Protoconch enlargement in Western Atlantic turritelline gastropod species following the closure of the Central American Seaway

Abstract The closure of the late Neogene interoceanic seaways between the Western Atlantic (WA) and Tropical Eastern Pacific (TEP)—commonly referred to as the Central American Seaway—significantly decreased nutrient supply in the WA compared to the TEP. In marine invertebrates, an increase in parental investment is expected to be selectively favored in nutrient‐poor marine environments as prolonged feeding in the plankton becomes less reliable. Here, we examine turritelline gastropods, which were abundant and diverse across this region during the Neogene and serve as important paleoenvironmental proxies, and test whether species exhibit decreased planktotrophy in the WA postclosure as compared to preclosure fossils and extant TEP species. We also test for differences in degree of planktotrophy in extant sister species pairs. Degree of planktotrophy was inferred by measuring the size of protoconchs, the species' larval shell that represents egg size. Protoconch size was compared between extant postclosure WA and TEP species and preclosure fossil species. To compare extant sister species, we reconstructed the phylogeny of available WA and TEP species using one nuclear (H3) and three mitochondrial markers (12S, 16S, and COI). Compared to the preclosure fossils, protoconch size increased in WA species but remained the same in the TEP species. In the two extant sister species pairs recovered in the phylogenetic analysis, the WA species are inferred to be nonplanktotrophic while the TEP species are planktotrophic. This suggests that decreased nutrient availability and primary productivity in the WA may have driven this change in developmental mode, and was the primary selective force resulting in postclosure turritelline extinctions.

The closure of the Central American interoceanic seaways and associated changes in nutrient conditions in the WA allows us the opportunity to directly test the relationship between decreased ambient nutrient supply and nutrient apportionment in gastropod eggs. A decrease in marine nutrient supply is expected to result in decreased planktotrophy success and therefore favor increased parental investment (Fortunato, 2004;Jablonski & Lutz, 1980;Lessios, 1990Lessios, ,2008Marshall et al., 2018;Miura, Frankel, & Torchin, 2011;Vance, 1973). Even if larvae still spend some time feeding in the plankton, larger offspring are better buffered against starvation and may need to spend less time in the plankton before settlement (Marshall & Keough, 2007;Marshall et al., 2018).
We chose turritelline gastropods ( Figure 1) to test our hypothesis that decreased nutrient availability selects for larger eggs.
F I G U R E 1 Turritella banksii (PRI 68087), a postclosure turritelline from the Tropical Eastern Pacific. Scale bar = 1 cm Inferences can be made about the larval mode of fossil and extant gastropods based on observations of the protoconch (larval shell), which is sometimes retained at the apex of the adult shell (Fortunato, 2002(Fortunato, ,2004Jablonski & Lutz, 1983;Lima & Lutz, 1990;Shuto, 1974;Thorson, 1950;Vendetti, 2007). Large, paucispiral protoconchs are presumed to be formed by larval gastropods that have spent little or no time in the plankton, and narrow, multispiral protoconchs are thought to indicate prolonged planktonic phases. Shuto (1974)  T. altilira a Panama, Colombia, Venezuela. Miocene. (Woodring, 1957) n/a n/a n/a n/a

T. clarionensis
Gulf of CA to Panama (Keen, 1971) T. gatunensis a Panama, Colombia, Venezuela (Conrad, 1857) n/a n/a n/a n/a

T. mariana = T. radula
Gulf of CA to southern Colombia (Keen, 1971) n/a n/a n/a n/a T. matarucana a Panama, Colombia, Venezuela (Hodson, 1926) n/a n/a n/a n/a  (Hodson, 1926) n/a n/a n/a n/a

| Taxon sampling
We sampled seven of the eight species of Turritella in the TEP and two of the four species in the WA (Table 1). Protoconch preservation can be rare, even in live-collected individuals from modern species.
Preclosure WA fossil species examined are those described from the late Miocene of Panama and the late Oligocene of Venezuela.
Although the only protoconch sampled for the extant species Vermicularia knorrii (Deshayes, 1843) was from a Pleistocene specimen, the specimen is considered part of the postclosure WA fauna.
Batillaria zonalis (Bruguière, 1792) and Cornell Lab of Ornithology. In molecular analyses, data sources are identified as UF = University of Florida, FLMNH collection, S = this study collected, and L = Lieberman, Allmon, and Eldredge (1993) from GenBank data.

| DNA extraction, sequencing, and alignment
Genomic DNA was extracted using the Qiagen DNeasy Kit from about 100 mg of tissue, following the manufacturer's protocol. We chose the mitochondrial 16S, 12S, cytochrome c oxidase subunit I (COI), and nuclear histone H3 regions for sequencing because 16S fragments are available from a subset of our species (Lieberman et al., 1993), and because there exist gastropod-specific primers for these genes (Miura,

| Phylogenetic analysis
Phylogenetic analysis of molecular characters was performed with parsimony, maximum-likelihood, and Bayesian methods. Parsimony analysis was run using PAUP* v. 4.0a141 (Swofford, 2002). Out of 2,328 total characters, 558 were parsimony-informative. Overall platform (Miller, Pfeiffer, & Schwartz, 2010) (Table   S1). In MrBayes, two runs were conducted with four chains each for 10 million generations. The first 25% of results were discarded as burnin. All other settings were left as default. Log files were combined and checked with Tracer v. 1.6 (Rambaut & Suchard, 2014). A statistical summary of the ML and Bayesian analyses is presented in Table S2.

| Protoconch measurements
Specimens with intact protoconchs were almost entirely found on juveniles less than one centimeter in length. Protoconchs are often abraded away in turritellines, even during the life of the organism . The protoconch is composed of two parts: protoconch I, which is the embryonic shell, formed prior to hatching and is unornamented, and protoconch II which is produced prior to metamorphosis, and which may be smooth or ornamented (Jablonski & Lutz, 1983;Robertson, 1971). Whole shells were sputter-coated with a thin layer of gold then imaged on a scanning electron microscope (JCM-6000 NeoScope Benchtop SEM) at the PRI.
Venezuelan specimens from the type and figured collection of the PRI were imaged without sputter-coating. Side and top view images were taken to identify the protoconch I-protoconch II boundary, which was then marked on the top view image. We used this boundary to find the total number of volutions (full 360-degree spirals) in protoconch I.

| Analysis of protoconch character divergence
Statistical comparisons were made among protoconch data obtained from preclosure fossil, postclosure Atlantic, and postclosure Pacific specimens in Past3 (Hammer, Harper, & Ryan, 2001). Both protoconch maximum diameter and diameter/volutions ratio were compared. Tukey's Q was calculated to make comparisons among means for all three data sets simultaneously, with significance estimated according to the method of Copenhaver and Holland (1988). The Mann-Whitney U test was applied to determine whether the samples were likely to be drawn from the same distributions.
Continuous character mapping of protoconch diameters on the molecular phylogeny was achieved using the "contMap" function in the "phytools" (Revell, 2012) package for R. The "contMap" function estimates character states at internal nodes using ML methods (function "anc.ML"). From the Bayesian tree, multiple individuals for each species were collapsed into one tip using the "delete subelements" function in TreeGraph2 (v. 2.14.0-771) (Stöver & Müller, 2010) to create a consolidated backbone. The average protoconch diameter for each species was then mapped onto each tip.

| Molecular phylogeny
Two extant sister species pairs are consistently identified in the molecular trees (Figures 2-4). The first pair is T. exoleta (WA) and T. radula (TEP), which was discovered in all three methods (parsimony, F I G U R E 2 Majority rule parsimony tree (consensus of 81 trees) generated from mitochondrial and nuclear sequences. All species are from genus Turritella, except for out-groups Batillaria zonalis and Lampania cumingi. L1 = sequence from Lieberman et al. (1993); S1 or S2 = specimen collected for this study; UF1 or UF2 = specimen from FLMNH and T. nodulosa (TEP), is identified in both the maximum-likelihood and Bayesian results (Figures 3 and 4), but exists as a "sister species cluster" with T. rubescens under the parsimony method ( Figure 2).
All methods find three major clades within these turritellines: (a) T. exoleta and T. radula sister to all other taxa, (b) T. bacillum and T. terebra sister to the remaining taxa, and (c) all other species. Most of the incongruence is located within this last clade due to unstable placement of T. banksii, T. leucostoma, and T. rubescens among methodologies.

| Protoconch size changes after closure of the Central American Seaway
Protoconch size data were obtained for the species identified in Table 3. We found that postclosure WA turritelline species as a whole experienced significant change in both protoconch diameter (Tables 4 and 5) and in diameter/volutions (D/Vol) compared with preclosure values (Figure 5; Tables 4 and 6). No significant change was found in TEP species relative to the preclosure fossil species in maximum diameter (Table 5) or diameter/volutions ratio (Table 6). (Table 6)

| Evolution of developmental mode in Central American Isthmus turritellines
We find evidence of increased nonplanktotrophy in WA species and conclude that this was likely a response to decreased nutrient availability in the WA after the closure of the interoceanic seaways. The F I G U R E 3 Maximum-likelihood tree generated from mitochondrial and nuclear sequences. Unless noted, bootstrap values at each node are 100. All species are from genus Turritella, except for out-groups Batillaria zonalis and Lampania cumingi. L1 = sequence from Lieberman et al. (1993); S1 or S2 = specimen collected for this study; UF1 or UF2 = specimen from FLMNH. Scale bar represented mean number of nucleotide substitutions per site

Lampania cumingi
Batillaria zonalis similarity of preclosure protoconch diameters to postclosure TEP protoconch diameters affirms that the observed difference in the modern populations is not due to a decline in mean protoconch size in the Pacific. The minimum protoconch size observed (157.5 μm) was from a TEP species and is nearly half of the minimum observed size for WA species (316.2 μm). Maximum observed protoconch sizes were similar between modern WA (475 μm) and TEP (470 μm) species. This indicates that selection against small protoconch size was the likely driver of this change. Because we show that the WA species examined are sister to EP species in our molecular phylogeny, we regard the evolution of increased nonplanktotrophy as separate, independent occurrences within each lineage. The observation that these surviving lineages have independently increased protoconch sizes relative to preclosure along with the observation of similar changes in other taxa (Fortunato, 2004;Jackson, Jung, & Fortunato, 1996;Lessios, 1990;Miura et al., 2011;Moran, 2004;Wehrtmann & Albornoz, 2002) strongly supports the adaptive significance of these changes.
This phylogeny updates the only existing molecular phylogeny of turritellines that was based only on partial mitochondrial 16S sequences (Lieberman et al., 1993). We recover a different topology than presented in Lieberman et al. (1993) due to the additional genetic data; we double the read length for the 16S sequences and add in three other genes to our dataset. In that previous study, which included many of the same species, turritellines were used as a case study to investigate whether there were signals of species selection favoring increased diversification of nonplanktotrophic species. This hypothesis was motivated by the observation that nonplanktotrophic turritelline species outnumber planktotrophic species approximately 3:1 in the Neogene Gulf Coastal Plain, and worldwide today nonplanktotrophic species are twice as common as planktotrophic species (Allmon, 1992). Larval mode has been F I G U R E 4 Bayesian tree generated from nuclear and mitochondrial sequence data. Posterior probabilities at nodes are 100 unless noted. All species are from genus Turritella, except for out-groups Batillaria zonalis and Lampania cumingi. L1 = sequence from Lieberman et al. (1993); S1 or S2 = specimen collected for this study; UF1 or UF2 = specimen from FLMNH T. nodulosa S1 T. terebra S1 T. bacillum S1 T. bacillum S2
While the Lieberman et al. (1993) topology is markedly different from the ones shown in this paper, both studies indicate that the ancestral turritelline condition was planktotrophy. This conclusion is further strengthened for the observed taxa by our finding that all preclosure forms in the Central American Isthmus region had protoconch sizes indicative of planktotrophy. Additionally, both this paper and Lieberman et al. (1993) find that nonplanktotrophy arose within single species instead of at the base of nonplanktotrophic clades, and so likely did not drive increased speciation in the sampled taxa.
TA B L E 3 Turritelline protoconch diameters and diameter/volutions ratios observed in this study The pattern of increased protoconch size in postclosure WA species is consistent with our hypothesis that decreased nutrient availability in the WA selected for nonplanktotrophy, and the phylogeny indicates that at least two of the three Recent WA species evolved larger protoconch sizes independently (Figure 7).
Further research is, however, necessary to determine the underlying  ing planktotrophy also may bias the long-term accumulation of nonplanktotrophy in a clade, if the transition to nonplanktotrophy has no consequences for speciation (Duda & Palumbi, 1999;Krug et al., 2015). There are two chief difficulties in assessing which of these macroevolutionary mechanisms was involved in the transition to larger protoconch sizes in WA turritellines. First, additional fossil protoconch data would need to be incorporated into a phylogenetic framework to distinguish between these evolutionary histories.
Data from additional protoconchs, with both high-resolution stratigraphic data and confident species assignments, are obviously vital to assess the possibility of anagenetic selection for protoconch size increase. Efforts should be made to document protoconch sizes in the literature where possible, even maximum diameters from fragmented protoconchs, and collecting efforts should take special care not to neglect small apical fragments which may be rapidly screened for protoconchs using light microscopy. Second, the present status of turritelline systematics presents a further difficulty. It has been the operational assumption of many studies that long-distance dispersal events among turritellines are rare (e.g., Marwick, 1957). Both Lieberman et al. (1993) and the present study suggest that this assumption should be treated with some caution as there appear to be two clades in the neotropics, one of which is sister to a clade of species from South-East Asia. A global molecular phylogeny of Recent turritellines is needed to assess the validity of this assumption in regard to fossil species from the tropical Americas, and to aid in determining what morphological characters may be informative in assigning species to these clades.
Regardless of the evolutionary mechanisms involved in achieving nonplanktotrophy, decreased planktotrophy and diversity of turritellines after the closure of tropical American interoceanic seaways will likely have long-term consequences for the evolution of turritellines. Species-poor clades are more likely to be subject to stochastic extinction, and low-dispersal larvae may result in shifts in speciation rates, or, in punctuational systems, shifts in rates of morphospace exploration (Jablonski, 2017;Krug et al., 2015). Our phylogeny indicates that the two WA species examined are not closely related, and therefore, loss of either would substantially decrease the phylogenetic diversity (Faith, 1992) present in the region. The loss of planktotrophy has also been considered subject to Dollo's law, with limited opportunities for reversal due to the complex of characters necessary for larval feeding (Krug et al., 2015). This may not be the case as even direct-developing gastropods may pass through a veliger stage within the egg, without loss of associated characters (e.g., larval velum; Collin, 2004;Collin, Chaparro, Winkler, & Veliz, 2007;Collin & Cipriani, 2003;Collin & Miglietta, 2008). If increased nonplanktotrophy decreases net diversification rates, selection toward higher parental investment in WA turritelline clades may have contributed to the overall decline in WA diversity as well (Krug et al., 2015). Investigating the evolution of WA and TEP turritelline protoconch size in a phylogenetic context may distinguish whether nonplanktotrophy has led to decreased net diversification rates in WA turritellines. If such a decrease is observed, then this shift in protoconch sizes may be evidence that WA turritellines represent two "dead clades walking" (Jablonski, 2002;Krug et al., 2015) following the Pliocene extinctions.

ACK N OWLED G M ENTS
We would like to thank A. Hendy, J. Santiago, L. Londono, and C. Jaramillo and the Smithsonian Tropical Research Institute for their assistance in facilitating collecting in Panama. We also thank Ricardo Perez for donating the Toyota vehicles used for fieldwork and the

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N
Stephanie Sang drafted the initial manuscript, performed molecular analyses, collected specimens, and performed analysis of both modern protoconchs and fossil protoconchs from Panama.

DATA ACCE SS I B I LIT Y
All molecular sequence data used in our analyses are available in GenBank, as outlined in Table 1.