INTRODUCTION
Chordates including vertebrates are close relatives of echinoderms and hemichordates. Hemichordates have been particularly important in investigations into the origin of vertebrates, since early studies suggested several morphological similarities (Bateson, Reference Bateson1885, Reference Bateson1886). Most of the proposed morphological similarities are no longer accepted as the synapomorphies between chordates and hemichordates, except for pharyngeal slits (Gerhart et al., Reference Gerhart, Lowe and Kirschner2005; Rychel, et al., Reference Rychel, Smith, Echimamoto and Swalla2006). Nonetheless, molecular phylogenetic studies, as well as recent investigations into the molecular developmental biology of enteropneust hemichordates (Lowe et al., Reference Lowe, Wu, Salic, Evas, Lander, Stange-Thomann, Gruber, Gerhart and Kirschner2003, Reference Lowe, Terasaki, Wu, Freeman, Runft, Kwan, Haigo, Aribiwicz, Lander, Gruber, Smith, Krischner and Gerhart2006), have strengthened their importance. Pterobranchs, one of two major groups of hemichordates, are filter feeders and are mostly colonial. They are initially recognized as bryozoans in the early years after their discovery (Allman, Reference Allman1869; Sars, Reference Sars1874; M'Intosh, Reference M'Intosh1887) until later detailed morphological analyses revealed close relationships between pterobranchs and enteropneusts. For example, they have a typical tripartite body consisting of an anterior coelom (protocoel) and two posterior coeloms (metacoels) with two mesocoels (collar coelom) in between (Harmer, Reference Harmer1887; Fowler, Reference Fowler1892, Reference Fowler1904). Pterobranchs are nowadays accepted as a group of hemichordates together with enteropneusts, which is also suggested by recent molecular phylogenetic analyses (Halanych, Reference Halanych1995; Winchell et al., Reference Winchell, Sullivan, Cameron, Swalla and Mallatt2002; Cameron et al., Reference Cameron, Garey and Swalla2000; Boulat et al., Reference Boulat, Nielsen, Lockyer, Littlewood and Telford2003).
Despite this phylogenetic significance, the reproduction and development of pterobranchs has been poorly studied, primarily because of the habitat of most species in the deep sea. Early studies of developmental biology were limited to cleavage stages, gastrulae and swimming larvae of Cephalodiscus, and depended on chance finds of ripe zooids and embryos among colonies dredged from cold, deep water during oceanographic expeditions around the turn of the last century (Lankester, Reference Lankester1884; Masterman, Reference Masterman1900; Anderson, Reference Anderson1903; Harmer, Reference Harmer1905; Schepotieff, Reference Schepotieff1909; Harmer & Ridewood, Reference Harmer and Ridewood1913; Gilchrist, Reference Gilchrist1915, Reference Gilchrist1917; Harmer, Reference Harmer1915). Those fragmented studies caused considerable disagreement. Swim-ming larvae were also described in Rhabdopleura (Burdon-Jones, Reference Burdon-Jones1957), but there has been no detailed description of these.
A breakthrough in reproductive and developmental biological studies of pterobranchs was brought about by the discovery of a population of Rhabdopleura compacta at approximately 20 m depth, off Stoke Point in Devon (Stebbing, Reference Stebbing1968). Plenty of live specimens became available from this relatively shallow site, which allowed detailed observation of some of the early developmental stages and larval behaviour (Dilly, Reference Dilly1973; Stebbing, Reference Stebbing1970a). This species has also been discovered from the east coast of Northern Ireland and from the north coast of Brittany (Stebbing, Reference Stebbing1970b). Subsequently, two other species of pterobranchs, Rhabdopleura normani and Cephalodiscus gracilis, have been discovered living at shallow depths (0.1–1 m and at low tide, respectively) around Bermuda (Barnes, Reference Barnes1977; Lester, Reference Lester1985). The development of Cephalodiscus is still largely unknown. On the other hand, some aspects of the development of R. normani have been studied in detail, particularly the development of internal organs (Lester, Reference Lester1988a, Reference Lesterb), by serial sectioning of various stages of larvae. However, the observations on the two species, R. compacta and R. normani, still showed some discrepancies, probably because of the lack of chronological observation of development. To study developmental biology further, an understanding of reproductive biology and life cycle is necessary. Lankester found that the reproductive season of Rhabdopleura normani in Lervik had almost ended in August, and he expected that the peak reproductive season was in July (Lankester, Reference Lankester1884). The first numerical study of seasonality in the Devon population of R. compacta involved counting the number of zooids, embryos, and dormant buds in November 1967, February 1968 and April 1968 (Stebbing, Reference Stebbing1970a). The author concluded that R. compacta was probably capable of successful sexual reproduction throughout the year (Stebbing, Reference Stebbing1970a). However, the data showed an increase in the relative proportion of active zooids and embryos from winter (November or February) to spring (April), suggesting a peak season of reproduction after April. To test this hypothesis, I examined reproductive status in large samples of R. compacta collected at regular intervals from November 2004 to February 2007. My study revealed that the production of embryos increased between April and June, but started to decrease between June and July.
MATERIALS AND METHODS
The animal
The pterobranch Rhabdopleura compacta Hincks is a tiny, colonial animal (Figure 1D), living in brown tubes (coenecium) attached to the concave side of a disarticulated lamellibranch shell (Figure 1A, B; Stebbing, Reference Stebbing1968). Mature females, recognized by a mature ovary almost as large as the rest of the body, incubate embryos and larvae in their coenecium until the larvae start swimming (Stebbing, Reference Stebbing1970a; Dilly, Reference Dilly1973). Zooids are sometimes found in a regressed dormant state, as differentiated buds in the basal region of coenecium (Stebbing, Reference Stebbing1970a).
Colonies, adult zooids and larvae of Rhabdopleura compacta. (A) Concave side of a lamellibranch shell (Glycymeris glycymeris). The black dots are the colonies of R. compacta. Scale bar: 10 mm; (B, C) a part of a colony, showing mature female zooid (fz), ova (ov), embryos (em) and larva (la). Scale bar: 100 μm; (D) ventral view of an adult zooid; (E) ventral side of a female zooid showing developing ovum in the posterior part of the trunk (metacoel). Note that female zooids are obvious because of the ova and less well developed tentacles (cf. zooid C). Tc, tentacle; cs, cephalic shield; s, stalk; ov, ova. Scale bar in C and D: 100 μm.
Disarticulated lamellibranch shells, mostly Glycymeris glycymeris (dog cockle), were collected at 21–24 m depth off Stoke Point (50°17′N 4°03′W) by dredging. In every collection, most of the shells were encrusted by one or two colonies of R. compacta.
Assessment of reproductive status
The colony form of R. compacta, consisting of crowded aggregations of tubular zooids often without clear mother–daughter budding relationships, sometimes makes it very hard to delimit colonies or to detect the co-occurrence of two (or more) colonies with their zooids intermingled. Therefore, reproductive activity was assessed at the level of all the colonies encrusting a particular shell.
To measure reproductive activity, I considered at least three factors: (a) relative frequency of shells with active colonies; (b) relative frequency of shells with mature zooids; and (c) relative frequency of embryos and larvae. Hence, shells were placed in four categories (Figure 2). Firstly, to examine the changes in active or dormant status, shells were categorized depending on whether they had colonies with zooids (active) or not (dormant). Active colonies were relaxed by adding an equal amount of magnesium chloride solution (0.37 M in water) to the natural seawater. To investigate the pattern of sexual maturity, colonies were then dissected to further categorize the shells according to whether colonies had any sexually mature zooids. In contrast to mature female zooids, mature male zooids are generally difficult to recognize when they are alive because of the body colour in their trunk, which obscures the internal organs. Therefore, whether the colonies on the shell were mature (MF) or non-mature (NMF) was determined by the presence of mature females (Figure 1E) which show developing oocytes in their metacoels. Further, to measure the degree of female reproductive activity, the numbers of embryos and larvae found in each shell were also counted. Data were statistically analysed using a program MiniTab.
Four categories of shell according to the reproductive status of Rhabdopleura compacta colonies on them. The numbers refer to the collection in April 2005 (shown in Table 5) as an example.
RESULTS
Stebbing's data revisited
A previously published numerical analysis of reproductive seasonality of the pterobranch species R. compacta was undertaken in November 1967, February 1968 and April 1968 on specimens collected off Stoke Point, Devon (Stebbing, Reference Stebbing1970a). My re-analysis of these data showed seasonal changes. For example, a Chi-square contingency test on the relative frequencies of active zooids to dormant zooids indicates that ratio of active to dormant zooids varies significantly between collections (Figure 3; χ2 = 12.923, df = 2, P = 0.002), being low in November (Figure 3). Chi-square analysis after excluding data from November indicates that there is no significant difference between the data from February to April (χ2 = 1.490, df = 1, P = 0.222). Therefore, these data imply that there was an increase in the relative frequency of active zooids from November 1967 to February and April 1968.
Observed frequency of active and dormant zooids from November 1967 to April 1968 (after Stebbing, Reference Stebbing1970a). Number in parentheses is the total number of tubes examined. Note that the relative frequency of dormant shells is high in November compared to other months.
In addition, Stebbing found more embryos in April than in November and February from almost the same number of zooids observed. This suggests that production of embryos per active zooid increased from November 1967 and February 1968 (Figure 4), although these data are not amenable to statistical analysis.
Observed frequency of embryos (after Stebbing, Reference Stebbing1970a). Number in parentheses is the total number of tubes examined.
Taken together, these results suggested the possibility that zooids became active in early spring, which is then followed by increased production of embryos from February to April. If this is a seasonal pattern, a peak season might be expected some time between April and November. To test this hypothesis, I undertook seasonal collections of R. compacta in 2004–2005 from the site studied by Stebbing (1970a).
Collections in 2004–2005 and implications of the change in status between spring and autumn
In the first year of investigation, shells were collected once in every season, i.e. November, April, July and September (Table 1). Colonies collected in November were mostly dormant and embryos and larvae were rare. However, as the analyses below clearly show, colonies became active and started reproduction in spring.
The number of colonies, zooids, embryos and larvae collected in each season during collections in November 2004–September 2005.
ND, not determined.
DORMANT VERSUS ACTIVE STATE
Comparison between the data from April, July and September 2005 showed that the proportion of dormant shells differs between the months (Figure 5) (Table 2; χ2 = 29.095, df = 2, P < 0.01) but no statistical difference was apparent in direct comparison between April and July (χ2 = 1.284, df = 1, P = 0.257). It is concluded that colonies were more active in April and July than in September. Given that Stebbing's data showed an increase in the relative frequency of active zooids from November to February, the data thus far suggest that colonies are dormant from autumn to winter and become active in spring and summer.
Frequencies of active and dormant shells in April, July and September 2005. The relative frequency of dormant shell is higher in September than in April and July.
Contribution of each frequency to the Chi-square count. Expected counts are printed below observed counts and the contribution to the Chi-square counts are printed in parentheses below expected counts.
RELATIVE FREQUENCY OF MATURE FEMALES
Another factor that would reflect the total extent of reproduction is the frequency of mature females. Is the frequency of mature females constant on active shells? To test this, the numbers of ‘mature shells’ (MF, those having mature females) and ‘non-mature shells’ (NMF, the remainder of the shells with active zooids), were compared between April, July and September (Figure 6). A contingency test showed that there was no significant difference in the frequencies of MF among the active shells (χ2 = 0.715, df = 2, P = 0.699). Thus, it is concluded that the occurrence of mature females has a peak coincident with the peak season of active shells.
Observed frequencies of mature (MF) and non-mature shells (NMF) amongst active shells in April, July and September 2005.
PROPORTION OF LARVAE TO THE TOTAL NUMBER OF EMBRYOS AND LARVAE
A previous study estimated that embryos are brooded for two to three weeks before the larvae are released from the metacoel and escape from the coenecium (Dilly, Reference Dilly1973). Because of this period of development, it is expected that at the end of a reproductive season, females would have more larvae, and fewer embryos. If there is a peak in reproductive activity, therefore, there would be a transition from embryos predominating to mostly larvae being present during the course of a season. To test this, I analysed the proportion of larvae to the total number of embryos and larvae in a shell between April and September (Figure 7; Table 3). Shells with embryos or larvae were categorized as follows: embryos only, larvae only or mixture. A Chi-square contingency test suggested that the relative frequency of the shells with ‘embryos only’ was high and the shells with ‘larvae only’ was low in April whereas the relative frequency of shells with ‘larvae only’ was high in September (Figure 7; χ2 = 13.394, df = 4, P < 0.01). After excluding the data from April, the data from July and September shared similar proportions (χ2 = 3.771, df = 2, P = 0.152). Therefore, a relative shift away from shells with embryos only occurred after April. This suggests that Rhabdopleura was at the beginning of reproduction in April therefore there were a greater proportion of shells with embryos only than in later months. Together with the implication that the reproductive status of zooids started decreasing by September, it is likely that there is a peak season in summer, some time between April and September.
Numbers of reproductive shells with embryos only, larvae only or a mixture in April, July and September 2005. Note that no shells had only larvae in April.
Numbers of shells with embryos only, larvae only, or a mixture. Expected counts are printed below observed counts and the contributions to the Chi-square counts are printed in parentheses below expected counts. Boxes indicate major contributions to the total Chi-square value.
Frequencies of each status of shell and the number of embryos and larvae in April 2006–February 2007.
Collections in 2006–2007 and peak season of reproductive activity during summer
To investigate the peak season of reproduction during summer, 150–250 shells were collected on four occasions in 2006 (April, June, July and November) and one in 2007 (February) (Table 4), and the reproductive status of the R. compacta colonies on them was analysed.
I first examined the changes in relative frequency of active to dormant shells and MF to NMF to confirm the result from 2005. A Chi-square contingency test indicated that the frequency of dormant shells varied significantly with month of collection (Table 5; χ2 = 22.372, df = 4, P < 0.01). Unexpectedly, however, the pattern did not suggest the dormancy in winter and active status in spring and summer. Instead, the relative proportion of active shells showed a decrease in June and little or no difference between the other collections (Figure 8; with June omitted after χ2 = 0.483, df = 3, P = 0.923).
Observed frequencies of active and dormant shells in the collections made in April 2006–February 2007.
Observed frequencies of active and dormant shells in the collections during 2006–2007 and their contributions to the Chi-square count. Expected counts are printed below observed counts and the contributions to the overall Chi-square values are printed in parentheses below expected counts. The collection which makes the largest contribution to the significant Chi-square value is marked by a rectangle.
Despite the lower relative frequency of active shells in June, the proportion of MF shells within active shells did not differ significantly between June and other months in 2006, as indicated by a Chi-square contingency test (Figure 9; χ2 = 3.307, df = 3, P = 0.347). This would support the hypothesis that the low relative frequency of active shells in June is most likely to be due to some abnormal conditions. If it had been because of some unknown factor that caused colony inactivation in June, it would be expected that the relative frequency of MF shells would also decrease in June. However, addition of the February 2007 data to the 2006 samples yields a significant Chi-square count (Figure 9; χ2 = 11.470, df = 4, P = 0.022), arising from the high frequency of MF shells in February.
Observed frequencies of mature (MF) shells and non-mature (NMF) shells from April 2006–February 2007.
SHELLS HAVING MATURE FEMALE WITH EMBRYOS OR LARVAE
If there is a peak season of reproductive activity, i.e. production of eggs, there would be a change in the relative frequency of females incubating embryos or larvae. Therefore, I examined the frequency of shells having mature females bearing embryos or larvae (reproductively active: RA; Figure 10). I found a significant difference in the relative frequency of RA shells (Figure 11, Table 6; χ2 = 29.863, df = 4, P < 0.01). Although the observed frequency of non-reproductively active (NRA) shells seems low in July, comparison of the data between only June, July and November and February without the data in April showed no significant difference between the months (χ2 = 6.425, df = 3, P = 0.093), confirming that the low frequency of RA shells in April is the major trend in the data. Thus, I concluded that the relative frequency of RA shells increased from April to June. This result agrees with the data in 2005 and suggests that Rhabdopleura is at the beginning of reproduction in April.
Nested classification of shells with active zooids.
Frequencies of reproductively active shells (RA) April 2006–February 2007.
Chi-square test on the relative frequency of reproductively active (RA) shells. Expected counts are printed below observed counts and the contributions to the Chi-square counts are printed in parentheses below expected counts. The collection that makes the greatest contribution to the significant Chi-square value is marked by a rectangle.
PROPORTION OF LARVAE TO THE TOTAL NUMBER OF EMBRYOS AND LARVAE
By categorizing the shells into ‘embryos only’, ‘larvae only’, and ‘mixture’, the data from 2005 suggested that the peak season of production of eggs was some time from April to September. In order to investigate the trend in this reproductive index during summer, similar analyses were performed in 2006–2007 (Figure 12; Table 7). A Chi-square contingency test indicated significant differences between months (Table 7; χ2 = 17.733, df = 8, P < 0.05); the frequency of the shells with embryos only was high in June and low in July. Thus, it is suggested that females were actively producing eggs in June, and this reproductive activity decreased by July.
Frequencies of shells with embryos only, larvae only, and a mixture April 2006–February 2007.
Numbers of shells with embryos only, larvae only, or a mixture. Expected counts are printed below observed counts and the contribution to the Chi-square counts are printed in parentheses below expected counts. Boxes indicate major contributions to the total Chi-square value.
Did this observed change reflect changes in the reproductive activity of individual females? To infer the activity of individual females from data scored per shell, it is necessary to confirm that there is no significant difference in the data sets counted by shells or counted by zooids. Therefore, the contributions of embryos and larvae in each tube were scored in July 2006 and the results were compared with those from scoring shells (Figure 13). A Chi-square contingency test indicated that there is no significant difference between the two data sets (χ2 = 1.451, df = 2, P = 0.484), suggesting that data counted by shell reflects the activity of individual females. Thus, it is concluded that embryo production of a female has its peak between June and July.
Composition of broods obtained by scoring embryos and larvae per shell, versus per tube (zooid), based on the data in July 2006.
DISCUSSION
The present study showed that the relative frequency of shells with mature female zooids is most likely constant throughout the year. Although this frequency showed a significant decrease in June 2006, this may have been due to some abnormal environmental conditions. One indication that sea conditions were unusual at this time is the fact that ‘needle plankton’ (thin diatoms), which are usually abundant only in autumn, appeared temporarily in June 2006, disappearing by July 2006 (G.T. Boalch personal communication). Therefore, I suggest that relative frequency of the shells with mature females is constant throughout the year. Despite this constancy, I found that production of eggs has its peak between June and July. My suggestion of a summer peak of reproductive activity revises the previous speculation that R. compacta is capable of successful sexual reproduction throughout the year at Stoke Point (Stebbing, Reference Stebbing1970a). Instead, my results also agree with the finding of Lankester (Reference Lankester1884) that most male gonads are probably spent by August in R. normani.
Seasonality of reproduction is observed in many shallow-water temperate invertebrates; its occurrence in benthic suspension feeders is reviewed by Coma et al. (Reference Coma, Ribes, Gili and Zabala2000). Eggleston (Reference Eggleston1972) provides data on seasonality in a temperate assemblage of one such group, the Bryozoa, a phylum of colonial suspension feeders including species that appear ecologically very similar to R. compacta. A potential generalization arising from these studies is that species close to the southern limit of their geographical range often have an inactive period during summer and a reproductive peak in the cooler months, while those near their northern limit frequently show winter dormancy, with peak sexual reproduction in the warmer months. Rhabdopleura compacta appears to follow the latter cycle. A similar broad pattern has been noted in British phoronids (Emig, Reference Emig1979). British ascidians generally have a longer period of breeding, but reproduction is still generally restricted to the warmer season (Millar, Reference Millar1970).
What triggers changes in the reproductive activity of R. compacta? Several environmental factors such as temperature, day length, food and salinity, have been investigated in relation to the reproductive activity of marine invertebrates. Correlations have been described and experimentally tested in laboratory conditions (e.g. temperature reviewed, Giase & Pearse, Reference Giese and Pearse1974; day length, Boolootian, Reference Boolootian1963; food, Crisp & Spencer, 1958; Starr et al., Reference Starr, Himmelman and Therrianlt1990). However, these factors are related to each other in nature, and organisms react to their total environment rather than to a single factor (Coma et al., Reference Coma, Ribes, Gili and Zabala2000). It is not possible to conclude from the present study what triggers the reproductive activity in R. compacta. However, there is a correlation between changes in these factors and the reproductive seasonality of R. compacta. For example, the temperature of seawater off the coast of Devon, at about 25 m depth, about the depth at which R. compacta was collected, is approximately constant from January to May (below 10) and starts increasing in May (2005 data from the Western Channel Observatory: http://www.npm.ac.uk/rsg/projects/observatory/l4_ctdf/). In relation to food availability, studies have shown that pterobranchs feed on phytoplankton and zooplankton. Previous studies found occasional algal cells, diatoms and a copepod in the stomach of R. compacta (M'Intosh, Reference M'Intosh1887; Stebbing & Dilly, Reference Stebbing and Dilly1972), while in R. normani, diatoms, radiolarians and crustacean larvae were found (Schepotieff, Reference Schepotieff1907). The amount of phytoplankton is known to respond to the increased light and temperature of the spring months and increases rapidly (Wickstead, Reference Wickstead1976), followed by an increase of zooplankton. Weekly observation of plankton at Station L4 in Plymouth showed that, although it is not easy to generalize, a dense population of zooplankton appeared in May 2003 and 2004 (Boalch, Reference Boalch2003, Reference Boalch2004). Testing the causal relations between these and the reproductive activity in R. compacta will provide us further knowledge into the general biology of pterobranch hemichordates.
ACKNOWLEDGEMENTS
I am extremely grateful to John Bishop, Christine Wood, David Dixon, Linda Dixon, Georgina Budd, Jefferson Murua, Patricia Masterson, and Stephen Hawkins for access to laboratory and boat facilities at the Marine Biological Association, Plymouth and for generous help and hospitality while collecting animals. John Bishop, Peter Holland and Robin McCleery gave much helpful advice during this research and critical comments on the manuscript. This study was funded by a research grant from Research Institute of Marine Invertebrates, Japan, by an Ishizaka Scholarship from Nippon Keidanren, Japan, by an Oxford–Kobe Scholarship from St Catherine's College (University of Oxford) Kobe Institute and by Merton College, Oxford.
