Preliminary cladistic analyses of relationships among loliginid squids (Cephalopoda: Myopsida) based on morphological data Frank E. Anderson DepartmentofBiology, University ofCalifornia, Santa Cruz, California95064U. S. A. Abstract: Most species in the cephalopod taxon Loliginidae have anear-shore habitatandare commercially important, yet phylogenetic relationships withinthegrouphavenotbeenexamined. Inthisstudy,relationshipsamong loliginidspeciesareanalyzedusingcladisticmethods. Forty-eightmorpholog- icalcharactersfor48species(40loliginidspeciesandeightoutgrouptaxa)werecollected throughexaminationofmuseumspecimensandprimary literature andcoded intoamatrixforcladisticanalysis. Bothunweightedandsuccessive weightingmaximum-parsimonyanalyseswereundertaken,andthe phyloge- netic signal ofthedatawasevaluated. Unweightedanalyses supportthehypothesis ofmonophyly forLoliginidae, andsuggestsome well-supportedsister speciesandcrown-claderelationships(suchasAlloieuihisWiilkerandSepiotcuthis Blainville),butthepositionsofthesegroupsrelativetooneanothercan- not be resolved due to the large number ofmost-parsimonious trees. Successive weighting analyses showed support for some additional major clades (PhotololigoNatsukari andNipponololigo Natsukari),and provided insight intothecladistic information valueofthecharacters in the analysis. Continued collection ofmorphological and internal anatomical data for these species for all stages ofthe life cycle, as well as the addition ofmoleculardata to the analysis,couldhelpresolverelationshipswithinthegroup. The cephalopod taxon Loliginidae contains over 40 phcnetic analysisofloliginids as the basis ofa generic-level species ofneritic squids found on most tropical and temper- taxonomy. Others, including Augustyn and Grant (1988) ate continental margins around the world (Roper et ah, and Brierley and Thorpe (1994), have used allozyme elec- 1984). In many regions, these squids are found in high trophoresis to address problems of loliginid relationships. abundance near shore, particularly when spawning, and Despite such efforts, these authors admit that further work some species form integral links in coastal marine ecosys- is necessary to clarify loliginid phylogeny. tems (such as Loligo opalescens Berry, 191 1, in Monterey No researchers have explicitly addressed loliginid Bay, California; see Morejohn etai, 1978). Many loliginid phylogenetic relationships using cladistic methodology. species are commercially harvested (Roper et al, 1984). Cladistic analysis allows many discrete character data to be In addition, the giant axons ofcertain loliginids (such as L. considered simultaneously using an explicit, simple opti- pealei, L. vulgaris, and L. opalescens) have served as mality criterion in which the preferred tree is the one which important model systems in neurophysiological research requires minimal assumptions of convergence and reversal (e. g. Young, 1938; Gilbert et ai, 1990; Rosenthal and across all characters in the analysis (the criterion of maxi- Gilly, 1993). mum parsimony) (Edwards and Cavalli-Sforza, 1964; Despite the ecological, economic, and scientific Camin and Sokal, 1965; Kluge and Farris, 1969; Fitch, importance of loliginid squids, their phylogeny remains 1971; Farris, 1983; Sober, 1988; see Wiley et ai, 1991, for unresolved, and their taxonomy is confused (Voss, 1977). a review of cladistic philosophy and techniques). Cladistic Numerous recent works (Natsukari, 1983, 1984; analysis yields a hypothesis of phylogenetic relationships Brakoniecki, 1986; Alexeyev, 1992; Vecchione et al., in with which to interpret biogeographical patterns and char- press) have sought to clarify loliginid taxonomy using key acter evolution (Brooks and McClennan, 1991; Maddison morphological characters. Brakoniecki (1986), who exam- and Maddison, 1992). Phylogenetic hypotheses can also be ined loliginid hectocotylus morphology, grouped the used to construct taxonomic schemes (de Queiroz and species he studied into six hectocotylus types. He based a Gauthier, 1990, 1992). new generic-level classification on his findings, and pro- For this study, many aspects of loliginid morpholo- posed anevolutionary zoogeographic scenario forthe radia- gy (particularly the hectocotylus, arm and tentacle-club tion of the group. Vecchione et al. (in press) have used a sucker rings, spermatophores, fins, and some aspects of AmericanMalacological Bulletin,Vol. 12(1/2)(1996):113-128 113 114 AMER. MALAC. BULL. 12(1/2) (1996) internal anatomy) were coded into a matrix for cladistic between loliginids and various active nektonic oegopsids analysis to examine species-level relationships within the like Todarodes. Due to this uncertainty, a diversity of family. These data were analyzed using a maximum-parsi- cephalopod taxawere included asoutgroups in this study. mony algorithm program, and the results were compared Many characters included in this analysis (such as with traditional taxonomic schemes and recent reclassifica- hectocotylus morphology, sucker ring dentition, and fin tions. The strengths and problems of using morphological shape) have been used in traditional studies ofloliginid tax- characteristics for examining loliginid evolution are onomy, but have never been objectively analyzed simulta- addressed, and topics for future research are briefly out- neously. In certain cases, some of these characters have lined. been presumed to be informative at some taxonomic levels but not at others. For example, arm-sucker ring dentition generally has been used to distinguish between very similar MATERIALS AND METHODS species (Natsukari, 1983; Brakoniecki, 1986), but it has not been used as a taxonomic character above this level. In DATA COLLECTION other cases, some characters have been examined only in those supraspecific taxa where they help unite or separate Forty-eight morphological characters for 40 species species (e. g. number oftrabeculae per marginal club suck- of loliginid squids and eight outgroup taxa were coded into er in Alloteuthis Wiilker; Hanlon et al, 1992), and might a data matrix in MacClade 3.04 (Maddison and Maddison, not have been thoroughly examined in all loliginid species. 1992) (Appendix I). Character states were determined Still other characters (e. g. spermatophore morphology) through direct study of museum specimens at the National have been examined widely in loliginids, but have not fig- Museum of Natural History, the California Academy of ured importantly so far in cephalopod systematic studies Sciences, and the Invertebrate Museum at the University of (Hess, 1987; deMaintenon, 1990). Some characters were Miami Rosenstiel School of Marine and Atmospheric found to vary among loliginid species but were consistent Science, as well as published species descriptions within species, and were included in this analysis. Certain (Appendix III). Some described loliginid species [including characters of traditional importance were avoided because Loligo arabica (Ehrenberg, 1831) and newly described they appeared to vary within certain species, and too few species such as Photololigo robsoni (Alexeyev, 1992)] were specimens were available to resolve these inconsistencies. not included in this study because descriptions of these Forexample, thickenings ofthe lateral edges ofthe vane of species were insufficient to code many characters with con- the gladius are for some authors important diagnostic char- fidence, and specimens were not available for examination acters for the genera Doryteuthis Naefand Loligo Lamarck at the institutions I visited. but have been found to be variable within species (Cohen, The outgroup method (Watrous and Wheeler, 1981; 1976; Toll, 1982). Despite this, some polymorphic charac- Maddison etal, 1984) was used to polarize character trans- ters (characters that vary within some terminal taxa) were formations in this study. The eight outgroup taxa represent included in the analysis. Characters exhibiting intraspecific abroad range ofdecapod cephalopod diversity. Fouroegop- variation can contain strong phylogenetic signal (although sid taxa {Moroteuthis Verrill, Todarodes Steenstrup, generally not as strong as fixed characters) and thus should Ctenopteryx Appellof, and Bathyteuthis Hoyle), two sepiol- not be ignored or simply coded as fixed in cladistic analy- id taxa {Euprymna Steenstrup and Rossia Owen), and one ses (Wiens, 1995). Aprioriassumptions about the informa- sepiid taxon {Sepia Linne) were included, along with tion value of characters (other than inclusion of "tradition- Pickfordiateuthis pulchella Voss, 1953. Pickfordiateuthis al" well-studied characters in the analysis) were avoided, was originally described by Voss (1953) as a monospecific but inevitably (as in all phylogenetic studies) some charac- taxon closely related to Loliginidae as another member of ters that could be phylogenetically informative have been the Myopsida. Recently, two new species of Pickfordi- excluded. Appendix II lists the characters, argumentation ateuthis have been described, and Pickfordiateuthis has and coding scheme used in this analysis. been subsumed within Loliginidae (Brakoniecki, 1996). Data for some characters are either not applicable For this analysis, P. pulchella was used as a representative for certain taxa ("n" in Appendix I) or could not be deter- ofthis groupofsquids. These outgroups were selected fora mined ("?" in Appendix I). Inapplicable characters usually variety of reasons. In some cases, earlier authors have sug- refer to some aspect of a structure which is not present in gested that certain oegopsid taxa are close relatives of all taxa in the analysis (e. g. "hectocotylus dorsal row suck- Loliginidae (e. g. Ctenopteryx; Young, 1991). In contrast, er morphology" in taxa which do not possess hectocotylus- Berthold and Engeser (1987) suggested that Loliginidae, es). In some cases, coding of "inapplicable" characters in Sepiolidae, and Sepiidae are all closely related members of this manner can cause problems in cladistic analyses the Myopsida. Morphological similarities also exist (Maddison, 1993), but the solution advocated by Maddison ANDERSON: LOLIGINID SQUID PHYLOGENETICS 115 (1993) - combining all such characters into one character phylogenetic utility of individual characters. The "reweight with many states - is often impractical and can lead to the characters" option in PAUP was used to successively weight loss of phylogenetic information. For example, in the case characters after each heuristic search (following the ofthis analysis, fusing all nine hectocotylus characters into approach of Farris, 1969). Farris (1989) proposed the one multistate character with many states was not per- rescaled consistency index (RCI) and suggested its use in formed. Fusing all hectocotylus characters into a single successive weighting analyses. In this analysis, characters character with many distinct unordered states does not were reweighted on the basis of their best rescaled consis- allow homology statements for individual aspects ofhecto- tency index value across the trees from the previous search. cotylusmorphology. It is possible, for instance, that species A heuristic search (following the same parameters as that possess a particular type of modified sucker, irrespec- described above) was performed using the reweighted char- tive of the region of the arm that bears the modification, acters. Rounds of successive weighting were repeated until constitute a monophyletic group relative to species with overall strict consensus tree topology did not change from othertypes ofsucker modification, orvice versa. Ifthe hec- one round to the next, or until character weights did not tocotylus characters were fused into one unordered multi- change after reweighting. As in preliminary analyses, strict state character, this information would be lost - only consensus cladograms from the final round of weighting species with almost exactly the same hectocotylus morphol- were combined, and a grand strict consensus cladogram ogy (i*. e. the same coded state for the single hectocotylus wascomputed. This allowed commonelements found in all character) would be grouped together. Maddison's (1993) successive weighting analyses tobe determined. method is valid, but maximally conservative, and a great A recently described technique called "safe taxo- deal ofphylogenetic information could be lostby collapsing nomic reduction" (Wilkinson, 1995) was also used in an characters in this way. effort to reduce the number of trees found. Analysis of matrices containing taxa with many missing data can result DATA ANALYSES in an inordinately large number of equally parsimonious trees, because taxa with a large percentage of missing data These data were analyzed using the maximum parsi- (termed negatively underdetermined taxa) can occupy a mony program PAUP 3.1.1 (Swofford, 1993). When termi- number of equally parsimonious positions (Wilkinson, nal taxa were coded as having multiple states for one or 1995). Consensus methodscan be used to find common ele- more characters, these characters were interpreted as "poly- ments across multiple most-parsimonious trees (MPT's), morphic." All characters were unordered binary or multi- but negatively underdetermined taxa can obfuscate patterns state characters, and were weighted equally for all initial of relationship among other taxa that are found in all trees, analyses. The use of equal weighting does not mean that yielding an extreme lack of resolution in strict consensus each character in the matrix is ofequal informative value. I cladograms. The goal of safe taxonomic reduction is to have chosen to use equal weighting simply because I have remove negatively underdetermined taxa from the analysis no compelling reason to use any particularaprioridifferen- without losing information about relationships (/. e. without tial weighting scheme (see discussion in Eernisse et ai, altering patterns of relationships among the remaining 1992). Heuristic searches were performed with 100 replica- taxa). This reduction in the number of negatively underde- tions of random stepwise addition of taxa using tree bisec- termined taxa often reduces greatly the number of MPT's tion-reconnection swapping with one tree held at each step. found by parsimony analysis. Increased resolution in con- The maximum number of trees stored for each search sensus cladograms is often found after safe taxonomic (MAXTREES) was 10,000. The COLLAPSE option was reduction. In these analyses, the only taxa that could be turned offfor some analyses in an effort to find all regions safely removed from the ingroup using Wilkinson's tech- of "islands" of most-parsimonious trees (Maddison, 1991; nique were Alloteuthis africana and A. media, which were Swofford, 1993). Following Maddison's (1991) suggestion, taxonomic equivalents ofA. subulata, and Loliolus affinis, a total of ten heuristic analyses were done to search for which was a taxonomic equivalent ofL. hardwickei. other islands and toexamine the level ofsupport forvarious The phylogenetic signal of the data was evaluated branches within cladograms. A strict consensus cladogram using the gi test of Hillis and Huelsenbeck (1992). Based was computed from the trees from each of the ten heuristic on simulationdata, Hillis and Huelsenbeck (1992) proposed searches, for a total of ten strict consensus cladograms. A the use of the gi statistic (a measure of skewness) as one strict consensus ofthese ten strict consensus trees (a"grand way toevaluate the ratioofsignal torandom noise in phylo- strict" consensus cladogram) collapsed all ambiguities and genetic data. They found that high degrees of left-skew in revealedelementscommon toall trees from all ten searches. plots of random tree distributions or total tree distributions After these preliminary analyses, two methods were obtained through exhaustive searches correlated well with used in an attemptto reduce tree numberand investigate the the success of parsimony methods in finding the true phy- . 116 AMER. MALAC. BULL. 12(1/2) (1996) logeny in simulation studies. Ten thousand random trees Table2. Maximally andminimally weightedcharactersacrossall weight- were generated based on this matrix using PAUP 3.1.1 with ed analyses (basedon bestrescaledconsistency index fit). Charactersthat multistate taxa interpreted as polymorphic. The gi from this varywithintheingrouparedenotedbyanasterisk(*). random tree distribution was compared to 95% and 99% Maximallyweightedcharacters confidence-limit values obtained from simulations with 50 CharacterNumber Description binary or multistate characters and 25 taxa performed by Hillis and Huelsenbeck (1992), which should provide an 1-1 rlaatcehriadlicaunscpussopfsroafdrualdaula approximate conservative comparison. In addition, all 5 numberofarmsuckerrows clades or sister-species groupings found across all trees (i.e. 14 retractiletentacles groupings retained in the grand strict consensus) were con- 15 trabeculaenumber* strained to examine if phylogenetic signal was clustered 20 photophoresoninksac* 25 ventralrowmodification* within these groups. Ifthe gi value (a negative value in left- 29 fusedcrestinventralrowofhectocotylus* skewed distributions) increases greatly after these con- 38 spermatophoreplacement* straints are applied, a large proportion of the phylogenetic 40 spiralfilament signal in the matrix can be clumped within the universally 43 poresoninksac* 46 muscularseptum supported clades, and is not evenly distributed across all 47 nuchalcartilage data (Hillis and Huelsenbeck, 1992). 48 digestivegland Minimally weightedcharacters(weightedtozero) RESULTS CharacterNumber Description 4 armsuckerrings All unconstrained, unweighted analyses of all char- 10 centralmanussuckerteeth* acters—and all taxa resulted in 10,000+ most-parsimonious 13 marginalmanussuckerteeth* trees the MAXTREES limitof 10,000 was reached in all 16 buccalmembranelobes 18 buccallappetsuckerteeth analyses. The strict consensus cladogram of all sets of 19 buccalmembraneformula 10,000 trees from all ten unconstrained and unweighted 27 lengthofhectocotylus* heuristic searches is shown in Fig. 1. Tree lengths and sta- tistics ofthe constituent trees are shown in Table 1 The strict consensus cladogram ofthe ten final strict consensus trees from successive weighting was marginally the same as their positions when all taxa were included in more resolved than the strict consensus of the unweighted the analysis. analyses (Fig. 2). Characters that were maximally and min- The gi skewness test of Hillis and Huelsenbeck imally weighted aftermultiple rounds ofsuccessive weight- (1992) suggests the presence of significant phylogenetic ing are shown inTable 2. signal in the unweighted data matrix. The gi value for Employing safe taxonomic reduction and removing 100,000 random trees derived from this data matrix was Alloteuthis africana, A. media, and Loliolus qffinis from the -1.275647. This value was appreciably below the critical analyses had no apparent effect on the number of trees values for 50 binary or 50 four-state characters for 25 taxa found in either unweighted or successive weighting analy- [95% confidence limit for binary characters = -0.10, for ses, or in the topology of the strict consensus cladograms four-state characters = -0.12; 99% confidence limit for from these analyses. Ten thousand trees were still found in binary characters = -0.11, for four-state characters = -0.13, all analyses following the removal of these taxa. The rela- based on simulation studies (Hillis and Huelsenbeck, tive positions ofA. subulata and L. hardwickei alone were 1992)]. When this procedure was repeated with the univer- sally supported clades retained (using the grand strict con- sensus as a constraint), the gi value for a distribution of 100,000 random trees was -0.502625. This value suggests Table 1. Tree statistics for 10,000 most-parsimonious trees found in the tenunweightedheuristicanalyses. that even when universally supported clades are con- strained, significantphylogenetic signal remains. Indicesforunweightedtrees Consistencyindex (CI)=0.582 Homoplasyindex (HI)=0.572, DISCUSSION Retentionindex(RI)=0.663 Rescaledconsistencyindex (RCI)=0.386 A small number of ingroup sister-species and sub- Treelength(TL)= 194 clade relationships were supported in the multiple . ANDERSON: LOLIGINID SQUID PHYLOGENETICS 117 Todarodes pacificus Moroteuthis robusta Pickfordiateuthis pulchella Ctenopteryx Bathyteuthis Sepia Rossia Euprymna Alloteuthis media Alloteuthis subulata Alloteuthisafricana Loligoopalescens Loligo pealei Loligo ocula Loligo plei Loligo roperi Loligo surinamensis Loligo sanpaulensis Loligogahi Loligo vulgaris Loligo reynaudii Loligoforbesi Loligo bleekeri Loligo duvauceli Loligo vossi Loligojaponica Loligo beka Loligosumatrensis Loligo uyii Loligo sibogae Loligo singhalensis Loligopickfordae Loligo reesi Loligo kobiensis Loliolus noctiluca Lolliguncula argus Lolliguncula mercatoris Lolliguncula brevis Lolliguncula panamensis Uroteuthis bartschi Loliolopsis diomedeae Loligo edulis Loligo chinensis Loliolus hardwickei Loliolus ajfinis Sepioteuthis sepioidea Sepioteuthis lessoniana Sepioteuthis australis Fig. 1.Grand strictconsensuscladogramoftenstrictconsensuscladogramsderived from tenheuristic searchesyielding 10,000treeseach. Treestatistics forthetreesuponwhichthisstrictconsensusisbasedareshowninTable 1 118 AMER. MALAC. BULL. 12(1/2) (1996) unweighted heuristic analyses (Fig. 1), but deeper branch- many authors to reduce the number of most-parsimonious ing orders and relationships among loliginid squids remain trees, to increase resolution in consensus trees, as a heuris- unresolved, despite the relatively high phylogenetic signal tic tool to investigate the cladistic informativeness ofchar- suggested by the gi test. Successive weighting analyses acters, and to study the effect of homoplastic characters in consistently suggested support for some higher-level rela- cladistic analyses (Farris, 1969; Carpenter, 1988, 1994), but tionships (Fig. 2), but even with successive weighting and some authors have criticized the use of successive weight- safe taxonomic reduction, tens of thousands of most-parsi- ing techniques (Swofford and Olsen, 1990) or have urged monious trees were found, and, despite some interesting caution in the interpretation of results from successive findings, the resulting grand strict consensus cladograms weighting analyses (Maddison and Maddison, 1992; are largely unresolved. Swofford, 1993; Suter, 1994). Some investigators (e. g. Monophyly of Loliginidae was supported by all Suter, 1994) have found that the parameters used in succes- unweighted and weighted analyses. Myopsida (traditionally sive weighting analyses can have an effect on the outcome comprised ofLoliginidae + Pickfordiateuthidae) was found of successive weighting analyses. In addition, successive to be paraphyletic in all unweighted and weighted analyses. weighting does not necessarily reduce the number ofmost- Brakoniecki's (1996) inclusion of Pickfordiateuthis within parsimonious trees - in this analysis, the tree buffer limit Loliginidae is not justified based on these analyses. Within was reached on all analyses, unweighted or weighted. the ingroup, only fourclades were consistently supported in Successive weighting can reduce the weight of highly all unweighted analyses, and two of these are sister-species homoplastic characters to zero, effectively removing them groupings. Monophyly of the genus Alloteuthis, consisting from the analysis [seven characters were weighted to zero of three species of small, slender squids found in the east- by the final round ofweighting in these analyses (Table 2)]. ern Atlantic along the coasts of Europe and Africa (A. As the number ofcharacters actually included in the analy- africana, A. media, and A. subulata) was supported in all sis drops, resolution in strict consensus cladograms likely analyses, and was the sistertaxon to the rest ofLoliginidae. will drop, particularly when the number of characters is Alloteuthis appears to have diverged from the other lolig- small relative to the numberoftaxa in the analysis. inid species early in the history ofthe group. Monophyly of Farris (1969) and Carpenter (1988, 1994) have the genus Sepioteuthis Blainville, which is comprised of S. strongly supported successive weighting as simply an australis and S. lessoniana (both Indo-West Pacific extension of the concept of cladistic reliability, or the species), and S. sepioidea (a Caribbean species) was also degree offit between acharacter and the phylogeny (Farris, supported in all unweighted analyses. Two synapomorphies 1969). Successive weighting allows aposteriori weighting appear to unite the three species ofSepioteuthis - a longitu- based on the cladistic information value ofthe characters in dinally oval fin shape uniquely derived within Loliginidae, the matrix. Characters that show little homoplasy when and the large size of the dorsal row of papillae relative to evaluated in conjunction with all other characters in the the ventral row in the modified portion ofthe hectocotylus. matrix are increased in weight (or set at a maximum base In addition to these three-taxon clades, two sister-species weight in PAUP 3.1.1) relative to characters exhibiting pairings were consistently found - Loligo chinensis and L. more homoplasy in subsequent rounds of analysis, while edulis in one pairing, Loliolus hardwickei and L. affinis in highly homoplastic characters are reduced in relative the other. All other relationships within Loliginidae are weight. Carpenter (1988, 1994) has argued that successive unresolved in the unweighted analyses. weighting "allows the characters of a given data set to There are a numberofpossible explanations for this judge themselves in terms oftheirreliability; that is, best fit lack ofresolution. First, the numberofcharacters employed to the solution supported by all the characters" (Carpenter, (48) is relatively low compared to the number of terminal 1994: 216). taxa (48) included in the analysis. Another important factor Successive weighting analyses supported all clades might be that many of the characters used in the analysis found in earlier, unweighted analyses, and suggested three are highly homoplastic, showing evidence ofmultiple con- other groupings not found in unweighted analyses. Two vergences or reversals throughout the evolution of this major clades were supported in all final weighted analyses group. Several of the characters included in this analysis - a clade consisting of all loliginid species possessing have been used extensively in decapod cephalopod taxo- paired bioluminescent organs on the ink sac [Loligo edulis, nomic studies, and appear to be very useful for distinguish- L. chinensis, L. duvauceli, L. sihogae, L. singhalensis, L. ing species, but when relationships among all loliginid pickfordae (Adam, 1954), L. reesi, Loliolus noctiluca, and species are studied and characters are atomized forcladistic Uroteuthis bartschi, all found in the Indo-WestPacific], and analysis, individual characters can exhibit high levels of a clade consisting of seven other Indo-West Pacific species homoplasy. (Loligo beka, L. japonica, L. sumatrensis, L. uyii, L. kobi- Successive weighting techniques have been used by ensis, Loliolus hardwickei, and L. affinis). In addition, ANDERSON: LOLIGINID SQUID PHYLOGENETICS Todarodes pacificus Moroteuthis robusta Pickfordiateuthis pulchella Ctenopteryx Bathyteuthis Sepia Rossia Euprymna Alloteuthis media Alloteuthis subulata Alloteuthis africana Loligo pealei Loligo ocula Loligo plei Loligo roperi Loligo surinamensis Loligo sanpaulensis Loligogahi Loligo vulgaris Loligo reynaudii Loligoforbesi Loligo vossi Lolliguncula argus Lolliguncula mercatoris Lolliguncula brevis Lolliguncula panamensis Loliolopsis diomedeae Loligo opalescens Loligo bleekeri Sepioteuthis sepioidea Sepioteuthis lessoniana Sepioteuthis australis Loligojaponica Loligo beka Loligo sumatrensis Loligo uyii Loligo kobiensis Loliolus hardwickei Loliolus affinis Loligo duvauceli Loligo sibogae Loligo singhalensis Loligopickfordae Loligo reesi Loliolus noctiluca Uroteuthis bartschi Loligo edulis Loligo chinensis Fig.2.Grandstrictconsensuscladogramoftenstrictconsensuscladogramsfromsuccessiveweightinganalysesfromeachoftenheuristicsearches. 120 AMER. MALAC. BULL. 12(1/2) (1996) Loligo opalescens and L. bleekeri were found to be sister these analyses, the lack of suckers on the tops of the slabs species in all final weighted analyses. was revealed as a synapomorphy uniting these two species The majorputative synapomorphy uniting the biolu- as sister taxa. In the ventral row of the hectocotylus, all minescent loliginids are the paired bean-shaped bacterial species in the Nipponololigo clade possess minute, appar- photophores (luminescent organs) on the ventral side ofthe ently suckerless papillae. Vecchione et al. (in press) have ink sac. This clade is similar to the proposed genus proposed the name Loliolus to include all members of Photololigo Natsukari (1984), which includes five of these Natsukari's Nipponololigo as well as L. affinis and L. hard- species {Loligo edulis, L. duvauceli, L. chinensis, L. sing- wickei. These species are divided into two subgenera - halensis, and L. sibogae) as well as L. arabica, which was Loliolus (Loliolus) (comprised of L. affinis and L. hard- not included in this analysis. Natsukari's Photololigo, how- wickei) and Loliolus (Nipponololigo) (comprised of the ever, does not include L. reesi, L. pickfordae, Loliolus noc- species in Natsukari's Nipponololigo, plus L. sumatrensis). tiluca, or Uroteuthis bartschi. Results of this weighted As with Photololigo, these analyses generally support their analysis support Vecchione eta/.'s (in press) Photololigo as conclusion, although a paraphyletic Nipponololigo (with a monophyletic group. However, these authors divide respect to L. affinis and L. hardwickei) is a possibility that Photololigo into two smaller groups - a subgenus cannot be addressed with these dataalone. Photololigo and a subgenus Uroteuthis (consisting only of Loligoopalescens andL. bleekericonstitute asister- one species- U. bartschi). It is possible that U. bartschi is a species pairing in all weighted analyses. Brakoniecki derived member of the photololiginid clade. If this is true, (1986) anticipated this result. He proposed that the epithet Vecchione et a/.'s (in press) subgenus Photololigo is para- Doryteuthis (subgenus Doryteuthis) be applied to six phyletic with respect to their subgenus Uroteuthis. Un- species of loliginid squids. Five of these species (Loligo fortunately, cladistic analysis of these data cannot address plei, L. roperi, L. sanpaulensis, L. gahi, and L. opalescens) this question. Alexeyev (1992) has reported that some spec- are found in American waters, while one species (L. bleek- imens ofLolliguncula mercatoris and a single specimen of eri) is found only in Japanese waters. Doryteuthis Loligoforbesi appeared to possess photophores on the ink (Doryteuthis) and Sepioteuthis are the only geographically sac. In this analysis, these species are considered to lack disjunct groupings described by Brakoniecki (1986). photophores, pending further investigation (as suggested by Brakoniecki proposed a causal explanation for the distribu- Vecchione et ah, in press). If Alexeyev's (1992) findings tion of Doryteuthis (Doryteuthis) - he suggested that a are accurate, they must be accounted for in future phyloge- slight rise in water temperature in the northern Pacific netic studies ofthis group. Ocean could have allowed the L. bleekeri-opalescens com- More detailed studies ofthe primary synapomorphy mon ancestor to disperse from the eastern Pacific coast of that unites the species of Photololigo - the photophores North America to Japan via the Aleutians. The results of themselves - might help resolve these problems. In the weighted analyses support a sister-species relationship Euprymna, the bioluminescent organ is the product of a between L. bleekeri and L. opalescens, but, due to the lack complex interaction between the squid and symbiotic lumi- of resolution of relationships among other Doryteuthis nescent bacteria (McFall and Ruby, 1991). Further investi- (Doryteuthis) species, a monophyletic subgenus Dory- gations of this interaction and its effects on photophore teuthis (sensu Brakoniecki, 1986) remains a possibility, but morphology in all photololiginid squids (e. g. Haneda, is not directly supported. Due to the overall lack ofresolu- 1963; Pringennies and Jorgensen, 1994) could illuminate tion, the possible ancestral range of the L. bleekeri- species relationships within the clade. opalescensancestorcannotbe examined. The other major clade found in the weighted analy- In addition to these putative clades, certain species ses is very similar to Natsukari's (1983) Nipponololigo, a presently grouped in the genus Lolliguncula (L. panamen- proposed subgenus of Loligo comprised of L. japonica, L. sis, L. mercatoris, and L. brevis), together with Loliolopsis uyii, L. kobiensis, and L. beka. The successive weighting diomedeae, were found in all strict consensus trees from all analyses support the inclusion of L. sumatrensis and the weighted analyses. However, the position of Lolliguncula Loliolus affinis-hardwickei sister-species grouping within a argus was variable across these trees. In some consensus broader Nipponololigo clade. The two synapomorphies trees, L. argus was completely outside the clade comprised uniting these species are the suckermorphology ofthe dor- of the rest of the Lolliguncula species plus Loliolopsis sal and ventral rowsofthe hectocotylus. The pedicelsofthe diomedeae. In other trees, L. argus was found to be ahigh- dorsal row suckers are fused with their protective mem- ly derived member of the Lolliguncula + Loliolopsis clade. brane and widened into fleshly slabs (Natsukari, 1983; Due to the variable position ofthis taxon, the Lolliguncula Brakoniecki, 1986). In most of these species, the slabs + Loliolopsis clade collapsed in the overall strict consensus retain small suckers; inL. affinis and L. hardwickei, howev- cladogram (Fig. 2). er, the suckers are not present on the tops of the slabs. In Successive weighting analyses can provide heuristic ANDERSON: LOLIGINID SQUID PHYLOGENETICS 121 insight into the information value of the characters in the only through an examination of multiple sources of data, analysis. Most informative, consistent characters found including morphological, anatomical, and molecular after multiple rounds of successive weighting are invariant sequence data. within the ingroup (Table 2). Few characters that vary with- in the ingroup appear to have high rescaled consistency indices across all initial unweighted trees. Also, several ACKNOWLEDGMENTS characters show varying amounts of homoplasy across ini- tial unweighted most-parsimonious trees, and subsequently have low weights aftersuccessive weighting analyses. The authorwould like to thankJ. S. Pearse, A.T. Newberry, A. The overall lack of resolution in strict consensus J. Gong, M. deMaintenon, R. Guralnick, C. Hedegaard, C. Meyer, B. Simison, R. E. Young,andananonymousreviewerfortheircommentson trees found after these unweighted and weighted analyses the manuscript and other assistance. J. Voight, N. Voss, C. F. E. Roper, and the low weights of many characters after successive M. Vecchione, M. Sweeney, T. Gosliner, G. Pogson, D. Eemisse, and R. weighting highlight the limited utility of using only gross Krosigkallassistedwiththisprojectbyprovidinghelpfuladvice,accessto external morphological characters to investigate loliginid museumcollections,orcommentsonanalytical techniques. Thisresearch squid phylogeny. Despite this general conclusion, external was supported in part by awards from the Earl H. Myers and Ethel M. Myers Marine Biology and Oceanographic Trust Fund, the Friends of morphological characters should not be ignored in future Long Marine Laboratory, the American Museum of Natural History investigations of loliginid relationships. The Hillis and Lerner-Grey Fund for Marine Research, and UCSC Department of Huelsenbeck (1992) gi test shows significant phylogenetic BiologyIndependentStudyfunds. structure in the data matrix. Some of the characters used in this analysis do appear to carry appreciable phylogenetic information. Undoubtedly, more data must be gathered to testthe results ofthese analyses and to resolve relationships LITERATURE CITED among these squids. Little is known about comparative internal anatomy in loliginid squids, although excellent Adam, W. 1950. Notes sur les cephalopodes, XXII. Deux nouvelles studies have been done ofparticular species (e.g. Williams, especesde lacoteafricaineoccidentale. BulletinduInslitutRoyal 1909). Forexample, the anatomy ofthe nervous system and Adam,Wd.es19S5c4i.enCceepshNaaltouprocdlal,esIdV.eBCcelpghiaqlueop2o6d:e1s-9a.1'exclusiondesgenres circulatory system, and perhaps aspects of juvenile devel- Sepia,SepiellaetSepioteuthis.Siboga-Expeditie55c:123-198. opment, are particularly promising systems for inclusion in Alexeyev, D. O. 1992. The systematic position ofbioluminescent squids cladistic analysis. Many neurophysiological studies have offamily Loliginidae (Cephalopoda, Myopsida). Zoologicheskii been performed on a broad range of loliginid squids, Zhurnal71:12-23. including Loligo opalescens, L. pealei, L. vulgaris, Augustyn, C. J. and W. S. Grant. 1988. Biochemical and morphological systematics ofLoligo vulgaris vulgaris Lamarck and Loligo vul- Sepioteuthis lessoniana, Lolliguncula brevis, and garis rcynaudi d'Orbigny nov. comb. (Cephalopoda: Myopsida). Alloteuthis media (e.g. Brown et al., 1991; Chrachri and Malacologia29:215-233. Williamson, 1993; Fishman and Metuzals, 1993; Preuss Berry, S. S. 1911. A note on the genus Lolliguncula. Proceedings ofthe and Budelmann, 1995). Due to the ease of culturing some AcademyofNaturalSciencesofPhiladelphia 1911:100-105. Berry, S. S. 1929. Loliolopsischiroctes, anew genusandspeciesofsquid ofthese species in the laboratory (Lee etal, 1994), and the from the Gulf of California. Transactions ofthe San Diego giant axonsofmany loliginids, more comparative neurolog- SocietyofNaturalHistory5:263-282. ical studies undoubtedly will be performed. These data Berthold,T. andT. Engeser. 1987. Phylogeneticanalysisandsystematiza- could be combined with other morphological and anatomi- tion of the Cephalopoda (Mollusca). Verhandlungen des cal dataforcladistic analysis. NatunvisscnschaftlichenVereinsinHamburg29:187-220. Brakoniecki, T. F. 1980. Lolliguncula tydeus, a new species of squid In addition to internal anatomical information, mol- (Cephalopoda: Myopsida) from the Pacific coast of Central ecular data could aid in resolving loliginid relationships. America. BulletinofMarineScience30:424-430. Recently, Yeatman and Benzie (1994) have found genetic Brakoniecki, T. F. 1984. A full description of Loligo sanpaulensis evidence of cryptic speciation within Photololigo edulis Brakoniecki, 1984 and a redescription ofLoligogahi d'Orbigny, andP. chinensis, providing evidence ofthe powerofmolec- 1835,twospecies fromthesouthwestAtlantic.BulletinofMarine Science34:435-448. ular techniques in species-level research ofloliginid squids. Brakoniecki, T. F. 1986. A Generic Revision ofthe Family Loliginidae An ongoing sequencing study of two mitochondrial genes (Cephalopoda: Myopsida) Based Primarily on the Comparative DNA (the 16S ribosomal gene and the cytochrome c oxi- Morphology of the Heclocolylus. Doctoral Dissertation, dase subunit I gene) could shed light on loliginid relation- UniversityofMiami,CoralGables,Florida. 164pp. ships (Anderson, unpub.). Relationships among loliginid Brakoniecki, T. F. 1996. A revision ofthe genus Pickfordiateuthis Voss, 1953 (Cephalopoda: Myopsida). BulletinofMarineScience58:9- squids at the species level and investigations of cladogene- 28. sis and biogeography within this group will be possible Brakoniecki, T. F. and C. F. E. Roper. 1985. Lolliguncula argus, a new species of loliginid squid from the tropical eastern Pacific. . . 122 AMER. MALAC. BULL. 12(1/2) (1996) ProceedingsoftheBiologicalSocietyofWashington98:47-53. Gilbert, D. L., W. J. Adelman, Jr., andJ. M. Arnold, eds. 1990. Squidas Brierley,A. S. andJ. P.Thorpe. 1994. Biochemical geneticevidencesup- ExperimentalAnimals. PlenumPress,NewYork.516pp. porting the taxonomic separation of Loligo gahi from the genus Haneda, Y. 1963. Observations in the luminescence ofthe shallow-water Loligo.AntarcticScience6:143-148. squid Uroteuthis bartschi. Science Report ofthe Yokosuka City Brooks, D. R. and D. A. McClennan. 1991. Phylogeny, Ecology and Museum8:10-15. Behavior: a Research Program in Comparative Biology. Uni- Hanlon,R.T,S. v. Boletzky,T. Okutani,G. Perez-Gandaras,P. Sanchez, versityofChicagoPress,Chicago.434pp. C. Sousa-Reis, and M. Vecchione. 1992. SuborderMyopsida. In: Brown, E. R., Q. Bone, K. P. Ryan, and N. J. Abbott. 1991. Morphology "Larval" and Juvenile Cephalopods: a Manualfor Their and electrical properties ofSchwann cells around the giant axon Identification, M.J. Sweeney,C.F.E.Roper, K. M. Mangold,M. ofthe squids Loligoforbesi and Loligo vulgaris. Proceedings of R. Clarke, and S. v. Boletzky, eds. pp. 37-53. Smithsonian the Royal Society ofLondon Series B - Biological Sciences Contributions to Zoology 513, Smithsonian Institution Press, 243:255-262. Washington,D.C. Burgess, L. A. 1967. Loliolus rhomboidalis, a new species ofloliginid Hess, S. C. 1987. Comparative Morphology, Variability, and Systematic squid from the Indian Ocean. Bulletin ofMarine Science 17:319- Applications ofCephalopod Spermatophores (Teuthoidea and 329. Vampyromorpha). Doctoral Dissertation, University of Miami, Camin, J. H. and R. R. Sokal. 1965. A method for deducing branching CoralGables,Florida. 581 pp. sequencesinphylogeny.Evolution 19:311-326. Hillis, D. M. andJ. P. Huelsenbeck. 1992. Signal, noise andreliability in Carpenter, J. M. 1988. Choosing among multiple equally parsimonious molecularphylogeneticanalyses.JournalofHeredity83:189-195. cladograms.Cladistics4:291-296. Kluge, A. J. and J. S. Farris. 1969. Quantitative phyletics and the evolu- Carpenter, J. M. 1994. Successive weighting, reliability and evidence. tionofanurans. SystematicZoology 18:1-32. Cladistics 10:215-220. Lee, P. G., P. E. Turk, W. T. Yang, and R. T. Hanlon. 1994. Biological Chrachri,A.andR. Williamson. 1993. Electricalcouplingbetweenprima- characteristics and biomedical applications of the squid ry hair cells in the statocyst of the squid, Alloteuthis subulata. Sepioteuthis lessoniana cultured through multiple generations. NeuroscienceLetters 161:227-231 BiologicalBulletin 186:328-341. Cohen,A.C. 1976.ThesystematicsanddistributionofLoligointhewest- Lu, C. C, C. F. E. Roper, and R. W. Tait. 1985. A revision ofLoliolus ern North Atlantic, with description of two new species. (Cephalopoda: Loliginidae), includingL. noctiluca, anewspecies Malacologia 15:229-367. ofsquidfromAustralian waters.ProceedingsoftheRoyalSociety detvlaintenon, M. J. 1990. Comparative Morphology ofthe Cranium and ofVictoria97:59-85. Brachial Cartilage in the Cephalopod Suborder Myopsida Lu, C. C. and R. W. Tait. 1983. Taxonomic studies on Sepioteuthis (Cephalopoda: Teuthoidea). Masters Thesis, University of Blainville(Cephalopoda: Loliginidae)fromtheAustralianregion. Miami,CoralGables,Florida.63pp. ProceedingsoftheRoyalSocietyofVictoria95:181-204. de Queiroz, K. andJ. Gauthier. 1990. Phylogeny as a central principle in Maddison, D. R. 1991. The discovery and importance ofmultiple islands taxonomy - phylogenetic definitions oftaxon names. Systematic ofmost-parsimonioustrees.SystematicZoology40:315-328. Zoology39:307-322. Maddison,W.P. 1993. Missingdataversusmissingcharactersinphyloge- de Queiroz, K. and J. Gauthier. 1992. Phylogenetic taxonomy. Annual neticanalysis.SystematicBiology42:576-581 ReviewofEcologyandSystematics23:449-480. Maddison, W. P., M. J. Donoghue, and D. R. Maddison. 1984. Outgroup Drew, G. A. 1911. Sexual activities ofthe squid, Loligopealei (Les.). 1. analysisandparsimony.SystematicZoology33:83-103. Copulation, egg-laying and fertilization. Journal ofMorphology Maddison, W. P. and D. R. Maddison. 1992. MacClade: Analysis of 22:327-352. Phylogeny and Character Evolution. Version 3.04. Sinauer Edwards, A. W. F.and L. L. Cavalli-Sforza. 1964. Reconstructionofevo- Associates,Sunderland, Massachusetts. lutionary trees. In: Pheneticand Phylogenetic Classification. V. McFall, M. and E. G. Ruby. 1991. Symbiont recognition and subsequent H. Heywood and J. MacNeill, eds. pp. 67-76. Systematics morphogenesis as early events in an animal-bacterial symbiosis. AssociationPublication6,London. Science254:1491-1494. Eernisse, D. J., J. S. Albert, and F. E. Anderson. 1992. Annelida and McGowan,J. A. 1954.Observationsonthesexualbehaviorandspawning Arthropoda are not sister taxa: a phylogenetic analysis ofspi- ofthesquid,Loligoopalescens,atLaJolla,California.California ralianmetazoanmorphology. SystematicBiology41:305-330. FishandGame40:47-54. Farris, J. S. 1969. A successive approximations approach to character Morejohn,G. V.,.1. T. Harvey, and L.T. Krasnow. 1978.Theimportance weighting.SystematicZoology 18:374-385. ofLoligo opalescens in the food web ofmarine vertebrates in Farris, J. S. 1983. The logical basis of phylogenetic analysis. In: Monterey Bay, California. In: Biological, Oceanographic, and Advancesin Cladistics, Volume2, N. I. Platnick andV. A. Funk, AcousticAspectsoftheMarketSquid, Loligoopalescens(Berry), eds. pp. 7-36. Proceedings ofthe Second Meeting of the Willi C. W. Recksiek and H. W. Frey, eds. pp. 67-98. California HennigSociety,ColumbiaUniversityPress,NewYork. DepartmentofFishandGameFish Bulletin 169. Farris,J. S. 1989. Theretentionindex andtherescaledconsistency index. Nateewathana, A. 1992. Taxonomic studies on loliginid squids Cladistics5:417-419. (Cephalopoda: Loliginidae) from the Andaman Sea coast of Fields, W.G. 1965. The structure,development, foodrelations,reproduc- Thailand. Phuket Marine Biological Center Research Bulletin tion, and life history of the squid Loligo opalescens Berry. 57:1-40. Fishman,CaHl.ifMor.niaandFiJs.hMaentduzGalasm.e1B9u9l3l.etCina21+3-1i:n1d-u1c0e8d.axosome formation Natsukaris,quYi.ds19-7I6I..DTeasxcornipotmiiocnaannddmnoerwphroelcoogridcoaflLsotluidigeossoinbotghaeel(olAidgainmi,d in internally dialyzed giant axons ofLoligopealei. Biological 1954)fromFormosa.Venus35:15-23. Bulletin 185:292-293. Natsukari, Y. 1983. Taxonomical and morphological studieson the lolig- Fitch, W. M. 1971. Toward defining the course ofevolution: minimal inid squids - III. Nipponololigo, a new subgenus ofthe genus change for a specific tree topology. SystematicZoology 20:406- Loligo. Venus42:313-318. 416. Natsukari, Y. 1984. Taxonomical andmorphological studieson the lolig-