On "Molecular phylogeny of Vespidae (Hymenoptera) and the evolution of sociality in wasps"

20 Pages·2003·2.6 MB·English

by  Carpenter, James M. (James Michael), 1956-

AMERICAN MUSEUM
Novitates
PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY
CENTRAL PARK WEST AT 79TH STREET, NEW YORK, NY 10024
Number 3389, 17 pp., 9 figures February 19, 2003
On *“‘Molecular Phylogeny of Vespidae
(Hymenoptera) and the Evolution of
Sociality in Wasps’”’
JAMES M. CARPENTER!
ABSTRACT
The alignment of molecular sequence data published by Schmitz and Moritz (1998. Molec-
ular phylogeny of Vespidae (Hymenoptera) and the evolution of sociality in wasps, Molecular
Phylogenetics and Evolution 9: 183-191) supported closer phylogenetic relationship of Eu-
meninae to Polistinae + Vespinae than Stenogastrinae, from which they concluded that social
behavior has independently evolved twice in the wasp family Vespidae. However, their anal-
yses also showed the Vespidae as paraphyletic in terms of the bee family Apidae. Simultaneous
analysis of these molecular data with published morphological and behavioral characters is
presented. The resulting cladograms support monophyly of Vespidae, as well as monophyly
of social wasps, with the primitively social Stenogastrinae being more closely related to the
highly social Polistinae + Vespinae than the solitary Eumeninae. A realignment of the se-
quence data is also presented, which is more parsimonious than that published by Schmitz
and Moritz. Analysis of the realigned sequences also supports monophyly of Vespidae, as well
as monophyly of social wasps, with the Stenogastrinae being more closely related to Polistinae
+ Vespinae than are Eumeninae.
INTRODUCTION strong evidence that sociality has indepen-
dently evolved twice in the Vespidae.”’
Schmitz and Moritz (1998: 183) published
The data consisted of sequences from the
alignments and analyses of two molecular 16S Mer DNA-and 28S cDNAS lock forthe
datasets which, they claimed, “provide following sample of wasps of the family
' Curator of Hymenoptera, Division of Invertebrate Zoology, American Museum of Natural History. e-mail: carpente@
amnh.org
Copyright © American Museum of Natural History 2003 ISSN 0003-0082
2 AMERICAN MUSEUM NOVITATES NO. 3389
Vespidae: three Vespinae (Vespa crabro, in results is obviously significant for the
Provespa nocturna, and Vespula germani- evolutionary study of social behavior in
ca), three Polistinae (Belonogaster petiolata wasps: Whether social behavior evolved
and two species of Polistes), three Steno- once or twice independently is crucial for
gastrinae (Liostenogaster vechti, Eusteno- the interpretation of studies of behavioral
gaster fraterna, and Parischnogaster mel- features in primitively social versus highly
lyi), four Eumeninae (two species of Ancis- social wasps.
trocerus and a different species of the genus Schmitz and Moritz (1998: 183) cited
Eumenes for each sequence dataset). Ves- Carpenter (1981) to the effect that the mor-
pinae and Polistinae are all highly social phological characters grouping Stenogastri-
species, while species of Stenogastrinae are nae with Polistinae + Vespinae “‘may be
primitively social and those of Eumeninae prone to homoplasy’’, and indeed I have
are mostly solitary (for a review of behavior, pointed out that the morphological evidence
see Ross and Matthews, 1991). The datasets is not abundant and that the behavioral ev-
also included two bees of the family Apidae idence most convincingly supports the
(both of them species of Apis) and a differ- grouping (see Carpenter, 1988). The behav-
ent parasitoid outgroup for each sequence ioral characters relate mostly to social behav-
dataset (one species of Pteromalidae and ior, but of this Schmitz and Moritz (1998: 184)
one of Braconidae). stated:
For the 16S dataset, there were 169 infor-
mative characters out of 314 aligned base the use of social behavior in cladistic studies may be
difficult, since it is well known that sociality evolved
pairs (Schmitz and Moritz [1998: 187] ex-
independently at least eight times among bees ... .
cluded 62 base pairs from their analyses be-
There seems to be a reasonable risk that one is
cause of “‘poor alignments’’); for the 28S da- trapped by homoplasy if social behavior is used as
taset, 125 out of 331 aligned base pairs were aphylogenetic [sic] character in cladistic analysis of
informative. distantly related groups. Furthermore, behavioral
traits may show great plasticity, rendering them less
Schmitz and Moritz’s (1998) analyses in-
informative for phylogenetic studies.
cluded the usual (misguided) distance and
The claim of greater plasticity rendering
likelihood procedures commonly used in
molecular systematics, as well as invalid behavioral characters less informative is ac-
statistics. I will not repeat criticisms of such tually without foundation (Wenzel, 1992),
methods, which have been abundantly de- and behavioral characters do not show ele-
vated levels of homoplasy relative to mor-
tailed in the cladistic literature (Farris, 1981,
phological characters (De Queiroz and Wim-
1983, 1985, 1999; Farris in Werdelin, 1989;
Carpenter, 1990, 1992a, 1992b, 1996; Kluge berger, 1993). The critical test of whether the
and Wolf, 1993; Farris et al., 1996; Siddall, social behavior characters in these wasps are
1998; Siddall and Whiting, 1999). However, informative or not would be to include them
their parsimony cladograms indeed support- in a simultaneous analysis together with the
ed a closer relationship of the eumenine morphological and molecular characters
sample to the polistines + vespines than the (Nixon and Carpenter, 1996). A related fal-
stenogastrines, thus diphyly of social wasps, lacy in the quoted statement is the equation
as figures | and 2 show, and this was also of homoplasy and lack of phylogenetic in-
supported by analysis of the two alignments formativeness. But homoplasy in and of itself
combined. This result is in conflict with the does not prevent a character from being in-
cladistic analyses of vespid subfamilies by formative—a homoplastic character may still
Carpenter (1981, 1988) and Carpenter and be informative in a particular clade, a fact
Rasnitsyn (1990), which supported Steno- which is part of the foundation of numerical
gastrinae as the sister group of Polistinae + cladistics (the ‘“‘Wagner method” of Farris
Vespinae, with Eumeninae in turn the sister [1970], which permits evolutionary revers-
group to that clade, and thus monophyly of ibility) and phylogenetic analysis itself (see
social wasps. This latter arrangement is also Farris, 1983). As Wenzel (1997: 31) put it,
in line with traditional taxonomic treatment “Eliminating characters because they are ex-
of the group (Richards, 1962). The disparity pected to show high homoplasy is an unac-
2003 CARPENTER: MONOPHYLY OF SOCIAL WASPS 3
Cotesia glomerata
Liostenogaster vechti
Parischnogaster mellyi
Eustenogaster fraterna
Apis mellifera
Apis dorsata
Eumenes coarctatus
Ancistrocerus oviventris
Ancistrocerus nigricornis
Belonogaster petiolata
Polistes dominulus
Polistes saggitarius
Vespula germanica
Vespa crabro
Provespa nocturna
Fig. 1. Cladogram for the 16S mt-rDNA alignment of Schmitz and Moritz (1998). The length is
512 steps; consistency index = 0.52 and retention index = 0.54.
ceptable ad hoc protection of an hypothesis logical data. Schmitz and Moritz (1998: 189)
from a legitimate test.” termed their arrangement “‘unusual’’, and
Figures 1 and 2 also show the family Ves- stated, ““To clarify the exact position of the
pidae as paraphyletic, in terms of the bee ge- Stenogastrinae among the aculeate hymenop-
nus Apis. Vespids and apids are placed in dif- tera, a more extensive study, including a
ferent superfamilies and are not at all closely range of additional vespid and nonvespid
related (see Brothers and Carpenter, 1993; members of the Vespoidea, is required.”
Brothers, 1999). The separation of these two Nevertheless, Schmitz and Moritz (1998:
families is supported by abundant morpho- 190) concluded that their data provide
- AMERICAN MUSEUM NOVITATES NO. 3389
Nasonia vitripennis
Parischnogaster melly!
Liostenogaster vechti
Eustenogaster fraterna
Apis mellifera
Apis dorsata
Eumenes spec
Ancistrocerus oviventris
Ancistrocerus nigricornis
Belonogaster petiolata
Polistes dominulus
Polistes saggitarius
Vespula germanica
Vespa crabro
Provespa nocturna
Fig. 2. Cladogram for the 28S rDNA alignment of Schmitz and Moritz (1998). The length is 302
steps; consistency index = 0.76 and retention index = 0.80.
“strong evidence for the sister group rela- behavioral characters (see Nixon and Car-
tionship of Eumeninae to Polistinae + Ves- penter, 1996, for review). I now present such
pinae.”’ an analysis. I adduced the relevant morpho-
logical and behavioral data from Carpenter
(1981, on vespid subfamilies; 1987, on ves-
SIMULTANEOUS ANALYSIS
pine genera; 1988, on stenogastrine genera;
Whether the evidence of Schmitz and 1989b, on social behavior; 1991, on polistine
Moritiz is “‘strong’’ should really be assessed genera), Brothers and Carpenter (1993, for
through simultaneous analysis of the molec- aculeate family relationships), Ronquist et al.
ular data combined with morphological and (1999, for superfamily relationships), and
2003 CARPENTER: MONOPHYLY OF SOCIAL WASPS ey
Hunt (1999, on vespid subfamilies), for a to- group to this clade in the tree for both align-
tal of 125 additional variables. These vari- ments (the values are, respectively, 98% and
ables are listed in appendix 1 and in a matrix 58% if 16S sites are not excluded). Use of
scored for the sequenced taxa at the end of bootstrap values for assessing confidence is
that appendix. misplaced (see Farris in Werdelin, 1989; Car-
The matrix was produced by: (1) taking penter, 1992a, 1996; Kluge and Wolf, 1993),
each cited character matrix, reducing each to but in the present context it is perhaps worth
the pertinent family, subfamily, or generic pointing out that the monophyly of Vespidae
vectors using the program Winclada (Nixon, is likewise supported by 100% bootstrap val-
2002) through deletion of irrelevant termi- ues (1000 replicates) for each combination of
nals, (2) flagging the uninformative charac- alignment with morphological and behavior-
ters by use of the “‘mop uninformative al characters (figs. 3—5). Bremer support val-
chars’? command of Winclada, and then de- ues, on the other hand, show a discrepancy:
leting those characters; and (3) combining as calculated with Nona, for the combination
the pertinent scores as summary scores with of the two alignments, paraphyly of Vespidae
the sequenced exemplars by merging the ma- is supported by just 1 step, while for the
trices (see Nixon and Carpenter [1996] on combination of molecular with morphologi-
mechanics and terminology). None of the cal and behavioral characters, monophyly of
16S data were excluded. Vespidae is supported by 36 steps.
Analysis with the program Nona (Golo-
boff, 1999) results in a single cladogram for
REALIGNMENT
the 16S alignment combined with morphol-
ogy and behavior (fig. 3) and two cladograms The lack of stability of the results sup-
for the 28S alignment combined with mor- ported by the sequence alignments could be
phology and behavior (fig. 4 is the consensus due to a number of factors, but one obvious
tree); the minimum-length tree was found by possibility is the alignments themselves.
most searches. The results of the simulta- Schmitz and Moritz (1998: 186) produced
neous analysis do not indicate a reclassifi- their alignments “by using the CLUSTAL V
cation of Aculeata, nor reinterpretation of program ... which were improved by com-
vespid phylogenetic relationships. Vespidae parison of the secondary structure of the
are supported as a monophyletic family, and rRNAs.”’? Unfortunately, criticism of any
social wasps are supported as monophyletic; alignment raises a difficult issue: in general,
that is, Stenogastrinae are the sister group of there is no optimality criterion for alignment.
Polistinae + Vespinae. Simultaneous analysis Thus, the common practice of inputting se-
of both alignments combined with the mor- quences into the Clustal program, obtaining
phology and behavior leads to the same re- an output alignment, and then changing that
sult (fig. 5). This is in accord with the results alignment by the user is acceptable, despite
published by Carpenter (1981, 1988), and it the fact that the changes are made on an ar-
is thus seen that the data of Schmitz and Mo- bitrary basis. Of course, authors may state
ritz (1998) are scarcely “‘strong’’. When oth- that they are taking into account secondary
er characters are considered, the sequence structure or applying a weighting scheme for
alignments are overruled. multiple substitution, and so on, but repli-
In support of their conclusions, Schmitz cability of such procedures is, to state the
and Moritz (1998: 189) observed that rear- obvious, not straightforward. To this is added
ranging their tree for both alignments to sup- the common practice of discarding data on
port monophyly of Vespidae required 31 ad- the grounds that they are “‘difficult to align’’.
ditional steps. Rearranging the cladogram of Even if a program is employed, with defined
figure 5 to support paraphyly of Vespidae, cost parameters that the user does not over-
with Stenogastrinae excluded, requires 49 ride ad libitum, those parameters are purely
additional steps. Schmitz and Moritz also cit- arbitrary. This has been cogently discussed
ed bootstrap values as confirming their tree, by Wheeler (1995), who proposed to deal
100% for the Eumeninae + (Polistinae + with the problem by means of sensitivity
Vespinae) and placement of Apidae as sister analysis, wherein the cost parameters are var-
6 AMERICAN MUSEUM NOVITATES NO. 3389
Cotesia glomerata
Apis mellifera
Apis dorsata
Eumenes coarctatus
Ancistrocerus oviventris
Ancistrocerus nigricornis
Parischnogaster mellyi
Liostenogaster vechti
Eustenogaster fraterna
Belonogaster petiolata
Polistes dominulus
Polistes saggitarius
Vespula germanica
Vespa crabro
Provespa nocturna
Fig. 3. Cladogram for the combined 16S data and the morphological and behavioral characters (see
appendix 1). The length is 652 steps; consistency index = 0.64 and retention index = 0.69.
ied and multiple reanalyses undertaken in or- Application of an alignment optimality cri-
der to assess how sensitive the output align- terion would obviate the necessity for sen-
ment is to particular parameters. He also pro- sitivity analysis, and in the realignment that
posed to select among parameter schemes by follows I propose to use parsimony, specifi-
means of congruence, with the costs chosen cally in the sense of preferring an alignment
based on minimization of incongruence _ that implies fewer steps than another. As with
among datasets. Thus he appealed to an op- the use of parsimony in phylogenetic analy-
timality criterion external to the alignment sis (Farris, 1983), parsimonious alignment
procedure to select the alignment. would maximize similarities accounted for
2003 CARPENTER: MONOPHYLY OF SOCIAL WASPS 7
Nasonia vitripennis
Apis mellifera
Apis dorsata
Eumenes spec
Ancistrocerus oviventris
Ancistrocerus nigricornis
Parischnogaster mellyi
Liostenogaster vechti
Eustenogaster fraterna
Belonogaster petiolata
Polistes dominulus
Polistes saggitarius
Vespa crabro
Provespa nocturna
Vespula germanica
Fig. 4. Consensus tree for the combined 28S data and the morphological and behavioral characters
(see appendix 1). The length is of the two underlying cladograms is 458 steps; consistency index =
0.81 and retention index = 0.85.
by ancestry (inheritance) by minimizing sep- then a bound could be set on changes. For
arate origins of features (= nucleic acid res- example, if the most parsimonious alignment
idues). Obviously, for this criterion, gaps is the one that has the fewest changes, set as
must be factored so as to avoid trivial in- lower bound the length of the longest se-
variant alignments, as could be obtained by quence (= maximum number of steps on a
treating gaps as “‘missing”’ (i.e., no cost) and bush). Then the difference between maxi-
merely inserting them so that the alignment mum steps minus the longest sequence could
implied no steps. Treatment of gaps as a be minimized.
“fifth state’? would obviate that problem, and However, Schmitz and Moritz (1998)
8 AMERICAN MUSEUM NOVITATES NO. 3389
Cotesia glomerata
Nasonia vitripennis
Apis mellifera
Apis dorsata
Eumenes coarctatus
Eumenes spec
Ancistrocerus oviventris
Ancistrocerus nigricornis
Parischnogaster mellyi
Liostenogaster vechti
Eustenogaster fraterna
Belonogaster petiolata
Polistes dominulus
Polistes saggitarius
Vespula germanica
Vespa crabro
Provespa nocturna
Fig. 5. Cladogram for the combined sequence datasets and the morphological and behavioral char-
acters. The length is 907 steps; consistency index = 0.68 and retention index = 0.75.
treated gaps as missing, and so for this re- alignment for 16S of 314 variables implies a
alignment will I. Rather than beginning with maximum number of steps of 802 and a min-
the raw sequences, I began with their pub- imum of 267, with the difference being 535;
lished alignment, which can be “‘improved”’ their 28S alignment of 331 variables implies
if the gaps can be rearranged so that the same a maximum of 600 steps and a minimum of
number of aligned variables is maintained, 231, with a difference of 369. These numbers
but fewer steps are implied. This can be read- were improved using Winclada, using the
ily accomplished. Schmitz and Moritz’s Alignment Insert/Delete function. The pro-
2003 CARPENTER: MONOPHYLY OF SOCIAL WASPS 9
ESTER CCC C. CCCOCU cq 7
Coy
‘2. Apis mellifera 3 ac:
3.Apis dorsata aM AIR AGRIC ICCA
4 Vespa crabro A
‘5.Provespa nocturna ACCC, eCEC \ | Bee ACCCHAHICCCCACA- - CGigI-
6. Vespula germanica AACGAMIGIBJACECAC. - - - - - - ACGAJANICCCGACA- - CGiy-
7 Polistes dominulus SCAICE EeA ------ EG GGCAGH- - Cec
8 Polistes saggitarius : e (GAME ------ <C CGGACG- - Ce
‘9 Belonogaster petiolata C ¢ piers igs SCIICCHIGAGACG- GCE
“10.Eumenes spec WGCEE AACGACARRIC
11.Ancistrocerus oviventricAACG HICECGACA- - CG
12.Ancistrocerus nigricormiGACG CCCe ACA- - C
13.Parischnogaster mellyi CeCe AGIGCA- GGYJAC
14 Liostenogastveecrh ti CGCC
15. Eustenogaster fraterna CGCCA
Fig. 6. Portion of 28S alignment of Schmitz and Moritz (1998) as displayed on the screen by
Winclada.
INFORM. 369 - 620022015601218§0000100233100645243215105
MAXSTEP 600: 93004312791442§0000210454222958474527216
MINSTEP 231 : 3100211123132180000110221122313231312111:
WTS 331 / 331: yB f Po UiF aS S ta Ct] td 1 eta sD as i i Fs
ACTIV JO: [331. VYYYYYYYYY YYY YYY YYYY YYY YYY YYY YYYBYYYYYY Y
AONB AO en ne A ae eee re ne ey me ere
| .Nasonia vitripennis
[2.Apis mellifera
[3.Apis dorsata
[4. Vespa crabro
(5.Provespa nocturna
(6. Vespula germanica
[7 Polistes dominulus
[8 Polistes saqgitarius
[9.Belonogaster petiolata
[1 0. Eumenes spec
CECGACA.-C
[1 1.Ancistrocerus oviventricA ACG
CGCEACA.- C
[1 2 Ancistrocerus nigricomiGACGA
[1 3.Parischnogastmeelrl yi C6CG
ARG. CCCAAAA.
(1L4 iostenogaster vechti CCCG
[15.Eustenogaster fraterna EcccAW e coma nel
Fig. 7. Screen display of same portion of the 28S alignment as in figure 6, with the show character
Statistics toggle of Winclada set to on. See text for explanation of the numbers displayed.
cedure is illustrated in figures 6—9. Figure 6 statistics for this alignment has been toggled
shows a screen display of a portion of the on, which shows at the left top, in order: the
28S alignment published by Schmitz and total information content (= difference be-
Moritz. In Figure 7, the display of character tween maximum and minimum steps), the
10 AMERICAN MUSEUM NOVITATES NO. 3389
INFORM. 366 - 201560121200001082331006452.
MAXSTEP 596 : 3127914424000021f4542229584
MINSTEP 230: 1112313212000011f§2211223132:
VTS 397 (397 PA Hse ee a ell
ACTIV JO: [397 V¥YYYYYYYYYYYYYYEYYYYYYYYYYS
PIO SRO On ne Beat Seed etal cae aen | Bee ee
€36 E46 | €56
[1.Nasonia vitripennis - GCCCGGeA-- - - COM- --- --- Gi
[2 Apis mellifera CCHICCG BA EAA faef ae T T fee
[3.Apis dorsata a CHICCCCANICAARICCHRIC IC ABICAM
[4. Vespa crabro a BCCCGA- - - - - - CO ACCC cc!
[5.Provespa nocturna OPEC A - - - - - CL PACCGRAMIC(::
(6. YVespula germanica OP ABP ACGCGA- - - - - - CL ACCCRANICE
[/.Polistes dominulus CONCHICGGA- - - - - - Cl RCC RICHRIC
[8.Polistes Saggitarius CGNIGHICCGARI- - - - - - CC Aaa
[9.Belonogaster peticlata COMBBECCGCA- - - - - - Cl CCRC.
[10.Eumenes spec Wg CUICCHIGAGG- - - - - - CAM ACCBBBAA:
[11.Ancistrocerus oviventrigl] Te s-s( SA- - - - - - ACCCIAMICE:
CRUA CCYAMICE:
12.Ancistrocerus nigric(o@rT f neteiteTter e
[13.Parischnogaster mellyi [fC- COC! WP™ -- - BARA AACGIL
[14. EDR ar eral u- COCAAA A- ~~ TieTTe7 “= Ae
am-C-
Fig. 8. Screen display of first step in realignment. See text for explanation of gap insertion.
maximum number of steps, the minimum gaps at the end of the matrix when the Align-
number of steps, the character weights (all = ment Insert/Delete function is toggled on; at
1), the character activities (all active), and the end of the procedure the trailing gaps are
character additivites (all nonadditive). The stripped). Figure 9 shows the subsequent
same numbers are shown for each character step, where gaps were inserted in the place
above that character’s column. I merely in- of the adenine in character 51, moving it to
serted or deleted gaps, checking the step to- character 50, and deleted in front of the thy-
tals each time and saving changes when these mines in character 50, moving them to char-
numbers went down. This is shown for one acter 49. This reduces the maximum number
of the characters in figure 8: gaps have been of steps to 594.
inserted in front of the cytosines in character Improvement of the alignments following
45, moving them to character 51. This results this procedure was rapid in each case, up to
in a reduction in maximum number of steps a reduction of around 100 maximum steps,
to 596, an improvement of 4, with decreases after which it became more difficult, and so
as well in other step counts (note that the was halted. This indicates a potential limi-
values for weights and additivities now ap- tation of the method, similar to the well-
pear higher; this is because Winclada adds known problem with local optima during tree