Austinixa aidae Righi, 1967 and A. hardyi Heard and Manning, 1997 (Decapoda: Brachyura: Pinnnotheridae) synonymized, with comments on molecular and morphometric methods in crustacean taxonomy

While conducting a systematic study of Austinixa Heard and Manning, 1997, we encountered difficulties in distinguishing A. hardyi Heard and Manning, 1997 from A. aidae Righi, 1967, based on characters discussed in the current literature. We re‐examined morphological characters previously reported as diagnostic, and assessed their variation within and between these putative species. We expanded the analysis to include three independent types of data: discrete morphological characters, morphometric characters, and mitochondrial DNA sequence data. Partial sequences of the mitochondrial 16s rDNA and cytochrome oxidase I (COI) genes were almost identical between A. aidae and A. hardyi, raising questions about the validity of A. hardyi as a distinct species. Analysis of morphometric characters using multivariate (principal components analysis) and univariate allometric adjustment methods, support the mitochondrial DNA data in suggesting A. aidae and A. hardyi are a single species. Finally, we provide a brief review of the morphometric and molecular methods used in this study and comment on the usefulness of these methods in crustacean taxonomy.


Introduction
Crabs of the family Pinnotheridae live as commensals or parasites of other invertebrates, and are found in nearly all of the world's oceans (Schmitt et al. 1973). Taxonomy of this group has been historically difficult because of small size, cryptic habitat, sexual dimorphism, and often complicated life histories. For example, the males, females, and juveniles of Fabia concharum Rathbun, 1893 were described originally as three distinct species (Davidson 1968). Many species have been recorded as the most abundant macroinvertebrates in the areas where they are found, yet little is known about them differentiating species (Moritz 1994;Sites and Crandall 1997;Avise 2000). There are numerous cases where morphologically cryptic species were first detected by large genetic discontinuities. When combined with other types of data (i.e. morphological or nuclear DNA markers), mtDNA markers can be useful in differentiating species. We applied this method for the first time in pinnotherid crabs.
Three main objectives are addressed in this study. First, the taxonomic status of A. hardyi with respect to A. aidae is assessed through the analysis of discrete morphological, morphometric, and molecular characters. Second, morphometric variation is quantified within and between these putative species. Third, we provide a brief review of the morphometric and molecular methods used in this study and comment on the usefulness of these methods in crustacean taxonomy.

Samples used
Museum specimens of A. hardyi (13 males, 13 females) were obtained from the National Museum of Natural History, Smithsonian Institution (USNM). A paratype male was first photographed for morphometric analysis, and then destroyed for DNA extraction. Museum specimens of A. aidae (eight males, one female) were obtained from the Museo de Zoologia, Universidade de São Paulo (MZUSP). In both cases specimens examined included type specimens. Two additional males of A. aidae were collected by Alberto Souza south of the type locality of Rio Juqueriquere, Caraguatatuba, São Paulo, Brazil, and preserved in 95% ethanol. One of these specimens was used in morphometric analysis as described, and subsequently destroyed for DNA extraction.

Molecular data collection
DNA extraction was performed with the Qiagen Qiamp Tissue Kit following the procedure outlined by the manufacturer (Qiagen Inc.), with the following exceptions: before addition of proteinase K, whole crabs (<20 mg) were ground in extraction buffer inside 1.5 ml tubes. After digestion with proteinase K, tubes were centrifuged and buffer containing digested tissue and DNA was transferred to a new tube to eliminate remaining exoskeleton. For the museum specimen of A. hardyi, the final DNA elution step was performed with a reduced volume of 100 ml and repeated three times. This increased the final concentration of DNA but reduced overall yield. There was no possibility for cross-contamination between samples as DNA extraction, polymerase chain reaction (PCR) amplification, and sequencing for each sample were performed 6 months apart, with samples never having the possibility of contact.
PCR amplification was performed in 50 ml reactions containing 5 ml 106 PCR buffer, 2.5 mM MgCl 2 , 0.2 mM dNTPs, 2-6 ml template DNA, 0.5 mM of each primer, and 1.25 units of Taq polymerase. Reactions were performed with an initial denaturation step at 94uC for 3 min, then 30-40 cycles of 94uC, 30 s/50uC, 90 s/72uC, 120 s, and a final extension step at 72uC for 10 min. After processing with exonuclease I and shrimp alkaline phosphatase, double-stranded PCR products were sequenced in both directions, using an ABI 373/377 automated sequencer and an ABI Big-Dye Ready Reaction Kit, following the standard cycle-sequencing protocol. COI was sequenced in both directions using the same primers used in PCR amplification. 16s was sequenced with internal primers (16saint 59-TTAATTCAACATCGAGGTCG-39 and 16sbint 59-GATATAAGAAGTCTAGCCTG-39) designed for this study. Sequences were aligned using the program Sequencher 3.0 (Gene Codes Corp.) and observed for nucleotide differences.

Discrete morphological characters
The following characters were compared between A. aidae (10 males, one female) and A. hardyi (13 males and 13 females): patch of setae on dorsal surface of carapace present or absent, ridge on the posterior surface of P4 dactylus present or absent, number of carinae on propodus of P4, movable spine on the basal segment of antennular (A1) flagellum present or absent, and apex of male gonopod sharply curved laterally or straight. These characters were reported as diagnostic in the description of A. hardyi (Heard and Manning 1997).

Morphometric data collection
Morphometric data were taken by capturing digital images with a JVC KY-F30 highresolution video camera attached to an Olympus SZH10 dissecting microscope. Images were changed to grayscale images and saved as TIFF files using Adobe Photoshop 4.0 (Adobe 1997). An image of a scale was taken by the same method at each magnification. Morphometric measurements were then acquired from the digitized images using the program Scion Image (Scion 1998). All images taken at the same magnification and the corresponding scale image were opened simultaneously. The scale was first calibrated and then all images were calibrated simultaneously to this scale. This was repeated for all magnifications used. Character and scale images were saved as calibrated, assuring that necessary re-measurements or calibration of future images are identical in calibration. This method increases both accuracy and precision in taking morphometric measurements.
Characters were measured for males only because only one female A. aidae was available. The following measurements were performed on the digitized images: carapace length (CL)5the distance from the tip of the rostrum to the cardiac ridge; carapace width (CW)5the distance of the greatest width, measured between the lateral extremes of the carapace just above the coxae of the pereiopods; telson width at the proximal end (TW); width of the sixth abdominal segment at the proximal end (Ab6WP); minimum width of the sixth abdominal segment (Ab6WMN); cheliped length-posterior (ChLP)5the distance along the posterior side from the tip of the fixed finger to the proximal end of the propodus; cheliped length-anterior (ChLA)5the distance along the anterior side from the dactyl/ propodus joint to the proximal end; maximum cheliped height (ChHMX); and cheliped height at the proximal end (ChHP). The following measurements were made on pereiopods 2-5 (P2, P3, P4, or P5 preceding measurement abbreviations to indicate appropriate pereiopod number): anterior merus length (MLA)5distance from the proximal to the distal end of the merus along the anterior side; posterior merus length (MLP)5distance from the proximal to the distal end of the merus along the posterior side; merus width (MW)5distance at the greatest width of the merus; anterior propodus length (PLA); posterior propodus length (PLP); propodus width (PW); and dactyl length (DL). All cheliped and pereiopod measurements were taken along the ventral surface. All measurements were taken to the nearest 0.01 mm.

Morphometric analysis
To compare morphometric characters it is important to use a methodology able to separate ontogenetic and evolutionary allometry (Cock 1996;Klingenberg 1996). In addition, because most measures of shape are in some way related to size, heterogeneity in mean body size between samples must be taken into account before meaningful comparisons can be made (Reist 1985). Because samples of the two nominal species differed in mean body size, a common multivariate technique (PCA), and univariate technique for allometric correction (Thorpe 1975;Reist 1985) were employed. Reist (1985) suggested that the best approach to use in allometric situations is the univariate allometric adjustment method (ALLOM) of Thorpe (1975). With this method individual characters are adjusted to predictions of what an individual's size for a particular variable would be if that individual were the overall mean standard size. The character adjustments are made by an allometric formula: where Y i is the predicted character value of the ith specimen and the exponent k is the log of the adjusted measurement (e) derived from the equation: Here, y i is the original unadjusted measurement of the ith specimen, b is the allometric coefficient (the slope of the relationship between the log y i and the log x i ), x i is the standard size measurement of the ith individual, and X SSM is the grand mean of the standard sizes. Log is the base-10 logarithm. For this study, the standard size measurement used was carapace width (CW). Use of a univariate adjustment of characters based on CW (i.e. use of CW as a standard size measurement) is justified because CW was correlated strongly with generalized multivariate size in the data set (logCW and PC1 extracted by a PCA using the covariance matrix r50.98, P,0.0001). After allometrically adjusted values for each character were determined, the values were then compared between samples by a one-way ANOVA.
PCA of covariance matrices was used to explore the distribution of individuals in multivariate morphospace and to determine variability within and among samples. Many authors have suggested the use of covariance matrices when data sets are composed of the same kind of measurements (e.g. lengths) (Blackith and Reyment 1971;Wiley 1981). PCA was performed on log-transformed data to equalize standard deviations of measurements across different-sized variables and to help ensure multivariate normality. Because size is usually the variable with highest variance in a data set, the first principal component extracted from a covariance matrix is often a generalized size factor. The remaining components account mostly for shape differences. The PC coefficients of a covariance matrix can therefore be used to account for allometry and size differences (Humphries et al. 1981).
PCA of a correlation matrix does not allow information about allometries to be derived from resulting PC coefficients (Rincon 2000), but it does allow ordination of individuals in multivariate morphometric space. Additional PCA, using the correlation matrix of allometrically adjusted characters derived using Equation 1, was carried out to provide additional perspective and complementary insight and contrast to PCA of the covariance matrix.

Mitochondrial DNA
No difference was detected between individuals of either species in the 554 bp region of the 16s rDNA gene. In the 693 bp region of COI, individuals of A. aidae and A. hardyi differed by only two base pairs (0.28%).

Discrete morphological data
No consistent differences in discrete morphological characters were found between specimens identifiable as A. aidae and A. hardyi. Morphological character data are summarized in Table I for both males and females of each species. Heard and Manning (1997) stated that A. hardyi differed from other species of Austinixa in possessing dense patches of setae on the carapace anterior to the cardiac crest in males. This bilobed patch of setae was present in all male specimens of A. aidae examined here. However, the type specimen of A. aidae (a male) had considerably fewer setae in this area than did other male specimens of A. aidae. This may be a consequence of setae being partially lost in preservation. It is possible also that the number of setae may decrease in larger specimens, as the type specimen of A. aidae is 2.5 mm wider than any other individual observed. Nonetheless, a remnant of the bilobed patch of setae is clearly visible in the type specimen. Heard and Manning (1997) also stated that a ventral ridge on the posterior surface of the P4 dactylus distinguished A. hardyi from A. aidae. We found that a dorsal and ventral sulcus was present on the posterior surface of the dactylus, with ridges varying from barely detectable to more pronounced in specimens of both nominal species. Consequently, this character appears not to be diagnostic.
Gonopod tip structure has been found to be extremely useful in the identification of cryptic species in pinnotherid crabs (Manning and Felder 1989). Scanning electron microscope (SEM) images of gonopod tips showed identical structure for specimens of both species (Figure 1).  (Table II). The subsequent components were of mixed sign and variable weight. This suggests that the variation along PC1 is size related, thus not informative with respect to shape; whereas PC2 and PC3 can be treated as shape vectors (Blackith and Reyment 1971;Wiley 1981).
Projections of individual scores along the first and second principal components showed divergence, with slight overlap along both PC1 and PC2 (Figure 2). Projection of individuals along second and third principal components showed near-complete overlap along PC3 and divergence, with some overlap along PC2 (Figure 3). Separation of specimens of each nominal species along PC2 was due largely to propodal width of third pereiopod (P3PW) and anterior cheliped length (ChLA). Other characters weighting heavily are summarized in Table II. A second PCA, using the correlation matrix of allometrically adjusted characters, indicated that the first, second, and third principal components accounted for 33.5%, 17.7%, and 14.1% of the total variance, respectively (Table III). Here, the effect of size had been previously removed. PC1 is now a component that reflects mostly the contribution of the dactyl length of P5 (P5DL), carapace length (CL), and the propodal length of P3 (P3PLA). Additional characters that appeared to weight heavily are summarized in Table III, but were generally measures of the P3 propodus and P5.
Scatterplots of PC1 versus PC2 of the correlation PCA showed divergence with some overlap, along PC1, with complete overlap along PC2 (Figure 4). Projections of all other principal components showed complete overlap, providing no additional discrimination between A. aidae and putative A. hardyi.
The allometric model generally revealed a good fit to the data. Carapace width (CW) explained a high proportion of variation in morphometric characters (81.3-96.5%), and allometric coefficients differed significantly from zero. Results of an ANOVA of allometrically adjusted characters are summarized in Table III. The majority of the morphometric characters showing significant differences between A. aidae and putative A. hardyi were associated with P3 and P5. These included all morphometric variables of the propodus, merus width, and dactyl length. Propodal width of the second pereiopod and carapace length also were found to differ significantly.

Taxonomy and distribution
Although morphometric variation was found when comparing types and syntypes of A. aidae and A. hardyi, no consistent diagnostic differences in either morphology or mtDNA sequences were found. The data do not support the existence of two species, and we conclude that A. hardyi should be considered a junior synonym of A. aidae.
The known distribution of A. aidae is expanded from its previously only known location, Rio Juqueriquere, Caraguatatuba, São Paulo, Brazil to the island of Tobago. Rio Juqueriquere appears to be the southern limit of this species. It is replaced to the south by a congener, A. patagoniensis Rathbun, 1918. Based on Heard and Manning (1997), A. aidae appears to be more common in Tobago than in the southern part of its range. The host of A. aidae has not been recorded in Brazil, but like most other species of Austinixa, specimens from Tobago prefer a thallassinidean shrimp of the Callichirus major complex as a host (Heard and Manning 1997

Morphometric variation
The data presented here indicate morphometric differences between specimens from Tobago and those from Brazil. These included width of the P3 and P5 propodus, and P5 dactylus length. The ALLOM method indicated that specimens from Brazil had a significantly longer carapace than those from Tobago, an observation also made by Heard and Manning (1997). Morphometric variation has been observed in other species of pinnotherid crabs. Individuals of Pinnixa franciscana Rathbun, 1918 found in protected coastal habitats differ from those found in open-coast habitats in extent of lateral elongation of the carapace, length of the P4 propodus, and degree of pronouncement of a cardiac ridge. This intraspecific variation was hypothesized to stem from ecological differences (Rathbun Table III. Weights for the first three principal components computed from the correlation matrix from the allometrically adjusted data; associated P values for an ANOVA of allometrically adjusted characters (ALLOM) comparing Austinixa aidae and putative A. hardyi. 1918; Zmarzly 1992). Alves and Pezzuto (1998) noted different growth rates between biyearly recruitment classes of A. patagoniensis, possibly due to lower water temperatures during one of the recruitment periods. A similar pattern was observed in Pinnotheres ostreum Say, 1817 (Christensen and McDermott 1958). Differences in growth rates have the potential to affect growth patterns and resulting shapes of morphological characters. Morphometric variation observed in A. aidae may be due to ecological adaptations to factors such as host or substrate differences between locations. Studies of development and a fine-scale geographic assessment of morphology are needed to better understand patterns of variation in A. aidae.

Review of morphometric methods
Results of the PCA of the covariance matrix provided questionable results that were not consistent with results obtained from allometric character adjustment methods. PC2 was the main axis indicating variation between A. aidae and A. hardyi. Three of the four cheliped measurements weighted quite heavily along PC2, and ChLA had the second highest weight on this axis. Other methods of comparisons employed not only contradicted this, but strongly indicated no statistical differences in these characters with P values ranging from 0.75 to 0.98. Simple observation of log-log bivariate plots also gave no indication of systematic bias in these characters. Some authors have suggested that PCA may produce spurious results when PC1 accounts for extremely high levels of the total variation. In this situation, the remaining PCs, which are assumed to contain the variation of interest (shape variation), account for such a small amount of the total variation that each PC may not have a unique meaning, PC1 may contain shape information, and the remaining PCs may also contain size information (Blackith and Reyment 1971;Humphries et al. 1981;Reist 1985;Rincon 2000). Although there is no consensus on how much variation is too much in PC1 and too little in the remaining PCs, with 93.7% of the variation accounted for by PC1, and only 6.3% by PC2-PC13, PCA may produce misleading results with this data set. Numerous univariate methods have been applied to taxonomic studies using morphometric data sets to deal with the problem of separating size and shape information. These methods include logarithmic transformations, ratios, logarithmic transformation of ratios, and allometric character adjustment methods (Reist 1985). Debate continues as to which method best separates shape information and size influences due to either heterogeneity in size between samples or different allometric patterns. A host of potential problems are known to exist when using ratios. However, ratios continue to be the most commonly used method. Potential problems with using ratios include compounding error terms, creating or increasing correlation between numerator variables, and most importantly not removing the effects of size if allometric relationships exist between variables (Atchley et al. 1976;Thorpe 1976;Atchley and Anderson 1978;Reist 1985).
The allometric adjustment method of Thorpe (1975) has yet to be used to analyze morphometric characters in systematic studies of crustaceans and most other invertebrate taxa. Reist (1985) compared numerous univariate and multivariate methods of morphometric analysis and determined this allometric adjustment method to be superior in generating shape variates that are independent of size and in having minimal adverse effects on the data. Future systematic studies of crustacean taxa, using morphometric data, would benefit from using this method in place of the common practice of using ratios.

Mitochondrial DNA sequence variation
The mitochondrial gene sequences compared in this study have been shown to be variable between species of Austinixa, with interspecies sequence divergence ranging from 3 to 9% for 16s and 13 to 22% for COI (Harrison 2004). The sibling species A. behreae Felder, 1989 andA. chacei Wass, 1955 in the northern Gulf of Mexico are almost indistinguishable morphologically, and show sequence divergences of 3 and 13% for 16s and COI, respectively (Harrison 2004). The amount of variation seen between individuals from Brazil and Tobago are characteristic of that between individuals of an interbreeding population. It is important to emphasize that only one individual was sampled from each location, which is significantly below the sample size needed to accurately estimate genetic diversity within and among populations (Crandall and Templeton 1993). Nonetheless, the observed lack of variation between individuals contributes important information to the question being addressed, and adds support to the morphological data in rejecting the hypothesis of the existence of two species.

Mitochondrial DNA data and taxonomy
Data from mtDNA have provided scientists with a wealth of information about natural variation within and among species and some comments on its use in studies of species discrimination are in order. mtDNA is haploid, non-recombining, and maternally inherited in most organisms. Therefore, the effective population size is one-quarter that of a singlecopy nuclear locus. This has important repercussions in that a mitochondrial locus will sort to fixation approximately four times faster than nuclear loci. This results in fixed diagnosable haplotypes being common between structured populations within a species (Sites and Crandall 1997). In addition, in the event of hybridization, the mtDNA haplotype of one species can rapidly introgress into the other, making the species indistinguishable with respect to mitochondrial markers (Sites and Crandall 1997;Slowinski and Page 1999). For these reasons, mtDNA markers should not be used exclusively in species definition, but can contribute important insight when used in combination with other types of data, including morphological data and nuclear DNA markers (Moritz 1994;Sites and Crandall 1997;Avise 2000).

Summary
This is the first study to combine morphology, morphometric, and genetic methods in a taxonomic study of pinnotherid crabs. Although the univariate allometric adjustment method employed in this study is commonly used in vertebrate taxa, this is the first study to use it for taxonomic comparison in decapod crustaceans. The use of a combination of three independent types of data provided strong support for the existence of only a single species while allowing quantification of morphometric variation that exists between extreme locations in the distribution.