeth into solid nDNA sequence data that may be able "to reach back". In this case the alignment of the nuclear DNA sequences could not be completed because of the absence of large parts of the sequence for many bats.

Summary: In summary, while the authors eliminate A+T bias as an explanation for their extraordinary phylogeny, base compositional bias could have contributed to their result. More data are needed to resolve this issue before one can convincingly reject the large body of evidence for the monophyly of microbats.

In comparison with the flying primate hypothesis, which is a coherent scenario with two separate lineages of flying mammals, this study proposes a less parsimonious scenario with three different bat lineages and much more uncertain scenarios concerning the origins of flight and echolocation.

4a: Paper with similar conclusions to 4, but based on an expanded data set.
30. Teeling EC, Madsen O, Van Den Bussche, de Jong W., Stanhope MJ, Springer MS 2002 Microbat paraphyly and convergent evolution of a key innovation in Old World rhinoophoid microbats. PNAS 99: 1431-1436

This paper is extraordinary in its willingness to overturn a large body of morphological data in favour of a 200kB molecular data matrix and then turn around to go into minute detail about a few selected morphological features that seem to support the controversial phylogeny offered. It also glosses over the difficulty that Rhinopoma now seems to have joined the Rhinolophoid clade that is supposed to be the sister to megabats!  In other words, in addition to asking us to swallow a new arrangement of microbat superfamilies with respect to megabats, they are also proposing another large change, by moving the Rhinopomatidae from its usual sister relationship to the Emballonuridae into a much more derived position within the Rhinolophoidea!  The lack of balance in the paper is perhaps best illustrated by the fact that this change, that would probably be unacceptable to anyone who has detailed knowledge of real live microbats, is not discussed, while there is detailed discussion about more minor morphological incongruities.
    As already pointed out in a number of studies, the grouping of AT-biased megabat DNA with that of the Rhinolophoids is attributable to similar biases in both groups. The reasons that one may not be able to see these shared biases in crude overall AT/GC ratios of the genes involved is dealt with in detail above. Appropriate tests of the impact of AT bias on a phylogenetic reconstruction include:-
i. Transcription of codons into amino acid sequence data (even this may not be sufficient to eliminate bias in severe cases where codon bias occurs. cf. Plasmodium)
ii. Testing the effect of AT bias at informative substitutions (e.g. by looking only G and C substitutions)...not by a crude measure of the AT content of the genes under study.

.5. Sequence Studies Supporting Bat Monophyly.

Whole Mitochondrial Genome from a Megabat Supports Monophyly of Chiroptera.

When I saw an abstract of this paper, I started to prepare a withdrawal from the debate. After reading the paper in detail, I decided that the main conclusion is still not strongly supported and believe that the best molecular data set is still the nuclear DNA study of McNiff and Allard (1998), which supports diphyly.

1. Track Record of mtDNA:
Arnasson and colleagues have studied a number of complete mitochondrial genomes of marine mammals without being able to settle all the phylogenetic questions. Since the divergence of marine mammals is considered to be relatively recent compared with the bats, one must be cautious about this first attempt with only two families of bats (Pteropodidae and Phyllostomidae) from the 18, both of which are known to have the greatest genomic modifications of all bats.

A recent study of the evolution and mutation of mt DNA in a number of generations of living nematodes has stressed the propensity of mT DNA for rapid, convergent evolution.

2. Base Compositional Bias:
As with many recent studies of bat DNA sequences data, lip service is paid to this issue in the present study, but no serious attempt is made to address it. As I have pointed below, it is not sufficient to estimate overall base composition if one wants to avoid bias. This is because base compositional biases are site specific. This was shown nicely by Paaboís group in their analysis of monotreme and marsupial mtDNA, where it can be seen that the third base has a higher A+T content than the other positions.

Because two recent papers have completely missed the point about the role of base compositional biases, I have provided an example below that shows the impact on phylogenetic inference of such a bias. I have purposely chosen an extreme case, from a hypothetical gene segment coding for a polymer of glycine in taxa with very strong mutational bias toward A and T to make the point clear. I have no doubt that there are comparable cases in the real world of huge A+T biases, such as found in megabats and Dictyostelium.

These hypothetical sequence data illustrate the point that base compositional biases are local or site specific-
It is not therefore sufficient to consider only bulk measurements of base bias.

Imagine a protein that needs a polyglycine stretch (?as part of a backbone).

Amino Acids ......G  G  G  G  G  G  G  G  G  G  G  G  G  G  G 
                      |        |  |     |           |                                       

Note the following:-

A. Overall A+T content ~=33%
B. A+T content at 3rd base = 100%
C. 9 3rd sites sharing the same substitution; 6 3rd sites with A or T.

It is obvious that a crude assessment of base composition would conclude that these taxa did not have an A+T bias....a conclusion that would be very wide of the mark, given the 100%A+T composition at the unconstrained site.

Similarly, it would be a foolhardy phylogeneticist who would venture an opinion concerning the origin of the shared A and T substitutions above. While it is possible that they are shared derived characters that link the ancestry of two monophyletic taxa to the exclusion of other taxa not sharing these substitutions, it is also very plausible to propose that Taxon 1 and Taxon 2 independently acquired their high A+T biases at the 3rd position, so well recognised in mtDNA, and that the shared A and T substitutions are coincidental.

A final point to be made is that is is impossible to decide between the two alternatives (common ancestry vs. convergent acquisition of A+T bias) on the basis of the two DNA sequences alone. Just as with convergences of form and function, convergences of DNA sequence may not be recognisable in their own right, but require phylogenetic information from another source to reveal them.

3. Ancient origin of Megabats:
An important contribution of the paper is to provide more evidence on the ancient origin of bats. If we set aside the megabat-microbat divergence, and look instead at the divergence of bats from other mammals, the present data place that bifurcation at 80 MyA. This is remarkably in accord with other information that supports the presence of microbats in the Cretaceous. Dixie Pierson had excellent imuunological data supporting a 80-100 MyA divergence of microbats but was persuaded to play this down by the prevailing view at the time that the mammalian divergence occurred predominantly at the K-T boundary, 65 MyA (Pierson 1983). My colleagues and I presented the zoogeographical evidence for Gondwanan microbats and Hoyís group showed that insect ears for detecting ultrasound probably evolved repeatedly in the Cretaceous, in keeping with fossil evidence for noctuid moths, a family defined by its bat-evading features, in the Cretaceous.
 The present study adds to the mounting evidence that the 17 diverse families of microbats diverged in the Cretaceous.

4. Impact of Genome Size Reduction on Local Base Biases in Ribosomal DNA:
Many bats, particularly megabats, having genomic modifications that are unusual in mammals. One of these, genome size reduction, may be evident in the Ryuku Pteropus data as a smaller genome size. Since genome size reductions oftern target functionally less inportant DNA, such as the A+T-rich loops of rDNA, another important analysis would be the consequences of these modifications. As in the example of polyglycine above, genome size reduction has consequences for base bias estimation, since the effect of the reduction on rDNA may be in an opposite direction from the A+T bias.............. because A+T-rich loop regions are eliminated preferentially over G+C-rich stem regions. In other words, it would be a mistake to lump ribosomal coding regions with proein-coding regions if one were analysing bas bias in bats.
    Just another detail to emphasise that measures of base bias have to be site specific.

5. Where is the Sequence Lodged?
I did not seem to be able to find the site where the complete sequence was lodged. There were significant differences between the microbat and megabat mtDNAs that would be interesting to explore in more detail. And of course one would like to tackle the base compositional bias question in a serious way.

B.  Springer MS Burk A Kavanagh JR Waddell VG Stanhope MJ 1997 Proc Natl Acad Sci 94:13754-13759
The interphotoreceptor retinoid binding protein gene in therian mammals: Implications for higher level relationships and evidence for loss of function in the marsupial mole.

The authors note that there is only very weak support for bat monophyly. This is to be contrasted with the100% bootstrap support for the paenungulate-elephant shrew clade.

Kupfermann H Satta Y Takahata N Tichy Klein 1999  Journal of Molecular Evolution 48:663-674
C. Evolution of Mhc-DRB introns:Implications for the origin of primates.

I have pointed out elsewhere the problems of aligning sequences by eye. In one celebrated case I showed that a small change in the alignment led to a complete change in the conclusions drawn from the study. In this study there were some puzzling, strange associations that do not sit well with the claims of the study. For example, in the prosimians, Otolemur sometimes pairs with Galago and sometimes pairs with other prosimians. Likewise,  two phyllostomid bats show variation in position with each other that indicate difficulties in recognition at the generic level, let alone the deep interordinal levels claimed. It is disturbing that this study sometimes has difficulties at familial and even generic levels. I would be very cautious about its claims to sort out the deep mammalian interordinal branchpoints. Bring on the tarsier!


1. It is obvious that the molecular data are not in complete agreement, with studies supporting a whole range of possibilities from fully-fledged "flying primates", .......through various paraphyletic arrangements............ to bat monophyly..
It is true that there are no DNA sequence data in support of "Flying Primates" per se, but there are good DNA sequence support for splitting the bats, in contradistinction to the statements made recently that there is overwhelming molecular support for the monophyly of bats.

Both the very large 12SRNA data set, as well a the alpha-Crystallin DNA data set (that has the advantage that it can be joined with the amino acid sequence data on the same gene) to support bat monophyly.

2. When there is enough DNA data to allow translation into amino sequences, note that there are no cases of support for bat monophyly. This raises doubts about an exclusive reliance on DNA data, given the extreme modifications of bat DNA in some cases (especially in the megabats and the phyllostomoid and rhinolophoid microbats).

On this point, note that the base compositional biases that confound the interpretation of slime mould and malarial phylogeny can be counteracted to some extent by the use of amino acid sequence data instead of base sequence data in the phylogenetic reconstruction. This should be attempted more commonly with the bat data....although the fragmentary nature of the DNA data at the moment make it only rarely possible to translate codons into sufficient amino acids. Note further that biases in the amino acids themselves may also confound the results, as shown most clearly in the case of Amphioxus by Naylor and Brown (1998).

For these reasons, the sequence data alone have to be treated with caution and examined always in the light of the other information about the taxa that has been provided by scientitsts who know them well. To take an example:-

It is obvious that the DNA sequence data are wrong in showing "with 10,000 bootstrap reliability" that Amphioxus is not  a cephalochordate , and that it must lie phylogeneticaly outside the echinoderms. The nonsense is fairly apparent in this case because of all the accumulated wisdom about lancelets, as organisms in their own right, as well as their relationship to chordates.

It becomes more difficult to recognise phylogenetic nonsense when the DNA sequence data concern little-known organisms that have never been studied in the laboratory and whose life in the wild is poorly known. Take the tarsier, perhaps the most controversial least as regards its phylogeny:-

Is Tarsius an anthropoid, as some morphologists claim, with fragmentary support from sequence data?
Or is Tarsius a basal primate with closer ties to the colugo than to the anthropoids, as the brain data and increasing sequence data indicate?

The Tarsius example is very relevant to the "flying primate" debate, since it is essential to know the features of basal primates (not those of anthropoids that may have recent features not shared by primates as a whole), if an early branch to the flying foxes is to be sought.  It is not a "flying monkeys" hypothesis, so derived characters that are shared by higher, but not lower, primates are useless for defining relationships early in the lineage (e.g. three cone photopigments),.
The lack of agreement about the position of Tarsius, as well as the difficulties of getting a clear picture of its molecular phylogeny until recently, illustrate the greater difficulties that may face the flying primate hypothesis with its even deeper branch.

The prediction made from the brain data that Tarsius is not an anthropoid establishes the lines of argument. If this brain-based phylogenetic perspective on Tarsius continues to be verified, as the recent molecular data of McNiff and Allard and of Jaworski resoundingly do so, then the bona fides of the same approach to the flying foxes are strongly supported.  Tests of the flying primate hypothesis  will need a comparably broad-based data set, not facile acceptance of the small, biased and probably misleading DNA sequence data.

3. The subsitutions supporting bat monophyly in the DNA data are suspicious. In addition to the A+T bias that I have already pointed out, there are a number of "bat synapomorphies" where all bat taxa (e.g. all 5 bat species in the von Willebrand Factor study) share exactly the same....C..... substitution in the third base position.  Since there is general agreement that there is a very deep divergence between the two kinds of bats (even different families of microbats are thought to have a deep divergence separating them), it is astonishing that 5 taxa should all have retained the same substitution at a number of different sites, instead of the heterogeneity that would be expected after such a long period of divergence. An alternative explanation for the shared noise-free C substitution is provided by other taxa with mutational biases toward A+T in their DNA, such as Plasmodium and Dictyostelium which have the highest A+T bias known in eucaryotes. In both these taxa, as well as increased A and T substitutions in the third codon position, there is an increase in third base C substitutions as a result of translational selection in the highly-expressed genes (Musto et al 1999).
Mutational biases in DNA may therefore be misleading analysis of DNA sequences. Perhaps this is a case where the codons should be translated into amino acid sequence is done for Plasmodium and Dictyostelium...before phylogenetic analysis is carried out . Even amino acid sequence data are not foolproof (as Gavin Naylor and Wes Brown have shown for Amphioxus). But the Table above shows that amino acid sequence data are at least in agreement about splitting the bats. Perhaps better understanding and correction for the mutational bias in the DNA of megabats and some microbats will lead to even better recognition of the affinities of megabat sequences with primate sequences.

4. The molecular studies that fail to support monophyly of bats (alpha crystallin and 12SRNA), also fail to support the position of Tarsius as an agreement with the brain data. It will be valuable to learn in what way these molecular data are able to provide such a different perspective, yet one which in in agreement with the brain data.

Carnivores and Megabats:

Many molecular studies show a link between flying foxes and carnivores, often to the exclusion of microbats (alpha-crystallin protein and DNA seuence data, 12SRNA, serum protein epitopes, von Willebrand). Coupled with some carnivore  features in the brain of both Tarsius (e.g. medial terminal nucleus), as well as the advanced visual features of carnivores (after flying foxes, carnivres have the most primate-like visual system), this carnivore-primate-flying fox link deserves more investigation.

Carnivores are placed in a new super-Order called Cetungulata (or variants on that) that includes artiodactyl ungulates, whales and carnivores. The evidence for the grouping is strong and growing, but many might consider it premature to take this into consideration while tackling another controversial proposal like "flying primates". If one accepts Cetungulata as a monophyletic grouping, the nearest clade is the primates! (leaving out the bats, as all analyses of Cetungulates presently do).

While carnivores and artiodactyls  share some brain characteres with primates (whales are not well-enough known), neither group is as close to primates as are flying foxes.

The perennial association between carnivores and flying foxes in molecular data could therefore reflect the fact that flying foxes lie between Cetungulates and Primates.
Here is one example of a phylogenetic analysis showing the megabat-carnivore connection.
This analysis took the data obtained by K. Bauer's group using monoclonal antibodies that recognised epitopes on serum proteins and converted them into a matrix of 88X15 (charactersXtaxa). This matrix could then be used in a parsimony analysis by Swofford's PAUP program. Apart from linking megabats (Pteropus spp.) to the representative primate (human) in all the trees, it is also noteworthy that the next closest sister group to the primate-megabat branch is the carnivores (Felis and Canis). The microbats (Taphozous, Macroderma) are not only separated from the megabats, but also from each other.


How does one recognise a convergence?  Hunting for Homoplasy.
My favourite example of a convergence is the spherical gradient-index lens and simple camera eye independently invented by cephalopods and fish. The recent molecular evidence indicates, against intuition, that both these eyes shared a common ancestry very early in evolution when the pax6 gene first organised the apposition of a nerve fibre to a photoreceptor. We can nevertheless infer that the two optical systems were separate inventions, after the period of shared ancestry of photoreceptor aggregation and innervation, as seen in planarian eyes (I heard Walter Goehring convincingly put his case for this scenario recently in Miamai).

Nevertheless, this example still provides a basis for thinking about convergence. How does one become suspicious that a convergence is present?

1. Not Unique:
One can often find examples of similar specialisations in other phylogenies. In the case of the aquatic eye of squid and fish, the widespread distribution of eyes, independently evolved to take advantage of the spatial precision offered by light, helps to put one on guard about the possibility of the convergent evolution of eyes.

In the case of bats, the specific association of different microbat families with megabats in different studies should ring alarm bells. Pierson found that emballonurids were closer to megabats than other microbats (using transferrin immunology), Stanhope and colleagues found that megadermatids were closer to megabats than other microbat families (DNA sequence of IRBP gene), Goodman and colleagues found that phyllostomids were closer to megabats (epsilon globin intron DNS sequence), and Kirsch and colleagues found that rhinolophoids were closer to megabats than other microbats (DNA-DNA hybridisation).

A+T bias in DNA sequence data are widespread. Notable examples include Dictyostelium (a slime mould) and Plasmodium (malarial parasite) which top the list at around 85% A+T, birds, some bats (megabats are 75%, the highest of any verebrate, while some microbats reach 65% such as Rhinolophus). One should be aware, therefore. of the possibility of a convergent change in DNA of both kinds of bats, especially since there is a lnk between high metabolism and A+T increase.

2. Differences:
Fine analysis often reveals differences in the detailed construction of the two convergent arrangements. In the squid/fish eye example, the nervous system gives the game away immediately because of the reversal of retinal layering:- because the squid has the ganglion cell layer closest to the brain (in contrast to the vertebrate case where light must pass through the ganglion cell layer before reaching the photoreceptors on the brain side of the retina) there is no optic disc where the optic nerve fibres have to penetrate the retina, as in the fish.

In bats, there are numerous cases where one can be suspicious of the flight adaptation having a common origin because of non-overlapping differences in structures associated with the adaptation (e.g. metacarpals and metatarsals, spinal cord modifications). In the case of the molecules there are some features that may suggest convergent evolution, such as the reduced similarity of microbats that do not have large genomic reaarangements (such as Myotis, which shows no association with megabat sequences)

3. Selective Explanation:
There will usually be some obvious selective explanation for a convergence, although the presence of such an explanation is no guarantee that it is a convergence rather than a shared derived trait. The presence of an adaptive explanation can help in the recognition of "correlated characters" that are all part of the same convergent complex. The absence of a selective explanation might help weigh the balance away from convergence.

While the selective pressures of flight on molecules are not so obvious as those on the muculo-skeletal system, they can nevertheless be sketched out. I was resoundingly mocked for suggesting that DNA could undergo adaptive evolution in reponse to the pressures of flight, but evidence gradually accumulates to support that view. e.g. 1. genome size is reduced in volant vertebrates to decrease cell size and increase efficiency of metabolic transfers across cellular boundaries; this reduction in DNA is selective with consequences not yet well-understood; in introns the G-C content may oppose the general trend so that stem cross-links are preserved in RNA complexes; 2. mutational bias tends to favour A+T because G is most sensitive to oxidation and 8-oxo-guanine is misread as A during repair. 3. recent work suggests that large A+T mutational biases can result in compesating "translational" selection

4. Tesseract:
Incorporating a convergence into a phylogenetic reconstruction as a shared derived feature can bring the two involved taxa much closer together than predicted from all the other phylogenetic information ("tesseract" for a an extra dimension that overarches all the other dimensions).
In the case of the squid eye, to take an absurd example, treating the convergent optical systems as shared derived features would enormously reduce the phylogenetic distance between cephalopods and vertebrates.
This criterion can be used more readily on molecular data than the others. There are plenty of examples in recent molecular debates where the distance indicated by one molecular comparison is much smaller than indicated by all the other considerations. I regard these cases with suspicion. Care must be taken to eliminate adaptive evolution of DNA (a kind of convergence) such as has been described in Amphioxus, Dictyostelium, Plasmodium and the lyzozyme gene of leaf-eating monkeys (to take a few examples).

A Specific Case from the Bat Problem that Suggests Microbat-Megabat Convergence Rather than Shared Ancestry:
Many tree-construction algorithms put increased weight onto the shortest distances in the tree. If one short distance is anomalous because of convergence, that one value can distort the whole tree, overriding all the longer distances. I think that this happens often with bats because the megabat-microbat distance is consistently under-estimated. An example of this is the DNA-DNA hybridisation data that shows an anomalously-short mega-micro distance....far too small to fit with the generally long distances obtained when one micro is compared with another (interfamilial micro-micro distances tend to be like inter-ordinal distances for other mammals). Are we to take the mega-micro distance at face value, even though it is much shorter than micro-micro distances and is certainly shorter than one would expect for a deep mammalian node)? To do so allows it to distort the pattern given by a host of other distances that are consistent with each other, but inconsistent with this one distance.
Alternatively, one can test the fit of all distances except the mega-micro one that we suspect is anomalously short. In this case there is a perfect fit with no residuals, lending some support to our contention (right side of diagram).
Further support was gained by reducing the A+T content of the DNA used for the hybridisation experiment. In this case, the separation between mega and micro increased (Not shown, see Pettigrew and Kirsch 1995).
Finally, when microchiropteran DNA was used from a taxon with little genomic modification (e.g. Noctilio) the separation from megabat was long (around 60...just like the estimated micro-mega distance), a result that implies convergence unless one splits Noctilio from the other microbats (See Pettigrew and Kirsch 1995).


last modified 9 Nov 1999