Introduction
Bryozoans are small benthic suspension feeders that play important roles in marine communities, especially in the Southern Hemisphere. Among the three classes of bryozoans, Phylactolaemata are found in fresh water, Gymnolaemata are predominantly marine, and Stenolaemata are exclusively marine. Most stenolaemate taxa are ancient and only known from the fossil record, with Cyclostomatida the sole surviving order.
Bryozoans are colonial animals, and each colony module, or zooid, is traditionally divided into a cystid (the zooid wall) and polypide (which can be retracted into the cystid or partly everted from it). The cystid includes the living portion of the zooid body wall and a nonliving, often calcified, portion (including cuticle) that provides structural support. The polypide comprises the gut, lophophore with tentacles, introvert, muscles, and ganglion (e.g., Borg Reference Borg1926; Ryland Reference Ryland1970) (Fig. 1).
Schematic drawings of the proximal parts of the bryozoan polypide, showing characters recorded in this study. A, A scheme of a sagittal section through the zooid with everted polypide. B, Top view of the protruded polypide. C, Types of the tentacle crown symmetry. D, Types of the crown shape, resulting from different tentacle curvatures.
Together, introvert and lophophore make up the specialized feeding apparatus of bryozoans. The introvert is the part of the polypide that serves as a flexible stalk supporting the lophophore when everted and as a tentacle sheath when retracted. The lophophore is a widened platform around the mouth that bears a fringe of ciliated tentacles and contains a coelomic cavity. In Phylactolaemata the lophophore is U-shaped, while Gymnolaemata and Stenolaemata have circular lophophores.
The tentacles are the main food-gathering elements of the lophophore; they bear cilia that generate currents and take part in capturing or rejecting particles. Each tentacle typically bears two or three types of ciliary bands: frontal, latero-frontal, and lateral (Mukai et al. Reference Mukai, Terakado, Reed and Harrison1997). The latter are the longest and play the principal role in feeding, because their strokes generate water currents (Borg Reference Borg1926).
Naturally, cystid and polypide morphology are interconnected (see, e.g., summary by McKinney Reference McKinney1990). Polypide size and shape are dictated by adaptation to feeding; the dimensions of the feeding apparatus are related to the dimensions of captured food (Winston Reference Winston, Woollacott and Zimmer1977; Sanderson et al. Reference Sanderson, Thorpe, Clarke, Cubilla and Jackson2000). Also, the characters of the tentacle crown—its diameter and the length and number of the tentacles—are directly related to the number of lateral cilia (Ryland Reference Ryland1970, Reference Ryland1976; Riisgård and Manríquez Reference Riisgård and Manríquez1997), and thus to feeding current speeds (Best and Thorpe Reference Best and Thorpe1986), pumping (Riisgård and Manríquez Reference Riisgård and Manríquez1997), and clearance rates (Strathmann Reference Strathmann1973; Winston Reference Winston, Woollacott and Zimmer1977).
In addition, the protrusion–retraction mechanism employed by bryozoans limits the height/depth of tentacle protrusion, thus determining some morphological relationships (Taylor Reference Taylor, Larwood and Nielsen1981). During protrusion, phylactolaemates and gymnolaemates change the shape of the zooid walls (either directly or by using thin-walled sacs filled with or emptied of seawater). By contrast, Recent stenolaemates have inflexible zooid walls and rely on their own body cavity fluid instead of seawater to compensate for lophophore protrusion. Their trunk coelom is split into exosaccal and endosaccal cavities, and the former compensates for shape changes of the latter (Nielsen and Pedersen Reference Nielsen and Pedersen1979).
Everted polypides of cyclostomates do not protrude from cystids farther than the lophophore base; the introvert is often not seen, and the mouth lies level with or below the aperture (in other classes, both the lophophore and introvert are revealed). Given these conditions, Taylor (Reference Taylor, Larwood and Nielsen1981) proposed that (1) clearance rate is lower in cyclostomates, because some of the cilia are hidden below the skeletal rim; and (2) cyclostomates have a narrow range of behaviors because introvert movements are nearly absent (the latter is partially supported by Shunatova and Ostrovsky Reference Shunatova and Ostrovsky2001).
Some parameters of zooid morphology appear to be nearly fixed within species, for example, the length of the cilia (data in Tamberg and Shunatova Reference Tamberg and Shunatova2017: Table 2); proportions and spacing of the tentacles, for example, distances between tentacle tips (Ryland Reference Ryland and Barnes1975; Herrera and Jackson Reference Herrera and Jackson1992); and mouth size (Herrera and Jackson Reference Herrera and Jackson1992). Others may vary more widely, for example, the size of the tentacle crown and tentacle length, number, and shape (Ryland Reference Ryland and Barnes1975, Reference Ryland1976; Tamberg and Shunatova Reference Tamberg and Shunatova2017), and the overall shape of the crown (Winston Reference Winston1978; McKinney Reference McKinney1990; Shunatova and Ostrovsky Reference Shunatova and Ostrovsky2001).
Several studies indicate that such soft-body parameters are also positively correlated with skeletal traits (McKinney and Boardman Reference McKinney, Boardman, Nielsen and Larwood1985; McKinney and Jackson Reference McKinney and Jackson1991), enabling theoretical reconstruction of the soft-body parts of fossil specimens. This in turn allows a range of paleoecological and paleofaunistic interpretations, from estimating the size range of available food particles to assessing the relationships between bryozoan orders through time. The existing formulas, however, cover very few zooid traits.
In Gymnolaemata the morphometry of the feeding apparatus is well understood, with many reported measures and established relationships (Winston Reference Winston, Woollacott and Zimmer1977, Reference Winston1978, Reference Winston and Broadhead1981; McKinney Reference McKinney1990). By contrast, Stenolaemata have not received nearly as much attention from researchers (Smith et al. Reference Smith, Taylor, Waeschenbach, Liow, Porter, Ostrovsky and Jenkins2017). It is a relatively large class with a rich fossil history (thousands of species from six extinct orders, dating back to the Ordovician) and about 543 living species in 98 genera in the extant order Cyclostomatida (Ryland Reference Ryland1970; Bock and Gordon Reference Bock and Gordon2013). Yet accounts of morphology of their soft parts and interrelationships with skeletal dimensions are extremely uncommon. Scattered morphometric data are given in various studies (Borg Reference Borg1926, Reference Borg1944; Ryland Reference Ryland1967, Reference Ryland and Barnes1975; Winston Reference Winston, Woollacott and Zimmer1977, Reference Winston1978, Reference Winston, Larwood and Abbott1979; Schäfer Reference Schäfer, Nielsen and Larwood1985; McKinney Reference McKinney and Bigey1991; Ryland and Hayward Reference Ryland and Hayward1991; Nielsen and Riisgård Reference Nielsen and Riisgård1998; Shunatova and Ostrovsky Reference Shunatova and Ostrovsky2001; Ramalho et al. Reference Ramalho, Muricy and Taylor2009). Rarely, however, do such reports combine cystid and polypide measurements. Equally rare is an indication of how many colonies were examined to reach the reported conclusions.
The aim of the present study is, therefore, to examine Recent cyclostomate polypide morphology and allometry, concentrating on the relationships between soft-part and skeletal morphology. We undertook this study on the southern New Zealand shelf (latitudes 45°S to 47°S), a place with abundant and diverse cyclostomates (Gordon et al. Reference Gordon, Taylor, Bigey and Gordon2009). We hope this will help paleontologists better interpret fossil stenolaemates.
Materials and Methods
We collected living colonies of cyclostomate species from around New Zealand, covering different taxonomic groups and a wide range of sizes and colonial forms. The majority of bryozoans were collected by dredge as part of a bimonthly sampling program in 2018 from a water depth of ~90 m on the Otago shelf of New Zealand (45°47.89′S, 170°54.5′E; Batson and Probert Reference Batson and Probert2000). Additional samples were taken in April 2018 from 58 m and 77 m around Stewart Island (46°54.87′S, 168°13.06′E and 47°07.70′S, 168°10.79′E, respectively).
Living bryozoan colonies were cultured in flow tanks in an isothermic room at ~13°C. Colonies were left to recover from dredging for at least 3 days, and kept alive for 1–4 weeks. Throughout this time, the colonies were constantly supplied with a mixture of natural particles and cultured algae (Rhodomonas salina and Dunaliella tertiolecta). Feeding animals with fully extended lophophores were imaged with a camera mounted on a dissecting microscope. From 9 to 54 photographs were taken of each specimen.
Measurements were made from the photographs using Inkscape 0.48.1 (Inkscape Project 2011). We recorded 10–12 parameters for each specimen, covering both skeletal and soft-body traits (Fig. 1). There were four skeletal characters: length and width of the aperture (AL and AW) and length and width of the zooid tube opening near the orifice (ZOL and ZTW). Among four tentacle crown parameters, two were quantitative: crown diameter (CD) and height visible above the aperture (CH); and two were qualitative: crown symmetry (Fig. 1C) and curvature (Fig. 1D). Finally, the three tentacle parameters were: tentacle number (TN), full tentacle length (TLF, sometimes simplified to tentacle length, TL, later in the article), and tentacle length visible above the aperture (TLP). For species with obliquely truncated tentacle crowns, we additionally recorded the length of the longest and shortest tentacles. We also measured mouth diameter (MD). The raw dataset is stored in full at PANGEA (PDI-21252). In this paper, the measured traits are reported with parametric descriptive statistics (mean ± SD, N) to ensure comparability with data from other authors. For the qualitative characteristics of the crown shape (symmetry and curvature), we give the dominant value and its frequency contrasted with the combined frequency of all other values.
The species we collected were divided into two groups. The first group contained abundant, easily identified, mostly large, and robustly calcified species (Table 1, Fig. 2B–D), examined in great detail and forming the main dataset of the study. In this group we recorded zooid traits in at least 10 colonies of each of the eight large species (a total of 83–209 zooids per species). An additional five species, which were smaller and less heavily calcified, formed the supportive dataset (Table 1, Fig. 2A); they were examined in lesser detail from fewer specimens (20–42 zooids per species). Assembled photographs of each colony yielded 5–25 measurements, but due to the varying angles of the photographs, not all characters were covered with equal detail.
Four of the species examined in this study, showing some measured traits: A, Idmidronea sp. B, Disporella pristis. C, Diaperoecia purpurascens. D, Favosipora rosea.
Cyclostomate bryozoan species used in this study, collected off Otago and Stewart Island, New Zealand.
Before proceeding, we checked whether these two datasets captured the natural variability of quantitative characters within each species equally well. To do this, we compared coefficients of variation of each character distribution using the Wilcoxon rank-sum test. As no differences were found (p > 0.26), we concluded that both datasets give a similarly detailed picture of zooid traits and treated them together in subsequent analyses.
To uncover relationships between skeletal and soft-body traits within our datasets, we used Pearson's correlations based on specimen measurement pairs, pooled across species. A correlation matrix of skeletal traits was made to find which traits could be used as independent predictors in subsequent regressions. Correlations between soft-body characteristics themselves were calculated to provide context for skeleton–polypide relationships. We proceeded to calculate univariate linear regressions between selected skeletal parameter and soft-body traits. Regressions were made for each species independently, as well as for all species together.
To establish relationships between zooid parameters in the wider context of all bryozoan classes, we created a hybrid dataset including measurements from published sources and our measurements from the full and supplementary datasets (means, medians, and SDs). Many cited sources were patchy in their coverage of the zooid parameters, which produced an inherently unbalanced dataset. Thus, to cover as many measurement pairs as possible, we performed five separate linear regression analyses with the zooid tube width and class as universal predictors and five soft-body traits (CD, CH, TL, TN, MD) as response variables.
Evaluating the resulting models, we learned that Cinctipora elegans exerted the strongest influence on regression parameters (Cook's distances between 1.8 and 5.5), invariably reducing the slope of the line. This is hardly surprising, given the trumpet-like shape of the zooids. Exclusion of this species from all regression analyses is justified from the mathematical point of view and also because it is very unusual among cyclostomates. While other tubuliporines, rectangulates, and cancellates in this study could be readily compared with known extinct species, C. elegans lacks an analogue in the Paleozoic fossil record.
To test the predictive abilities of the models, we measured photographs of 2–5 zooids from nine unidentified cyclostomate species from suborder Tubuliporina, collected in the same locations as the main material of this study. We made as many measurements as possible of the five soft-body traits (CD, CH, TN, TL, MD) and the zooid tube width. We put mean zooid widths (predictors) into the regression formulas and calculated predicted values of soft-body traits, which were later compared with actual measurements. To evaluate the predictions, we counted the number of testing values that fell into 95% predictive intervals of our modes and calculated the mean deviance between the predicted and testing values. Later, these measurements were included in the dataset. The resulting updated models are also reported here. We performed standard residuals checks on all regression models in this study: for normality (near-normality) of distribution, for patterns in residuals, and for influential observations.
Finally, we examined the levels of dissimilarity between bryozoan species from all three classes, taking into account several variables at once. To do this, we performed multivariate ordination in the form of nonparametric multidimensional scaling with Euclidean distances on a truncated hybrid dataset containing only complete measurement cases. To retain sufficient sample size, we had to restrict the list of variables to four: zooid tube width, crown diameter, and tentacle length and number. Further, whenever necessary, we complemented and/or averaged the cited measurements from different species of the same genus.
All statistical analyses were performed with R v. 3.4.4 (R Core Team 2018).
Results
Measured Zooid Parameters
All characters examined here showed strong variability both within and between species (Table 2.1 and 2.2). Aperture length was equal to or exceeded the length of zooid tube opening at the orifice level, especially strongly in case of Idmidronea sp., Cinctipora elegans, and Disporella pristis. Aperture width was similar to the zooid tube width in all species except C. elegans and Idmidronea sp.
Skeletal and tentacle crown characters measured in 13 cyclostomate species from New Zealand. All measurements are reported in micrometers (μm) as mean ± SD (sample size), while crown symmetry and curvature are given as the dominant value (its frequency/combined frequency of all other values). Coefficients of variation are reported for all measurements with sample size > 5. *See Fig. 1 for crown shapes. E, equitentacular; OT, obliquely truncated crown.
Characters of the tentacles and mouth. All measurements reported in μm as mean ± SD (sample size). Coefficients of variation are reported for all measurements with sample size > 5.
Characters of the tentacle crown tended to have somewhat greater dispersion than skeletal traits (Table 2). The smallest crown diameter was found in Hornera foliacea (287 ± 43.4 μm, 133), the largest in Diaperoecia purpurascens (584 ± 141.7 μm, 90), and there was variability of hundreds of micrometers within every species (Fig. 3A). Distribution of the crown heights was positively correlated with crown diameters (r = 0.74, p ≪ 0.001; Table 3), with larger species having taller tentacle crowns. Exceptions to this rule included Doliocoitis cyanea, Idmidronea sp., and H. foliacea, which demonstrated no correlation between these two traits (p > 0.5); the latter two species also had unusually flattened tentacle crowns.
Descriptive plots of selected polypide parameters: A, Tentacle crown diameter. B, Tentacle crown shape and symmetry. C, Tentacle number. D, Mouth diameter. On every plot the x-axis lists the studied species in alphabetical order: 1, Cinctipora elegans; 2, Diaperoecia purpurascens; 3, Diaperoecia sp.; 4, Disporella pristis; 5, Disporella sp.; 6, Doliocoitis cyanea; 7, Favosipora rosea; 8, Hornera foliacea; 9, Hornera robusta; 10, Idmidronea sp.; 11, Microeciella sp.; 12, Platonea sp.; 13, Telopora lobata. ET stands for equitentacular crown, OT, for obliquely truncated lophophore shape as depicted on Fig. 1.
Pearson's product-moment correlations between soft-body characters for 13 measured species combined (based on specimen measurement pairs). The p-values were corrected with the Holm method (Reference Holm1979).
The aspects of crown shape varied between polypides (even from the same colony) in all studied species (Fig. 3B). Obliquely truncated crowns dominated in all species except the two largest ones: C. elegans and D. purpurascens; but equitentacular crowns were also present in all species but one (Diaperoecia sp.). Regardless of the symmetry of the crown, however, tentacles were usually deeply curved into a bell or trumpet shape with splayed tips. Scoop-shaped crowns represented an interesting exception to this trend, being fairly common in Hornera robusta, H. foliacea, Platonea sp., and Telopora lobata, and even dominant in C. elegans.
Tentacle parameters, too, demonstrated considerable variability (Table 2.2). The tentacle number was unstable in 10 out of 13 species (Fig. 3C). Full tentacle length was either larger than or similar to the tentacle length revealed above the aperture. In the majority of species the differences were relatively small (<15% of the tentacle length), and the tentacle crown protruded almost fully from the zooid opening, with the mouth slightly under or level with the aperture rim (e.g., Fig. 2A,C). In four species (H. robusta, Diaperoecia sp., T. lobata, and D. purpurascens), the length difference was between 15% and 25%. Finally in Idmidronea sp. and C. elegans, 28% and 44% of the tentacle length, respectively, was hidden below the aperture, potentially removed from generating feeding currents.
Our studied species differ in mouth diameter from 17 μm (±0.6, 3) in Microeciella sp. to 48 μm (±3.0, 5) in C. elegans, but differences were also present within species (Fig. 3D). Small sample sizes reflect the difficulty of measurement; more data could resolve the variation better.
Relationships between Zooid Parameters in Our Dataset
The four skeletal parameters were strongly positively correlated with one another (Table 4), making it impossible to include all of them as independent predictors. Thus, we chose the width of the zooid tube near the orifice, rather than at the aperture, as a single measure that best approximates the internal diameter of the living chamber (after McKinney and Boardman Reference McKinney, Boardman, Nielsen and Larwood1985). The zooid chamber of most cyclostomates resembles a uniform, somewhat flattened cylinder that does not change shape near the orifice.
Pearson's product-moment correlations between skeletal traits for 13 measured species combined (based on specimen measurement pairs). The p-values were corrected with the Holm method (Reference Holm1979).
Measurements of the soft-body parts, that is, crown diameter and height, tentacle length and number, and mouth diameter, all demonstrated moderate to strong positive correlations with one another (0.67 to 0.80, all p-values ≪ 0.01; Table 3) with a single exception: we found no correlation between tentacle number and mouth diameter.
Correlations between the zooid tube width and soft-body traits were weaker, although still positive and statistically significant (0.48 to 0.78, all p-values ≪ 0.01; Table 3). For a more detailed analysis, we examined relationships between soft-body parameters and tube width separately for each species. The number of available observations was not always sufficient for linear modeling within each species, but when it was, we often found that tube width predicted parameters of the feeding apparatus poorly or not at all (adjusted R 2 between 0 and 0.37), with regression coefficients not statistically significant (Fig. 4).
Linear regressions showing relationships between dependent variables (tentacle crown diameter, height, mouth diameter, and tentacle length) with a single independent variable: zooid tube width. Regression lines are plotted both for individual species (colored) and for all species pooled (dotted black line with confidence interval shown as a gray band).
When species data were pooled, tube width became a somewhat better predictor of most soft-body traits (adjusted R 2 between 0.24 and 0.42, all p ≪ 0.01. Table 5). Regression lines for each character pair are given in Figure 4. Unlike other dependent variables, the relation between tube width and tentacle number in our dataset could not be modeled with a linear formula. Residuals from a regression test revealed violations of assumptions: severe divergence from normality and pronounced patterns in the residuals.
Details of regression analyses for zooid tube width (predictor) and soft-body characters (responses). Models are reported in the form: Intercept + Slope*Predictor. R 2 values are given as adjusted R 2/predicted R 2. ***p-value ≪ 0.001.
Relationships between Zooid Parameters in a Hybrid Dataset
We performed an analogous set of regression tests on a hybrid dataset that combined published data with our measurements (Figs. 5–7). Cinctipora elegans was removed from this dataset for reasons of biological (as having unusual zooid shape) as well as statistical nature (as an outlier with extreme leverage). Unlike the dataset containing only our measurements, tentacle number data from a combined dataset were successfully analyzed, and no violations of the assumptions were detected. We found that models including three predictors, that is, zooid tube width, class, and the interaction term, achieved the best fit (Table 5). This held true for all dependent variables except crown height, because there were too few available observations.
Because our formulas are intended to be used for predictions of fossil bryozoans, they may never be tested against the actual soft-body measurements. As the next best alternative, we tested the predictive power of the models by measuring photographs of nine unidentified cyclostomate species and comparing the test values with the values predicted by the models. In most cases (22 out of 27), the test values did not overstep the 95% predictive intervals of the regressions (insets in Figs. 5–7). The new measurements were adequately predicted by models describing the mouth diameter and tentacle number, but not crown parameters. The average of the deviations of real against predicted values was very close to zero (−0.2, 3) for tentacle number and mouth diameter, indicating a uniform spread of the data points around the regression line. For crown diameter, crown height, and tentacle length, however, the mean deviations differed from zero substantially (72, −38, and −22, respectively), indicating that test values appeared predominantly either above or below the regression line.
Linear regressions based on a hybrid dataset uniting our measurements and data from published sources. Presented variables include tentacle crown diameter and height plotted against zooid tube width. All regression lines are presented together with their 95% predictive intervals. Two insets represent correspondence between original models and models updated by the addition of the test values.
Linear regressions based on a hybrid dataset uniting our measurements and data from published sources. Variables include tentacle length and number plotted against zooid tube width. All regression lines are presented together with their 95% predictive intervals. Two insets represent stenolaemate-specific correspondence between original models and models updated by the addition of the test values. An asterisk in the upper inset represents an Ordovician trepostome Tetratoechus crassimuralis (Boardman and McKinney Reference Boardman and McKinney1976).
Linear regressions (mouth diameter vs. zooid tube width) based on a hybrid dataset uniting our measurements and data from published sources. All regression lines are presented together with their 95% predictive intervals. An inset represents correspondence between original models and models updated by the addition of the test values.
With the test values added into the hybrid dataset, we calculated updated regression models for all three bryozoan classes (Table 6A). We also had a sufficient number of observations to perform the regression analyses on the datasets containing only stenolaemates (Table 6B). Updated linear models, restricted to Stenolaemata, often demonstrated reduced fitness and lower predictive power compared with models based on all bryozoan classes (compare R 2 values in Table 6A,B), but they were free of nonlinear patterns and influential observations (Cook's distances < 0.8).
Details of a repeated regression analyses for zooid tube width (predictor) and soft-body parameters (responses) of bryozoan classes, based on a hybrid dataset updated to include the test values. Models are reported in the form Intercept + Slope*Predictor. ***p-value ≪ 0.001.
Finally, to visualize the amount of dissimilarity between bryozoan species using four variables at once, we performed a nonmetric multidimensional scaling (MDS). The resulting scaling had a low enough stress (0.113) to allow interpretation of the relative positions of data points on the plot (Fig. 8). Overall, the density of the data points indicate that bryozoans have relatively small range of proportions. The point cloud of the cyclostomates is smaller and overlaps considerably with that of gymnolaemates (Cheilostomata + Ctenostomata), indicating that the former group occupies only a part of the morphospace available to the latter.
Results of nonmetric MDS plotted in two main axes (stress = 0.122). Variables used in the analysis include zooid tube width, crown diameter, and tentacle length and number.
Discussion
This is the first detailed and coordinated report of morphometry in the cyclostomate Bryozoa. Among the cystid and polypide characteristics measured in this study, crown and tentacle traits are of the most interest.
Crown diameters, measured here, fall between 190 and 956 μm (individual measurements), with species means ranging from 287 to 584 μm. This agrees with previous studies (Ryland Reference Ryland and Barnes1975; Winston Reference Winston, Woollacott and Zimmer1977, Reference Winston1978, Reference Winston, Larwood and Abbott1979; Shunatova and Ostrovsky Reference Shunatova and Ostrovsky2001) and fits with the notion of the cyclostomates being generally smaller than gymnolaemates (60–1280 μm [species means]; data from Ryland Reference Ryland and Barnes1975, Reference Ryland1976; Winston Reference Winston, Woollacott and Zimmer1977, Reference Winston1978, Reference Winston, Larwood and Abbott1979; Buss Reference Buss1979; Herrera and Jackson Reference Herrera and Jackson1992; Shunatova and Ostrovsky Reference Shunatova and Ostrovsky2001).
In gymnolaemates and phylactolaemates, shorter tentacles tend to be straight, while long ones are usually curved: bent inward or outward or S-shaped (Ryland Reference Ryland1976; Sanderson et al. Reference Sanderson, Thorpe, Clarke, Cubilla and Jackson2000; Tamberg and Shunatova Reference Tamberg and Shunatova2017). By contrast, in cyclostomates, we did not see any relation between tentacle curvature and length. Cone-shaped crowns comprising straight tentacles were equally rare in species with short and long tentacles (Fig. 3B), and both within and across species, straight tentacles were not significantly shorter than curved ones.
Interestingly, in some cyclostomate species, an increase in tentacle length does not always correspond to an increase in crown height. For instance, H. foliacea and Idmidronea sp. have rather flattened, splayed tentacle crowns (regression slopes did not differ from zero). By contrast, in gymnolaemates, the proportions of the tentacle crowns are very conservative, even as their sizes vary (Ryland Reference Ryland and Barnes1975).
We used the formula TD = π*CD/TN to roughly estimate intertentacular tip distances (TD) from our measurements by reconstructing crown circumference from diameter (CD) and dividing it by the number of tentacles (TN). Admittedly, this approach does not take into account possible oval rather than circular shapes of the crown edge or uneven spacing of the tentacle tips. These imperfections, however, would lead to underestimated, rather than overestimated, intertentacular tip distances. Pooled, these distances in the studied Cyclostomatida varied from 55 to 240 μm (individual measurements), with medians ranging from 99 to 160 μm (13 species). It was not surprising to find such relatively large values, because many measurements come from obliquely truncated tentacle crowns with uneven tentacle lengths. We expected to see smaller intertentacular tip distances among strictly equitentacular crowns, and indeed the distances stabilized between 100 and 125 μm. This interval agrees with distances reported for Gymnolaemata, which hold tentacle tips about 110 (77–134) μm apart (Ryland Reference Ryland and Barnes1975). A distance of 140 μm and above is considered extreme for gymnolaemates (Riisgård and Manríquez Reference Riisgård and Manríquez1997; Sanderson et al. Reference Sanderson, Thorpe, Clarke, Cubilla and Jackson2000).
Among tentacle crown shapes, obliquely truncated crowns dominated in 10 of 13 examined species (except D. purpurascens, C. elegans, and Platonea sp.), while in the study by Winston (Reference Winston, Woollacott and Zimmer1978), 27 gymnolaemate species had equitentacular crowns and in 24 species at least some polypides were non-equitentacular. This suggests a much stronger presence of obliquely truncated crowns among cyclostomates, although the sample size is limited.
Winston (Reference Winston and Broadhead1981) and other authors (Lidgard Reference Lidgard, Larwood and Nielsen1981; Pratt Reference Pratt2004, Reference Pratt2008; von Dassow Reference Von Dassow2006) linked asymmetrical crown shapes with increased velocity of feeding currents, which is especially important in expelling already-filtered water away from the colony. Obliquely truncated lophophores often flank the excurrent chimneys in both gymnolaemates and cyclostomates (Cook Reference Cook1977; Winston Reference Winston and Broadhead1981; Shunatova and Ostrovsky Reference Shunatova and Ostrovsky2002), but in our study they seemed even more widespread, especially among species with few tentacles. The prevalence of nonsymmetrical crowns, together with larger intertentacle tip distances, may mean that tentacles are partly independent. But, as the number of tentacles increases, cyclostomates approach cheilostomates in the crown shape, intertentacle tip distance, and thus crown proportions. Is it possible that asymmetrically shaped crowns give specific advantage to smaller polypides?
Tentacle number varied considerably in some of the examined species, while remaining stable in others, both states having been found before in cyclostomates and gymnolaemates alike (e.g., Borg Reference Borg1926; Rogick Reference Rogick1949; Ryland Reference Ryland and Barnes1975; Winston Reference Winston, Woollacott and Zimmer1977; Schäfer Reference Schäfer, Nielsen and Larwood1985; Thorpe et al. Reference Thorpe, Clarke, Best, Nielsen and Larwood1985; McKinney Reference McKinney and Bigey1991; Ryland and Hayward Reference Ryland and Hayward1991). While changes in the interspecific tentacle number in Gymnolaemata were connected to ecological parameters, such as temperature (Amui-Vedel et al. Reference Amui-Vedel, Hayward and Porter2007), presence or absence of competition (Thorpe et al. Reference Thorpe, Clarke, Best, Nielsen and Larwood1985), and feeding conditions (Jebram Reference Jebram1973), little is known about the connection between tentacle number and ecological conditions in Cyclostomatida. It seems very likely, however, that such connections exist.
Сyclostomates are often reported as having the smallest feeding apparatus in the phylum, with tentacles ranging from 150 to 450 μm in length (Borg Reference Borg1926; Winston Reference Winston1978; McKinney Reference McKinney and Bigey1991). Our results show that in some species tentacles can be much larger: up to 700 (mean length within a species) or 900 μm (individual measurements). Because all measurements were made from extended polypides, we can conclude that in Cyclostomatida the polypides feed at very different sizes and start feeding while very small. Indeed, if a definitive “adult” zooid size exists, cyclostomates start feeding long before reaching it (Silén and Harmelin Reference Silén and Harmelin1974).
The ability to start protruding (and presumably feeding) while still developing and growing is also seen in phylactolaemate polypides (Y. Tamberg personal observations). Cyclostomates and freshwater bryozoans are thus in stark contrast with cheilostomates, which only begin feeding when fully grown (Marcus Reference Marcus, Grimpe and Wagler1926), because their polypides attain definitive size at the same time as the opercular apparatus becomes functional.
The full length of the tentacles in our measurements was equal to or somewhat larger than the length of the tentacle parts protruding above the aperture. Theoretical considerations point at free-walled species as potentially having smaller or no difference between full and revealed tentacle lengths, because they have a shared hypostegal cavity (Borg Reference Borg1926), which provides an additional reservoir for body cavity fluid and may increase protrusibility of tentacles (Taylor Reference Taylor, Larwood and Nielsen1981). Surprisingly, we found no evidence to support this view. There was no connection between zooid wall type and differences between full tentacle length and length above the aperture. Indeed, C. elegans, a species with mixed free- and fixed-walled zooids, had the most deeply set polypides. Our nonmetric MDS graph also did not reflect the grouping of the cyclostomates based on zooid wall type (Fig. 8).
Overall, within-species regressions based on our measurements show that characters of soft-body morphology are often independent from skeletal parameters (R 2 between 0 and 0.26). Although it seemed likely that increased level of detail would improve the predictive power of these models, nothing was gained through it. We found that species-specific morphometric traits are poorly connected across the skeleton–soft body divide and the natural intracolonial and intraspecific variability of polypide characteristics is very strong.
Analyses of the combined dataset with all classes incorporated were much more helpful in terms of explaining the values of the dependent variables. Compared with the models based only on our measurements, all models of the combined dataset demonstrated improved adjusted R 2 (i.e., increased amount of explained variance). In the cases of tentacle length and mouth diameter, the standard error of regression also improved (i.e., decreased), while for crown parameters it increased. Overall, however, the combined models are better at explaining variance.
The apparent advantage of the regression models based on a combined dataset, compared with solely our own measurements is mostly due to the considerable variability found within each species is collapsed into a single number—its central tendency, the mean. Thus, it is only means that these models explain and predict. The true variability remains hidden, and if there are smaller-scale relationships between soft-body and skeletal parameters acting within species, we failed to detect them.
Still, these models performed only adequately in predicting measurements (5 out of 27 test values fell outside the 95% predictive interval). Even worse, additional data points from the testing dataset altered the coefficients of the models. Undoubtedly, input of more species into the dataset would result in further changes. Therefore, we recommend caution while using our formulas for interpolations, and doubly so for extrapolations. Nevertheless, recognizing the need for predictions in the practical work of paleontologists and paleoecologists, we propose the following: for each value of independent variable (zooid tube width), report not only the predicted value of the response variable, but also the upper and lower boundaries of the appropriate 95% predictive interval (for an easy approximation of this interval, multiply the residual standard error S of the regression by 1.96).
In the past, several studies attempted to determine the nature of and generate formulas describing the relationship between the skeletal and polypide characters. Skeletal traits used in these studies as independent variables were aperture/orifice/living chamber width and distance to the nearest neighboring zooid. Among soft-body traits, tentacle number is the most commonly used, with crown and mouth diameter addressed as distant seconds (Winston Reference Winston and Broadhead1981; McKinney and Jackson Reference McKinney and Jackson1991).
Presence of a positive, moderately strong, and statistically significant relationship between polypide and skeletal traits was established in a number of studies (e.g., Winston Reference Winston, Woollacott and Zimmer1977, Reference Winston and Broadhead1981; Schäfer Reference Schäfer, Nielsen and Larwood1985; McKinney and Jackson Reference McKinney and Jackson1991) and confirmed in the present analysis. This in itself may be sufficient for some applications. Knowing that zooid tube width reliably correlates with polypide dimensions is sufficient to confidently compare skeletal remains from different bryozoan fossil communities, compare stenolaemate faunas (Reid and Tamberg Reference Reid, Tamberg and Zágoršek2019), and look for size distributions between stenolaemate orders.
In other situations it may be helpful to reconstruct the actual dimensions of the tentacle crown, because tentacle parameters (length and number) tie into food size and feeding parameters such as clearance rate (Strathmann Reference Strathmann1973) and particle speed (Best and Thorpe Reference Best and Thorpe1986). The latter can tell more about ecological roles and interactions between the members of benthic communities and, potentially, about the surrounding conditions. Thus, it may be important to choose the most suitable regression formula for specific reconstructions.
Existing regressions (e.g., in Winston Reference Winston and Broadhead1981; McKinney and Boardman Reference McKinney, Boardman, Nielsen and Larwood1985) and present data cover a similar zooid tube size range from ~60 to ~275 μm and include large living species (C. elegans and D. purpurascens measured by us, and Heteropora sp. from a hybrid dataset). The model coefficients themselves, however, differ notably both between studies and between our own analyses. We hesitate to single out any one formula as the best, because they all have their limitations, but we believe that ones that include the greatest number of species (i.e., updated combined dataset) are probably more suited for predicting soft-body parts from skeletal remains.
Necessarily, these predictions are additionally limited by the fact that only one order from the class Stenolaemata has survived to the present day. We may never know how deep were the differences in the soft-part morphology of the stenolaemate orders, and using one to predict the others carries inherent risk. However, some fossils retain what look like tentacles or tentacle-related structures that can be counted. For instance, 11 preserved tentacle remains were found in a zooid of the Ordovician trepostome Tetratoechus crassimuralis (Boardman and McKinney Reference Boardman and McKinney1976: Plate 13, Fig. 1). When plotted, it fell close to the regression line, supporting our model (inset in Fig. 6).
We also wish to point out the considerable interspecies variability within the Stenolaemata. In our nonmetric MDS plot (Fig. 8) cyclostomates occupied a substantial area in a multidimensional morphospace despite being outnumbered 10 to 1 by the flourishing modern gymnolaemates in the sea (543 vs. 5240 species; Bock and Gordon Reference Bock and Gordon2013) and 2 to 1 in the dataset (n = 29 vs. n = 52). The high intra- and interspecific variability of the modern stenolaemates revealed in this study necessitates proper respect for the underlying uncertainties of predictive modeling in this group.
Acknowledgments
We are most grateful to K. Currie (National Institute of Water and Atmospheric Research, Wellington, New Zealand) for generously sharing many sampling opportunities. We also wish to thank the staff at the Portobello Marine Laboratory (R. Pooley, L. Groenewegen) and the master and crew of RV Polaris II (B. Dickson, E. Kenton), University of Otago. Gratitude is due to C. Reid (Canterbury University) for discussion and to P. Batson (University of Otago) for providing valuable comments on the article. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
