Patterns of growth in cold-seep vestimenferans including Seepiophila jonesi: a second species of long-lived tubeworm


Erik E. Cordes, Organismic and Evolutionary Biology Department, Harvard University, 16 Divinity Ave, Cambridge, MA 02138, USA.


Seepiophila jonesi is a vestimentiferan tubeworm (Siboglinidae: Polychaeta) inhabiting the cold seeps of the upper slope of the Gulf of Mexico. It commonly co-occurs with Lamellibrachia luymesi, which is among the most long-lived non-clonal animals known. The growth pattern of S. jonesi is best described by a model including a size-specific probability of growth and an average growth rate that does not vary with individual size. This model, based on growth data from in situ staining and collection approximately 1 year later, predicts that S. jonesi is very slow growing and may attain ages comparable with L. luymesi. The efficacy of this model in describing L. luymesi growth rate was assessed, but the previously employed model of declining growth rate with individual size provided the better fit to the empirical data. Comparisons of both S. jonesi and L. luymesi growth rates among sites and among aggregations within a site indicate that there is some degree of habitat variability contributing to differences in growth rates. However, position of the anterior end of the worm within an aggregation did not have a significant effect on growth rate in comparisons among groups of L. luymesi from different distances from the center of an aggregation. The evolution of longevity in these species of vestimentiferans was favored by the relative stability of the seep habitat and sulfide sources, in contrast to the hydrothermal vent environment inhabited by relatively short-lived and fast-growing vestimentiferan species.


In animals, longevity is more likely to evolve if hazards in the environment are low (Williams 1957). Organisms that inhabit stable environments where resources are not limiting and lethal predation rates are low should have long life spans. The vestimentiferan tubeworm Lamellibrachia luymesi inhabits the hydrocarbon seeps of the northern Gulf of Mexico and is among the most long-lived non-clonal animals (Fisher et al. 1997; Bergquist et al. 2000, but see Schöne et al. 2005). Empirically determined growth rates of L. luymesi from the upper Louisiana slope in the Gulf of Mexico indicate that large individuals with over 2 m anterior tube lengths are in excess of 200 years of age (Fisher et al. 1997; Bergquist et al. 2000; Cordes et al. 2005a).

Vestimentiferan tubeworms are polychaetes of the Family Siboglinidae (Rouse 2001). They lose their digestive tract upon metamorphosis from the larval form and are reliant on sulfide-oxidizing bacterial symbionts for their nutrition as adults (Fisher 1990). The symbionts are intracellular and contained within bacteriocytes in a large internal organ, the trophosome (Jones 1981; Cavanaugh et al. 1981). The tubeworms transfer both sulfide and oxygen to the symbionts via specialized hemoglobin molecules contained in their vascular blood and celomic fluid (Arp et al. 1987; Childress et al. 1991). The symbionts fix carbon and nutrients are transferred to the host tubeworm through a combination of translocation of metabolites and intracellular digestion of the symbionts (Bosch & Grasse 1984; Felbeck & Jarchow 1998; Bright et al. 2000).

Unlike the more well-known vent species that rely on sulfide uptake from the water around their anterior plume (Childress & Fisher 1992), it has been demonstrated that the cold-seep tubeworm species L. luymesi is capable of using a posterior extension of its body and tube, the root, for uptake of sulfide from seep sediments (Julian et al. 1999; Freytag et al. 2001). Models of L. luymesi sulfide sources suggest that they release the sulfate generated by their sulfide-oxidizing symbionts into the sediments around their roots (Cordes et al. 2005a). The models predict sulfide production from sulfate at sufficient rates to match the high sulfide demand of L. luymesi aggregations (Cordes et al. 2003) in the presence of consistent sources of methane and higher hydrocarbons (oil) as electron donors (Cordes et al. 2005a). Seepage of oil and gas is stable in these habitats with estimates for the persistence of an individual seep site in excess of 10,000 years (Roberts & Aharon 1994). This ensures the persistence of the tubeworms’ ultimate energy source, preventing nutrient limitation and favoring the evolution of longevity in these organisms.

In addition to nutrient availability, mortality probability is related to the frequency of lethal predation. Unlike some vent vestimentiferans (Urcuyo et al. 2003), there are no visible signs of plume predation on the seep species in the Gulf of Mexico. Furthermore, analysis of tissue stable C, N, and S isotopes have thus far found no evidence of significant nutritional input from vestimentiferan tissue in the diet of any of the fauna found around the seeps in the Gulf of Mexico (MacAvoy et al. 2002, 2005). The lack of lethal predation on vestimentiferan tubeworms at the Gulf of Mexico seeps would also favor the evolution of longevity in these species.

Commonly co-occurring with L. luymesi (Bergquist et al. 2002; Cordes et al. 2005b), and occasionally dominating tubeworm aggregations (Cordes et al. 2006), is another species of vestimentiferan tubeworm, Seepiophila jonesi. There is little known about the habitat preferences, physiology, or growth rates of this species including whether it relies on its plume or roots for sulfide uptake. The paucity of growth data on this species is a result of its growth habit. Because this species grows with its anterior end close to the sediment–seawater interface (Fig. 1), the frequency of staining of this species is much lower than L. luymesi. Use of a new whole-aggregation staining device in 2002 and 2003 greatly increased the frequency of staining for S. jonesi. In this study, empirically determined growth rates and a growth model for S. jonesi are presented to examine its growth pattern and test the hypothesis that the stability of the seep environment would favor longevity in this species as well. We also examine the variability of growth rate in both S. jonesi and L. luymesi among three sites on the upper continental slope off the coast of Louisiana in the Gulf of Mexico.

Figure 1.

 Photograph of a tubeworm aggregation 14 months following staining in situ. This aggregation was stained with the whole-bush stainer (modified Bushmaster device) in 2003 and collected in 2004 from GC 234. The majority of the individuals are Lamellibrachia luymesi. Seepiophila jonesi (with large flares at the ends of the tubes) can be seen primarily around the periphery of the aggregation near the sediment surface.

Material and Methods

Growth data were obtained by staining tubeworms with a chitin stain (acid blue no. 148). Tubeworms were stained at three different seep sites within a 5-km radius of each other on the Upper Louisiana slope of the Gulf of Mexico: Green Canyon (GC) 234 (27°44.8′ N, 91°13.3′ W, 530 m depth), GC 232 (27°44.5′ N, 91°19.1′ W, 570 m depth) and Bush Hill (GC 185, 27°45.0′ N, 91°30.5′ W, 540 m depth). These are some of the most extensively studied seep sites in the Gulf of Mexico and complete geological, geochemical, and ecological descriptions may be found elsewhere (e.g.MacDonald et al. 1989, 1990; Aharon et al. 1997; Bergquist et al. 2003a; Sager et al. 2003; Cordes et al. 2006). In 1994, 1995, and 1997 portions of 15 tubeworm aggregations were stained using a small dome-stainer device at GC 234 and Bush Hill (Bergquist et al. 2000, 2002). In addition, six whole aggregations were stained using a modification of the Bushmaster Jr collection device (Urcuyo et al. 2003) fitted with an internal plastic liner. Four aggregations were stained at GC 234 and GC 232 in June 2002 and collected in August 2003, and two additional aggregations were stained at GC 234 in August 2003 and collected in June 2004. The Bushmaster device was constricted around an aggregation and stain was intermittently pumped into the device for approximately 5 min. After staining, the tubeworms appeared healthy with their plumes extended from their blue-stained tubes, and they appeared healthy prior to collection the following year (Fig. 1). The Seepiophila jonesi and Lamellibrachia luymesi growth data from the 1994, 1995, and 1997 stainings were presented previously (Bergquist et al. 2000, 2002) as were the L. luymesi data from the 2002 stainings (Cordes et al. 2003, 2005a). Seepiophila jonesi growth data from 2002 (n = 150) and all data from the 2003 stainings (n = 176) are presented here as well.

Two different approaches to modeling tubeworm growth were evaluated with respect to their efficacy in describing the observed growth patterns of S. jonesi and L. luymesi. The first model included size-specific growth rate based on nonlinear regression of anterior length to annual growth. The nonlinear regression model was as follows:


where g is growth rate in cm · year−1 and L is tube length in cm (Bergquist et al. 2000). In the second model, growth rate was held constant at the average of all non-zero growth rates measured for that species and the probability of positive (non-zero) growth was modeled as a function of individual size using weighted nonlinear regression. The function:


was used to describe the size-specific probability of positive growth (pg) for 5-cm size classes of individuals (L5). Model fits to the empirical data were judged based on the significance of the regression and the proportion of variance explained (r2). The best fitting model for each species was selected for use in the subsequent simulations.

Age estimates for tubeworms were obtained by simulating growth for 1000 individuals of each species using the two different models. As integration of the second model to arrive at a deterministic solution for size at age for the tubeworms is not possible, individual-based simulations of growth were adapted from the methods of Cordes et al. (2003, 2005a). In the simulations, the size of an individual is tracked over time and size-specific growth rate (model 1) or growth probability (model 2) determined at each annual time step. Size-specific growth rate was allowed to vary from an average rate as a normally distributed parameter with a standard deviation approximating the residuals from equation (1) with growth rates bounded by zero, such that g ≥ 0. The size-specific standard deviation in this error term was selected based on the best model fit to the absolute value of the residuals (r) where:


This function was used as the error term instead of the confidence intervals of the nonlinear regression in order to more completely approximate the observed variability in growth rate. In the second model, a random number with a uniform distribution between 0 and 1 is compared with the size-specific probability of growth. If the probability of growth exceeds the random number, then the individual in the simulation grows at a rate determined from a normal distribution with an average and standard deviation determined from all measured non-zero growth rates. Average size at age and maxima and minima of 1000 iterations are reported. Nonlinear regressions were performed using JMP software (SAS Institute) and all simulations were run in Delphi (Borland).

Differences in growth rates among sites and aggregations were assessed using analysis of variance (ANOVA) of the residuals of the nonlinear regression with Tukey's post hoc test for pairwise comparisons. This analysis tests the hypothesis that some aggregations comprised individuals exhibiting growth rates consistently above or below the size-specific average growth rate determined by the regression model in equation (1). The first model included residuals from the regression including all individuals measured with site as the factor. The second model compared each aggregation's residuals from the site-specific growth model. All analyses were performed for each species independently. Because there were only two aggregations sampled at GC 232, t-tests were used for these comparisons.

Variability of L. luymesi growth rate within an aggregation was assessed in another ANOVA to determine if growth rate varied with the position of the anterior end of a tubeworm in an aggregation. After collection of one of the stained aggregations, a wooden dowel was placed in the center of the aggregation. A line was attached to the dowel and the anterior ends of all stained individuals within the radius described by the line were marked with a certain color. Individuals were marked at 5-, 10-, and 15-cm intervals, resulting in four treatment levels: <5, 5–10, 10–15, and >15 cm from the center of the aggregation.

Results and Discussion

Seepiophila jonesi exhibited extremely slow growth rates (Fig. 2A). The majority of animals collected (59.6%) showed no detectable growth over the preceding year, although they appeared healthy at the time of collection. Growth rate declined with individual size in S. jonesi (P < 0.001, r2 = 0.07), although the proportion of variance explained by the function was low (Fig. 2A). This is in contrast to previous findings of no significant decline in growth rate with size (Bergquist et al. 2002). If the zero growth rate measurements are removed from the analysis, then there is no significant trend in growth rate with individual size (P = 0.493, r2 = 0.01), and the average and standard deviation of growth rate is 2.14 ± 1.55 cm · year−1. Larger individuals grew less frequently than smaller individuals (Fig. 3A), with a significant decline in the frequency of positive growth with individual size (P < 0.001, r2 = 0.90). Thus, the probability of growing in a given year declines with age, but if an animal grows the amount of yearly growth does not change with age.

Figure 2.

 Annual growth rate for (A) Seepiophila jonesi and (B) Lamellibrachia luymesi measured at three different sites in the Gulf of Mexico. Growth rates determined from in situ staining of tubeworm individuals and collection 12–14 months later. Points include all individuals measured in this study and data from previous studies. Lines represent growth model 1: the nonlinear regressions of equation (1) and 95% confidence intervals. Growth rates from GC 232 (plus sign) were significantly different from GC 234 (closed circles) and Bush Hill (open circles).

Figure 3.

 Frequency of positive growth for (A) Seepiophila jonesi and (b) Lamellibrachia luymesi individuals. Nonlinear regression lines indicate declining frequency of positive growth with individual size (model 2). Individuals are grouped into 5-cm size classes with center of size class shown.

Lamellibrachia luymesi growth was best described by the model of declining growth rates with individual size (Fig. 2B). This model fit the empirical data well (P < 10−134, r2 = 0.27), and remained significant when zero growth rates were removed from the data set (P < 10−75, r2 = 0.20). There was also a significant decline in the probability of positive growth with individual size in L. luymesi (P < 10−5, r2 = 0.40), but the variance around the regression line was relatively high for large individuals (Fig. 3B). Because the fit of the nonlinear growth model (model 1) to the L. luymesi data was comparable with the fit of the probability of growth model (model 2) and non-zero growth rates still declined with length, model 1 was used in all subsequent simulations of L. luymesi growth. This is the same function used in previous studies (Bergquist et al. 2000, Cordes et al. 2003, 2005a) and the parameter estimates continue to be refined with the inclusion of the new data here.

The method of modeling S. jonesi growth as a declining frequency of positive growth rather than declining growth rate with individual size is based on the normal habit of S. jonesi in situ. This species is typically found with its anterior end very near the sediment surface with the plume exposed to low levels of sulfide in the water column. The sulfide-binding affinity of one of the vascular hemoglobin molecules suggests that S. jonesi is poised to acquire sulfide at low environmental concentrations across its plume (Freytag 2003). Therefore, it would be to its advantage to maintain their plume as near to the sediment surface as possible because sulfide is often detected at the sediment–water interface, but concentrations rapidly decline with distance from the sediment surface (Bergquist et al. 2003b; Cordes et al. 2005b). Under typical conditions they do not grow much above the sediment surface (but could potentially grow at the posterior end, which has not been measured). However our data suggest that when they do grow, perhaps to avoid burial of their anterior end and to assure the supply of oxygen at their plume, they add approximately 2 cm to the anterior end in a single growth spurt. In a previous study of S. jonesi growth rate (Fisher et al. 1997), the median growth rate was zero (as it was in this study), and the average distance between growth ‘rings’ (successive flared portions of the anterior ends of the tubes) was 1.14 cm. This suggests that, on average, two of these rings are added during a year in which growth occurs. This strategy would allow S. jonesi to avoid competing for sulfide with L. luymesi by utilizing the sulfide in the benthic boundary layer rather than from deep in the sediments. If L. luymesi depletes sulfide in the deeper sediments (Cordes et al. 2003) and S. jonesi is scavenging the remaining sulfide along with that produced in surficial sediments and released at the sediment–water interface, environmental sulfide concentrations would be further reduced in older tubeworm aggregations, as has been previously demonstrated (Bergquist et al. 2003b; Cordes et al. 2005b; Cordes et al. 2006).

This model of S. jonesi growth provides age estimates of the same magnitude as L. luymesi. When S. jonesi growth is modeled as a declining frequency of positive growth (Fig. 4), an average age of 96 years (min = 38, max = 251) is predicted for a 50-cm individual. The largest S. jonesi that has been collected was over 1.1 m in length and S. jonesi individuals of similar sizes are occasionally observed in the middle of extremely large aggregations that are too large to be collected with existing sampling gear. Individuals over 1 m in length are predicted to be in excess of 300 years of age, based on an extrapolation of the model. However, it is possible that these individuals all experienced anomalously high growth rates beyond those measured in this study. These estimates could also be biased, as the error around the growth probability function increases with increasing distance from the average length of S. jonesi individuals for which growth measurements exist. When modeled with this large error term included, the probability of growth increases with increasing length. This is because of increasingly positive deviations as the error term is bounded by zero. Although this is contrary to existing data and observations, this alternative hypothesis of increasing growth rate cannot be rejected at this time. Until additional data on growth rates and growth probability for large S. jonesi are available (the largest in this study was 41.3 cm, but larger individuals are only occasionally observed and collected), these estimates of S. jonesi longevity should be considered preliminary. Regardless, modeled growth rates for S. jonesi follow a similar but slower trajectory than L. luymesi, suggesting that S. jonesi individuals attain ages at least comparable to, and potentially exceeding those of L. luymesi.

Figure 4.

 Age estimates for Seepiophila jonesi and Lamellibrachia luymesi. Average, maximum, and minimum size at age based on 1000 iterations of the two growth models are presented. Seepiophila jonesi growth was modeled using model 2 (size-specific growth probability) and L. luymesi growth modeled using model 1 (size-specific growth rate).

Growth rates of both species differed among the three study sites (S. jonesi, P < 0.001, r2 = 0.09; L. luymesi, P < 0.001, r2 = 0.16). In S. jonesi, growth rates at GC 232 were significantly different from GC 234, and in L. luymesi all sites were significantly different (Fig. 5). Within a site, growth rate varied among some of the aggregations (Fig. 6). Differences in growth rates among aggregations were prominent for both species at GC 234, and for L. luymesi at Bush Hill. These differences may be due to local environmental factors influencing the distribution and quantity of sulfide in the habitat. Alternatively, differences in growth rate could be due to local dispersal and population-level genetic effects. However, L. luymesi populations do not show genetic segregation at the aggregation or site level at the spatial scales examined here (McMullin 2003). Therefore genetic differences are likely to be less of a factor than nutrient availability. If growth rates are dependant on sulfide availability and sulfide concentrations and fluxes are increased in microhabitats supporting large, dense aggregations, these individuals could be growing faster than has been measured to this point. Further experimental manipulations of sulfide flux and examinations of the plasticity of growth rate are required to assess the degree of dependence of growth rate on sulfide availability. There were no significant differences in growth rate detected among S. jonesi in different aggregations from Bush Hill, or in either species from two aggregations at GC 232. Either the Bush Hill and GC 232 sites are more geochemically homogenous than GC 234, or the lack of sufficient replication within these sites resulted in the lack of significant differences detected.

Figure 5.

 Site-specific growth rates for (A) Seepiophila jonesi and (B) Lamellibrachia luymesi. Residuals of growth rates represent the average residual of all individuals within a site from the nonlinear regression of size-specific growth rate using all of the data. Average and standard deviation of residuals are shown.

Figure 6.

 Aggregation-specific growth rates for (A) Seepiophila jonesi and (B) Lamellibrachia luymesi. Residuals of aggregation-specific growth rates represent individual differences from the site-specific length-to-growth rate relationships. Average and standard deviation of residuals are shown. Horizontal bars above the points represent groups of aggregations within which there are no significant differences among relative growth rates.

There were no significant differences detected in L. luymesi growth rates in groups of individuals from different distances from the center of an aggregation (P = 0.738, r2 = 0.01, Fig. 7). Similar growth rates from a common point of attachment would lead to the hemispherical shape normally observed in vestimentiferan aggregations on the upper Louisiana Slope. The lack of a relationship between growth rate and the position of the anterior end of the tube and plume indicates that the environment is relatively homogenous above the sediment surface. It remains possible that there were subtle differences between different sides of the aggregation, as this factor was not taken into account. It is also possible that differences in microhabitat are only evident in the reproductive output of an individual rather than size-specific growth rate. The most likely explanation is that growth rate is determined to a certain degree by the position of the posterior end of the tube, the site of sulfide uptake in L. luymesi (Freytag et al. 2001). The high degree of spatial heterogeneity in seep sediments (Treude et al. 2003; Arvidson et al. 2004) suggests that the position and differential growth of the posterior end may be particularly significant.

Figure 7.

 Variability in Lamellibrachia luymesi growth rate at different positions within an aggregation. The distance from the center of the aggregation to the anterior ends of stained individuals was noted immediately after collection. There were no significant differences in the growth rates of individuals among the distance treatments shown.

Both L. luymesi and S. jonesi exhibit very slow growth rates and long life. Because they are thought to be continuously reproductive (Tyler & Young 1999), these species do not appear to abide by the long-standing theory of a trade-off between longevity and fecundity (Williams 1966; but also see Reznick et al. 2000). Rather, the life-history strategy in these species appears to be one of high reproductive effort over a long life-span to offset high losses in the dispersal and recruitment stages. If the probability of complete reproductive failure during any single reproductive event is very high, then multiple reproductive events are favored (Stearns 1992). The recruitment period for a new aggregation has been estimated to be between 20 and 60 years, with the majority of recruitment occurring early in this period (Bergquist et al. 2002; Cordes et al. 2003). Although this period is fairly long, the frequency of appearance of a new suitable seepage site at the seafloor, including both the presence of hydrogen sulfide and a hard substrate for settlement (Bergquist et al. 2003a), is likely to be rare, although this geologic process is poorly constrained. The low frequency of generation of primary substrate and the very patchy distribution of these microhabitats on the continental slope of the Gulf of Mexico (Kennicutt et al. 1988; Sager et al. 2003) make the probability of successful recruitment to an appropriate habitat extremely low for vestimentiferan larvae. The low probability of reproductive success has likely favored the evolution of high lifetime fecundity through continuous reproduction and extreme longevity in S. jonesi and L. luymesi.

The hydrothermal-vent tubeworm Riftia pachyptila is subjected to very different selective pressures. In contrast to the stable seep environment, hydrothermal vents are more ephemeral habitats, with microhabitats that can support vestimentiferans appearing and disappearing on the order of years to decades rather than centuries or more (Shank et al. 1998; Tsurumi & Tunnicliffe 2001). Higher adult mortality rates, in conjunction with low probability of successful recruitment because of the restriction of suitable habitat to small areas in active regions of the narrow axial summit caldera along a spreading center (Kim & Mullineaux 1998; Tyler & Young 1999), would favor the evolution of rapid growth to reproductive size. Although insufficient data exist to model growth of R. pachyptila, the appearance of individuals in excess of 1.2 m in tube length in <1 year (Govenar et al. 2002), and over 1.5 m in tube length in <2 years (Lutz et al. 1994) on the East Pacific Rise suggests that they are among the fastest growing animals known (Lutz et al. 1994). The appearance of such diametrically opposed life-history strategies in two organisms belonging to the same group within a single family of polychaete annelids suggests that growth rates in this group are quite plastic and the low larval survival rates result in strong selection pressures on growth and fecundity.


In conclusion, Seepiophila jonesi represents an example of another long-lived seep vestimentiferan species. The growth pattern of S. jonesi appears to be one of discontinuous growth, where the tubeworm maintains its plume near the sediment–water interface where it is exposed to sulfide. In some cases, growth rates of both S. jonesi and Lamellibrachia luymesi varied among sites and among aggregations within a site, but growth rates were consistent among individuals within an aggregation. Investigations of growth rates in other species of seep tubeworms are required to examine the generality of these findings. This also presents an interesting case study for the evolution of life-history strategies in that two groups of closely related siboglinid polychaetes (vent and seep tubeworms) appear to have diverged drastically in terms of growth rate and longevity.


Many thanks to all of those who contributed to the measurement of tubeworms both at sea and in the laboratory: Shawn Arellano, Ana Hilario, Johanna Jarnegren, Mike McGinley, Bettina Pflugfelder, Liz Podowski, Ben Predmore, and Anna Van Gaest. Tubeworm collections would have been impossible without the assistance of the crew of the RV Seward Johnson and Captain George Gunther, and the crew and pilots of the Johnson Sea-Link submersibles. We would also like to acknowledge Bob Carney and Craig Young for their assistance and informative discussions at sea. This work was supported by the Minerals Management Service contract 1435-01-96-CT-30813, ship and submersible time from the NOAA National Undersea Research Center at UNC Wilmington and the NOAA Ocean Exploration program, and National Science Foundation grant OCE0117050. E.E.C. also received support from the NOAA Nancy Foster Scholarship Program.