J. Phycol. 56, 110–120 (2020) © 2019 Phycological Society of America DOI: 10.1111/jpy.12918-19-050
SEASCAPE GENETICS OF THE STALKED KELP PTERYGOPHORA CALIFORNICA AND COMPARATIVE POPULATION GENETICS IN THE SANTA BARBARA CHANNEL1
Heidi L. Hargarten Tree Fruit Research Laboratory, US Department of Agriculture – Agricultural Research Service, Wenatchee, Washington 98801, USA Mattias L. Johansson Department of Biology, University of North Georgia, Oakwood, Georgia 30566, USA Daniel C. Reed Marine Science Institute, University of California, Santa Barbara, California 93106, USA Nelson C. Coelho Department of Computational and Systems Biology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, USA David A. Siegel Earth Research Institute and Department of Geography, University of California, Santa Barbara, California 93106, USA and Filipe Alberto2 Department of Biological Sciences, University of Wisconsin – Milwaukee, Milwaukee, Wisconsin 53201, USA
We conducted a population genetic analysis of the differentiation and pairwise differentiation were stalked kelp, Pterygophora californica, in the Santa similar among patches between the two kelp species, Barbara Channel, California, USA. The results were indicating that they have similar dispersal capabilities compared with previous work on the genetic despite their differences in rafting ability. These differentiation of giant kelp, Macrocystis pyrifera, in results suggest that rafting sporophytes do not play the same region. These two sympatric kelps not only a significant role in effective dispersal of share many life history and dispersal characteristics M. pyrifera at ecologically relevant spatial and but also differ in that dislodged P. californica does temporal scales. not produce floating rafts with buoyant fertile sporophytes, commonly observed for M. pyrifera.We Key index words: comparative population genetics; used a comparative population genetic approach with dispersal; genetic structure; giant kelp; marine con- these two species to test the hypothesis that nectivity; seascape genetics; stalked kelp the ability to produce floating rafts increases the Abbreviations: AB, Arroyo Burro; AH, Arroyo genetic connectivity among kelp patches in the Santa Hondo; AIC, Akaike information criterion is an esti- Barbara Channel. We quantified the association of mator of the relative quality of statistical models for habitat continuity and oceanographic distance with a given set of data; AQ, Arroyo Quemado; AR, Alle- the genetic differentiation observed in stalked kelp, lic richness standardized by sample size; B, Bullito; like previously conducted for giant kelp. We C, Carpinteria; CPG, Comparative population genet- compared both overall (across all patches) and ics; DEST, Estimator of genetic differentiation not pairwise (between patches) genetic differentiation. biased by the scale of genetic diversity (allelic rich- We found that oceanographic transit time, habitat ness or heterozygosity); FIS, Inbreeding coefficient; continuity, and geographic distance were all FST, Genetic differentiation or the proportion of associated with genetic connectivity in P. californica, the total genetic variance contained in a subpopula- supporting similar previous findings for M. pyrifera. tion relative to the total genetic variance; GeoDist, Controlling for differences in heterozygosity between Geographic distance; G, Goleta Bay; HabCont, kelp species using Jost’s DEST, we showed that global Habitat continuity; IBD, Isolation by distance model of genetic differentiation; IV, Isla Vista; M, 1Received 19 March 2019. Accepted 29 August 2019. First Pub- Mohawk; N, Naples Reef; TT, Minimum oceano- lished Online 12 September 2019. Published Online 6 November 2019, Wiley Online Library (wileyonlinelibrary.com). graphic transit time between two sites, the shortest 2Author for correspondence: e-mail [email protected]. transit time of the pair pop.i to pop.j or pop.j to Editorial Responsibility: M. Roleda (Associate Editor) popi
110 POPULATION GENETICS OF STALKED KELP 111
Dispersal is the universal mechanism that pro- Gene flow in kelps is intimately linked to their motes gene flow and population connectivity. heteromorphic life history characterized by a large Although several different dispersal mechanisms diploid sporophyte and a microscopic haploid have evolved in the ocean, the primary strategy gametophyte. Sporophytes produce haploid zoos- employed by species with sessile adult forms is the pores via meiosis that are released into the water production of planktonic propagules that are pas- column and disperse passively via oceanic currents sively dispersed locally or over long distances by over distances of meters to kilometers (Reed et al. ocean currents (Siegel et al. 2003, Sagarin et al. 1988, 1992, 2004, Gaylord et al. 2006, 2012). Upon 2006, Cowen and Sponaugle 2009, Selkoe et al. settling to the bottom zoospores germinate into ses- 2016, Padron et al. 2018, Xuereb et al. 2018). Con- sile, free-living male and female gametophytes that ventional theory predicts that organisms having produce gametes that disperse over distances of just long-lived planktonic stages have higher levels of a few millimeters (Muller 1981, Reed 1990). Thus, gene flow than those with short-lived planktonic genetic connectivity among kelp populations occurs stages (Siegel et al. 2003, Weersing and Toonen almost exclusively either through the dispersal of 2009, Selkoe et al. 2010). However, this is not always planktonic zoospores or reproductive sporophytes the case as cryptic ocean barriers, environmental that become dislodged and set adrift. The relative gradients, and temporal oscillations in oceano- importance of these two mechanisms in contribut- graphic circulation can create unexpected patterns ing to gene flow and population dynamics in kelps of connectivity and genetic structure across a range has been a topic of much discussion and debate of spatial scales (Gilg and Hilbish 2003, Johansson (Dayton 1985, Santelices 1990, Reed et al. 2004, et al. 2008, Treml et al. 2008, Alberto et al. 2011, 2006, Graham et al. 2007, Schiel and Foster 2015). DeFaveri et al. 2013, Liggins et al. 2013, Xuereb Integrating genetic studies of adult populations into et al. 2018). The effects of these factors on dispersal this discussion has the potential to provide addi- and gene flow depend on population history tional insight into the relative importance of these (Nesbo€ et al. 2000, Pelc et al. 2009), demography two mechanisms in promoting dispersal and gene (Dawson et al. 2002), life history (Shulman and Ber- flow in kelps (Alberto et al. 2010, 2011, Valero et al. mingham 1995, Sponaugle and Cowen 1997, Turner 2011, Robuchon et al. 2014, Evankow 2015). and Trexler 1998), and propagule behavior In the northeast Pacific, the kelps Pterygophora cali- (Pineda-Krch and Fagerstrom 1999, Paris et al. fornica and Macrocystis pyrifera commonly co-occur 2007, Pringle and Wares 2007, Woodson and McMa- in kelp forests from British Columbia, Canada to nus 2007, Cowen and Sponaugle 2009, Morgan and Baja California, Mexico (Abbott et al. 1992). The Fisher 2010, Pineda et al. 2010). As a result, com- stalked kelp P. californica grows 1–3 m in height and parisons of gene flow patterns between species or forms dense stands of ridged palm-like sporophytes, across different studies of the same species can be whereas the giant kelp M. pyrifera uses gas bladders challenging (Bird et al. 2007, Liggins et al. 2013, to extend through the water column to produce a Sexton et al. 2014). flexible floating canopy at the sea surface. The two Here, we use a comparative population genetics species co-occur in the same rocky habitat at similar (CPG) approach to examine the dispersal of two depths, experience many of the same biotic and abi- sympatric species of kelp (Order Laminariales). We otic pressures, and also share the same basic life his- define CPG as the study of genetic differentiation of tory. Moreover, the zoospores of both species are two or more taxa that share many life history traits released near the bottom, sink at similar rates, and and demographic history in a restricted geographic have similar planktonic durations (Reed et al. 1992, area. CPG has been used to examine the role of lar- Gaylord et al. 2002), which influence the distances val strategies (Lambert et al. 2003, Watts and over which they disperse (Norton 1992, Gaylord Thorpe 2006, Barbosa et al. 2013), life history char- et al. 2002). The primary difference in the dispersal acteristics (Criscione and Blouin 2004), and envi- potential of these two species is that unlike P. califor- ronmental gradients (White et al. 2011, DeFaveri nica, M. pyrifera is positively buoyant and intact et al. 2012) in understanding patterns of gene flow sporophytes that are dislodged by large waves create in a variety of marine and aquatic systems. More- floating rafts that are dispersed by winds and cur- over, comparing genetic differentiation between rents at the sea surface promote dispersal between populations of sympatric species with similar life his- populations that are tens or hundreds of kilometers tory traits over a limited geographic area has been apart from one another (Hobday 2000, Macaya used to make inferences about dispersal patterns for et al. 2005, Hern andez-Carmona et al. 2006, Gra- each species (Shulman and Bermingham 1995, ham et al. 2007, Gaylord et al. 2012). Sponaugle and Cowen 1997, Turner and Trexler For our CPG study, we performed a microsatel- 1998, Dawson et al. 2002, Nikula et al. 2011a,b). lite-based population genetics analysis of Pterygophora CPG has proved to be particularly useful for under- californica in the Santa Barbara Channel using the standing mechanisms that promote gene flow and same sites previously analyzed for Macrocystis pyrifera population persistence in kelps (Valero et al. 2011, (Alberto et al. 2010). We compared how among site Robuchon et al. 2014, Evankow 2015). genetic differentiation and its drivers, namely 112 HEIDI L. HARGARTEN ET AL. oceanographic transit time, habitat continuity, and and a final elongation step at 72°C for 20 min. All PCR reac- geographic distance, varied between the two species. tions were performed on a GeneAmp 9700 thermocycler (PE We hypothesized that if rafting adults is an impor- Applied Biosystems). An ABI PRISM 3130xl DNA analyzer was used to analyze fragment length using the GeneScan Liz 500 tant mechanism of dispersal, then P. californica size standard (Applied Biosystems). Raw allele sizes were should show more genetic structure among sites scored with STRand (http://www.vgl.ucdavis.edu/informatic than M. pyrifera. Conversely, if gene flow depends s/STRand), binned and reviewed for ambiguities using the R primarily on the transport of planktonic zoospores, package MsatAllele (Alberto et al. 2009). then we would expect the level of genetic structure To estimate genetic diversity in Pterygophora californica pop- among sites to be similar for the two species. To ulations, allelic richness, standardized for sample size, was cal- our knowledge, our study is the first seascape analy- culated using the R package “standArich” (Alberto et al. 2006). The null hypothesis of population Hardy–Weinberg sis of genetic diversity and population differentia- Equilibrium and inbreeding coefficients (FIS) not deviating tion in P. californica. from zero were tested using Genepop v.4 (Rousset 2008). Populations were checked for the presence of null alleles using Micro-Checker (Van Oosterhout et al. 2004). Excess MATERIALS AND METHODS homozygosity was identified at three loci (Pc-10, Pc-14, and Field sampling and genetic differentiation analyses. Tissue sam- Pc-17). To detect whether these loci were affecting pairwise ples were collected from approximately 50 individuals of Ptery- FST calculations, the program FreeNA (Chapuis and Estoup gophora californica in summer 2010 at the same nine sites 2007) was used to determine whether null alleles needed to previously sampled for Macrocystis pyrifera along the mainland be eliminated from the dataset, by comparing FST values of coast of the Santa Barbara Channel, California (Fig. 1); Bul- data with null alleles versus data without null alleles. FreeNA lito (B), Arroyo Hondo (AH), Arroyo Quemado (AQ), Naples corrects for the positive bias induced by the presence of null Reef (N), Isla Vista (IV), Goleta (G), Arroyo Burro (AB), alleles on FST, providing a more accurate estimation of FST in Mohawk (M), and Carpinteria (C). the presence of null alleles. Global FST estimates using the “eliminate null alleles” correction method (F = 0.0605) A single blade was collected from each individual by divers ST were roughly similar to uncorrected values (F = 0.0629). A and transported to the laboratory in a cool insulated con- ST paired t-test between all pairwise population corrected and tainer. Tissue from each blade was dried and stored in silica uncorrected F values revealed that the difference was not gel for DNA extraction. Genomic DNA was extracted using ST significant (t = 1.179, P = 0.247). Therefore, analyses were the Nucleospin 96 Plant Kit (Macherey-Nagel, Germany). All 35 run with non-corrected F values. samples were genotyped using seven microsatellite loci previ- ST ously designed for Pterygophora californica (see Appendix S1 Jost’s DEST estimator of genetic differentiation was used for and Table S1 in the Supporting Information for sequences comparisons between the two species and calculated using and GenBank accession numbers). PCRs were performed in the R package “diversity” (Keenan et al. 2013). Jost’s DEST volumes of 15 lL containing 20 ng of DNA, 0.1 lMof provides an estimate of among-population genetic diversity that is not affected by the within-population levels of diversity each primer, 0.8 mM of dNTPs, 2.0 or 2.5 mM of MgCl2 (for individual locus PCR conditions, see Table S1), 3.0 lLof59 (heterozygosity). This statistic is well suited for inter-specific PCR Buffer, and 0.4 U of GoTaq Polymerase (Promega, comparisons because it accounts for differing allelic richness Madison, WI, USA). Cycling conditions consisted of an initial and heterozygosity across species (Jost 2008). Global DEST was denaturing step of 5 min at 95°C, followed by 35 cycles of calculated for both Macrocystis pyrifera and Pterygophora califor- 30 s at 95°C, 30 s at annealing temperature, 45 s at 72°C, nica. Confidence intervals for Jost’s DEST were determined using 1,000 bootstrapped replicates. To account for the inherent bias in bootstrapping measures of genetic differenti- ation, we used the bias-corrected values for DEST and 95% confidence intervals in pairwise population comparisons (Keenan et al. 2013). We compared Pterygophora californica and Macrocystis pyrifera rate of change in Jost’s DEST genetic differentiation as a func- tion of three different predictors using simple linear regres- sion. The population pairwise predictor variables used for these analyses were as follows: (i) Euclidian distance (isola- tion by distance, IBD model), (ii) oceanographic transit time, and (iii) habitat continuity. We tested whether the slopes of these regressions were different using ANCOVA to determine whether rates of change in genetic differentiation differed between the two species based on these predictor variables. Oceanographic transit time. Oceanographic distance has pro- ven to be a better predictor of genetic connectivity than Euclidian distance in several marine systems (Weersing and Toonen 2009, White et al. 2010, Singh et al. 2018), including kelps (Alberto et al. 2011, Johansson et al. 2015, Durrant et al. 2018). In our previous work (Alberto et al. 2011), we combined data on population genetics, habitat availability, Euclidian distance, and oceanographic connectivity (derived using a Lagrangian particle simulation model) in a multiple regression model to determine the best predictors of among- FIG.1. Pterygophora californica sampling locations located in the site genetic differentiation of Macrocystis pyrifera in the Santa Santa Barbara channel, California, USA. Barbara Channel. We used this same approach to evaluate POPULATION GENETICS OF STALKED KELP 113 the importance of these predictor variables in explaining the relative importance of each predictor while controlling among-site genetic differentiation in Pterygophora californica. for the effects of the remaining ones. In this case, there were By doing so, we were able to evaluate whether the processes no comparisons between taxa with different levels of influencing the genetic structure of these two sympatric, heterozygosity; thus, pairwise genetic differentiation between albeit morphologically different, species were similar. populations was described using the formula FST/(1 FST)as Dispersal trajectories were simulated for over 50 million suggested by Rousset (1997), instead of Jost’s D. Simple passive Lagrangian particles in the Southern California Bight regression models identified spring months as the ones with domain over the period from January 1, 1996 to December the highest goodness of fit; therefore, spring oceanographic 31, 2002, with 135 uniformly distributed, near-shore circular transit time (average value in the months of April–June) was patches (5 km in radius) as release sites (see Mitarai et al. the predictor used in multiple regression. We compared 2009 for more information on the Lagrangian particle simu- models with different combinations of predictor variables lations). Monthly mean particle transit times were calculated using the Akaike information criterion (AIC). All regression in days for particles traveling between each pairwise cell in analyses were done in R (R Core Team 2015). the model and averaged again across the 7 years of simulated data. An important property of Lagrangian particle simulations RESULTS is the inherent asymmetry in transit times between popula- Comparative population genetics in Pterygophora cali- tions; particles traveling from site i to j might have different fornica versus Macrocystis pyrifera. Microsatellite transit times than particles traveling in the opposite direction, from j to i (Watson et al. 2011). Therefore, we used the data indicated moderate genetic diversity in Pterygo- shorter of the two oceanographic distances between pairwise phora californica (allelic richness range 5.95–8.35) populations to account for this asymmetry. By doing so, we across all sites. P. californica allelic diversity was lower created a measure for the minimum mean oceanographic than corresponding Macrocystis pyrifera populations transit time (TT), between two populations. Hereafter, we (M. pyrifera allelic richness 11.22–13.46; Table 1), refer to this measure as oceanographic transit time. The indicating lower genetic diversity in P. californica effect of intra-annual variability in oceanographic transit time compared to M. pyrifera in this region. Additionally, was investigated by averaging the 7 years of simulated data in – different time intervals (monthly, quarterly, and annually). for P. californica, there were deviations from Hardy Transit times during the months of June, July, August, Weinberg equilibrium, as measured by FIS, in only September, and October were removed from our models two sites (Table 1, Table S2 in the Supporting Infor- because Pterygophora californica does not produce spores dur- mation). In M. pyrifera, significant FIS values were ing these months (Reed et al. 1996) and including them attributed to the effect of null alleles in four loci would have led to biologically irrelevant interpretations. Two pairs of populations occur within the same 5-kilometer (see appendix B in Alberto et al. 2009). Populations oceanographic cell used to estimate transit time, AH and AQ, were significantly differentiated from one another and AB and M. This prohibits measures of oceanographic except for the neighboring populations Isla Vista connectivity from being estimated for those pairwise compar- (IV) and Goleta (G; P = 0.071), as shown by pair- isons; thus, sites AH and AB were removed from regression wise FST (Table S3 in the Supporting Information). analyses using oceanographic transit time. Genetic differentiation among all Pterygophora cali- Habitat continuity. The area of habitat between two fornica populations (global D = 0.0911) was patches is associated with the level of stepping stone gene EST flow in kelps (Billot et al. 2003, Faugeron et al. 2005, Alberto higher than genetic differentiation found among all et al. 2010, Coleman 2013, Robuchon et al. 2014). Here, Macrocystis pyrifera populations (global = habitat continuity was defined as area of kelp in hectares per DEST 0.0736), although 95% confidence intervals kilometer of linear coastline between all pairs of sites. overlapped between P. californica (global mean = = = Because Pterygophora californica is an understory species to DEST 0.1056, CIlower 0.0947, CIupper 0.1172) Macrocystis pyrifera occupying the same rocky reef habitat, = and M. pyrifera (global mean DEST 0.1108, habitat continuity values were the same as used in Alberto CI = 0.0942, CI = 0.1309). A paired t-test et al. (2010). We used a composite layer of annual giant kelp lower upper cover from 1988 to 2003 which represents the maximum used to determine whether pairwise population DEST extent of kelp habitat in the study region including years differed between the two species confirmed the dif- = = when kelp cover was near maximum. This layer reflects the ference was not significant (t35 1.76, P 0.086), submerged rocky reef area at giant kelp colonization depth. suggesting that genetic differentiation of the two There is no reason to believe that this rocky reef structure – species in the Santa Barbara Channel is similar. changed substantially from 1988 2003 to 2010. Because the ANCOVA provided additional evidence that P. cali- genetic structure sampled at any time is always the result of multiple, overlapping generations (and thus encompasses fornica and M. pyrifera dispersal distance was not sig- habitat continuity variation), we considered that a composite nificantly different. The rate of change in DEST for of the maximum kelp cover should represent the best synthe- the two species was similar for oceanographic transit sis for this effect on gene flow. The habitat continuity layer time (interaction P = 0.509; slope = 0.017; was characterized using the California Department of Fish R2 = 0.38), habitat continuity (interaction and Wildlife kelp cover GIS layer (https://map.dfg.ca.gov/ P = 0.603; slope = –0.006; R2 = 0.23), and geo- arcgis/rest/services/Project_Marine/Marine_Habitat/MapSe graphic distance (interaction P = 0.883; rver). = 2 = Modeling genetic differentiation in Pterygophora califor- slope 0.001; R 0.19; Fig. 2). nica. To obtain a single model of genetic differentiation for Modeling genetic connectivity in Pterygophora califor- Pterygophora californica, we used the predictor variables nica. Single regression models identified oceano- described above in a multiple regression model to estimate graphic transit time during the month of May as the 114 HEIDI L. HARGARTEN ET AL.
TABLE 1. Summary statistics for comparative population genetics between Pterygophora californica (Pc) and Macrocystis pyrifera (Mp) sampled at the same sites in the Santa Barbara Channel, California. Abbreviations are as follows: Pop = Population name (refer to main text for full names), N = sample size, AR = standardized allelic richness, He = expected heterozygos- = *≤ **≤ ***≤ ity, FIS inbreeding coefficient, 0.01 significance level, 0.001 significance level, 0.0001 significance level. Macro- cystis pyrifera values are from Alberto et al. (2010). FIS significance tests per population and per locus are shown in Table S2.
NARHeFIS Pop Latitude (North) Longitude (West) Pc Mp Pc Mp Pc Mp Pc Mp B34°27031.98″ 120°20000.36″ 52 52 7.43 11.42 0.611 0.767 0.016 0.183 AH 34°28018.72″ 120°08039.78″ 50 32 8.24 11.22 0.581 0.790 0.061 0.175 AQ 34°28007.62″ 120°07017.10″ 50 50 7.32 12.03 0.547 0.760 0.018 0.121 NP 34°25020.40″ 119°57010.56″ 50 49 6.06 12.93 0.567 0.777 0.005 0.215 IV 34°24010.20″ 119°51028.32″ 50 50 7.11 12.62 0.579 0/776 0.113 0.215 G34°24049.62″ 119°49020.64″ 50 47 7.17 12.11 0.585 0.768 0.101* 0.148*** AB 34°24000.42″ 119°44039.78″ 49 37 7.15 12.87 0.620 0.740 0.009 0.141*** M34°23039.60″ 119°°43048.00″ 50 44 8.35 13.46 0.616 0.795 0.228*** 0.156*** C34o23032.70″ 119°32037.68″ 52 50 5.95 11.78 0.513 0.780 0.041 0.134***
single best predictor of genetic differentiation (FST) differentiation between the two species were non- in Pterygophora californica (Table 2). Habitat continu- significant. We note that when using FST to run ity was a better predictor of genetic differentiation these comparisons (data not shown), M. pyrifera than geographic distance. The multiple regression genetic differentiation was higher than P. californica. model with highest goodness of fit included habitat However, the levels of heterozygosity are also higher continuity, geographic distance, and transit time in the former kelp and the comparison using FST is during the month of April (P = 0.002, R2 = 0.503, thus biased (Jost 2008). The disparity between two = AIC 125.04) and was considered the best overall different measures of genetic variation, FST and model (Fig. 3). DEST, highlights the importance of using directly Oceanographic transit time and gene flow in Ptery- comparable genetic measurements to produce gophora californica. During the winter months, meaningful data in cross-species comparative stud- oceanographic transit time was much longer than in ies. Additionally, it provides an example of the use the spring months, with May having the fastest over- of Jost’s DEST in studies of isolation by distance, and all transit times between populations (Fig. 4, shaded other genetic differentiation drivers, and exempli- gray area). Spore production and subsequent fies its usefulness when compared with traditional release (as measured by sorus area) increase during measures of genetic variation. Therefore, when the winter, when oceanographic current velocity is genetic differentiation was made comparable, we slow. Spore release has been inferred for the period did not find support for the hypothesis that rafting when currents speed up, by the decrease in sorus sporophytes extend gene flow in M. pyrifera at the area without new sporangial tissue production geographic scale of this study. (Reed et al. 1996). The best model predicting While we consider the positive buoyancy of dis- genetic differentiation was observed during this per- lodged Macrocystis pyrifera sporophytes carrying iod of declining sorus area and fast oceanographic viable propagules to be the key difference in disper- transit times (AIC, broken gray line, Fig. 4). sal potential between M. pyrifera and Pterygophora cal- ifornica, other life history differences may play a role explaining genetic differentiation between them. DISCUSSION For example, M. pyrifera, reproduces throughout the Comparative population genetics of Pterygophora cali- year, with seasonal peaks occurring twice annually fornica and Macrocystis pyrifera. Our CPG study during early winter and late spring (Reed et al. focused on the regional levels of genetic differentia- 1996). Continuous spore release year round maxi- tion between Pterygophora californica and Macrocystis mizes dispersal potential during periods of high pyrifera. Given life history similarities between these advective flow, which we would expect to decrease kelps, we tested the hypothesis that gene flow was genetic differences between and among M. pyrifera more restricted in P. californica because this kelp patches. Pterygophora californica, meanwhile, has a lacks floating rafts that may be important dispersal strict reproductive window with highly synchronous vectors in M. pyrifera. We found that (i) genetic dif- spore release from November to early May (De ferentiation between the two species was similar Wreede 1986, Reed 1990, Reed et al. 1996, 1997). when DEST was used to control for different levels of Synchronization in spore release is predicted to pro- within-species genetic diversity and (ii) differences mote gene flow by increasing the spore cloud that is in the levels of among population (global) and available to disperse per unit of time. These periods between population (pair-wise) genetic also coincide with favorable conditions for not only POPULATION GENETICS OF STALKED KELP 115
recruitment (Reed and Foster 1984, Deysher and Dean 1986, Reed 1990, Reed et al. 1996) but also increasingly extended dispersal as winter transitions to spring. The combination of these conditions could lead to lower levels of genetic differentiation in P. californica. The age structure of kelp popula- tions might also have an effect on gene flow and the genetic makeup of patches across a region. P. californica sporophytes tend to live longer than M. pyrifera sporophytes (Rosenthal et al. 1974, Hyman- son et al. 1990). Different lifespans mean different overlapping generation periods between these kelps, but its effect on effective population size, genetic drift, and differentiation is difficult to predict in natural populations because of the large array of unknown life parameters involved (Choy and Weir 1978, Emigh and Pollak 1979). Our results using a comparative genetic approach agree with previous studies analyzing the effects of kelp rafts on patch dynamics. For example, Reed et al. (2004) found no correlation between the size of the spore source provided by Macrocystis pyrifera rafts and the density of new M. pyrifera recruits on a large artificial reef, which instead was positively cor- related with distance from the nearest standing pop- ulation of M. pyrifera. This observation indicated that distant spore dispersal from extant populations, rather than local spore dispersal from drifting rafts, was more likely to be the source of new recruits that initially colonized the artificial reef. Alberto et al. (2011) found that oceanographic connectivity for late spring (June) had the best-fit predicting genetic differentiation in M. pyrifera. However, this time per- iod is when whole sporophyte dislodgement is mini- mal (Reed et al. 2008). The near absence of rafters during periods of environmental conditions that produce an optimal setting for extending dispersal distances could explain why floating sporophytes are a negligible component of gene flow among M. pyrifera patches. Although evidence argues that floating rafts con- tribute little to population dynamics of Macrocystis pyrifera, they may contribute to infrequent but nonetheless important episodes of gene flow that maintain genetic connectivity across greater geo- graphic distances and longer temporal scales (Gille- spie et al. 2012, Saunders 2014). Infrequent contributions from drifters may still be adequate to supply the “one migrant per generation” needed to maintain gene flow and dilute the effects of genetic drift that create high levels of differentiation among distant populations in the absence of long-range dis- persal capabilities. Genetic studies that target longer temporal scales and done across its global range
FIG. 2. Pairwise genetic differentiation DEST association with bolster this argument for M. pyrifera (Coyer et al. different predictors is compared between Pterygophora californica 2001, Macaya and Zuccarello 2010, Astorga et al. (open dots, thin regression lines) and Macrocystis pyrifera (closed 2012, Johansson et al. 2015). Species comparisons dots, thick regression line). Predictors of genetic differentiation, from top to bottom, are as follows: mean spring oceanographic of microsatellite genetic differentiation at larger spa- transit time (days, see text for details), habitat continuity (area of tial scales would help clarify the role of rafts in dis- kelp cover per km along shore), and geographic distance (km). persal at the biogeographic scale where M. pyrifera 116 HEIDI L. HARGARTEN ET AL.
TABLE 2. Modeling genetic differentiation between populations of Pterygophora californica using the following predictors: Geographic distance (GeoDist), habitat continuity (HabCont), and minimum oceanographic transit time (TT) during months of spore production. The first rows show single predictor regression models for GeoDist, HabCont, and each of the TT for different periods. The following rows show multiple regression models for all combinations of predictors. For two predictor models, different combinations are shown using TT averages for different months. A single TT month is shown for the three predictor models. This was overall the model with highest goodness of fit. For each section, the best model is shown in bold. We show results for both FST- and DEST-based genetic differentiation. Goodness of fit metrics shown are the regression F test P value, the adjusted coefficient of determination (R2), and the Akaike information crite- rion (AIC).