The Condor 99~139-150 0 The Cooper Ornithological Society 1997

MOLECULAR VARIATION AND BIOGEOGRAPHY OF ROCK SHAGS1

DOUGLAS SIEGEL-CAUSEY School of Biological Sciences and University of Nebraska State Museum, Lincoln, NE 68588-0118, e-mail: [email protected]

Abstract. Molecular analysis of the presentgenetic structureof Rock Shags indicates significant population subdivision probably caused by vicariant disjunction associated with the Llanquihue Glaciation (35,00&15,000 ybp). The formerly continuouspopula- tion was forced into refugia on the Pacific and Atlantic coasts,where they remained with- out contactfor approximately 20,000 years. With amelioration of the climate and conse- quent glacial retreat, populationsrecolonized rocky shorelinesin the central portion of the present day range and introgressedconsiderably. The Chubdt and Falkland popula- tions serve as genetic sourcesfor the others, whereas the Fuegian population acts as a genetic sink. The population that is resident on Isla Chiloe is enigmatic and in nonequi- librium, possibly the result of indirect effects by a yet unsampled population. Key words: vicariance biogeography,population genetics,Fuego-Patagonia, glacia- tion, Phalacrocoracidae, Stictocarbomagellanicus.

INTRODUCTION period of similar magnitude occurred at the Pliocene-Pleistocene boundary, 1.2-l .O million Vicariance biogeographic models postulate that ybp. At maximum, the Llanquihue glacial allopatric remnants of a formerly continuous sheets covered the southern Andes from ap- population are created by the interposition of a proximately 40”s latitude to the Fuegian archi- barrier to gene flow (Platnick and Nelson 1978). pelago, and east to the Atlantic coast to about The most widely studied of these vicariance 52”s latitude. During this period, the Fuegian events have been the Pleistocene glaciations of coastline (i.e., Pacific coast of southern Chile, the Northern Hemisphere (e.g., Hoffmann Tierra de1 Fuego, and the Atlantic coast of 1976). In typically studied cases, the range of a southern Argentina) probably was uninhabitable species was disrupted by the advancing ice by coastal breeding animals, because formerly- sheet which displaced the population into refu- used shoreline habitats were subsequently cov- gia of varying sizes. Once confined to a glacial ered by ice or drastically altered by changing refugium, the species experienced novel condi- sea levels (Mercer 1976, Vuilleumier 1985). tions of habitat and species mix that presented Based on geological and fossil evidence, coastal a unique set of selection pressures, likely dif- refugia existed in the Falkland Islands (Islas ferent for each refugium (Hoffmann 1981). Gla- Malvinas) as well as the adjacent Burdwood cial retreat removed the geographical barrier Bank that was exposed by lowered sea levels, and created a natural ecological experiment in the Atlantic coast of northern Argentine Patago- which populations once isolated were then re- nia, and in regions north of Isla Chiloe, Chile introduced. How isolation by vicariance may (Vuilleumier 1971). Evidence to date indicates have played a role in the present genetic struc- that the fauna of patagonian Chile, patagonian ture of species is a problem critical to under- Argentina, and Tierra de1 Fuego (i.e., Fuego- standing the role of biogeography in speciation Patagonia) experienced changes as profound as events. those studied in the northern hemisphere (Vuil- For southern South America, the Llanquihue leumier 1971, 1985, Bdz and Scillato Yane Glaciation (35,00&15,000 years before present 1979, Simpson 1979, Fjeldsg 1985, Haffer [ybp]) was the only major glaciation event of 1985, Rasmussen 1987, 1991). In order to as- the Pleistocene (Mercer 1976); the next glacial sess the magnitude and effect of the Llanquihue Glaciation on coastal animals, I analyzed the ’ ’ Received 8 December 1995. Accepted 1 October patterns of intraspecific genetic variation in 1996. Rock Shags (Phalacrocoracidae: Stictocarbo

U391 140 DOUGLAS SIEGEL-CAUSEY

1936, Siegel-Causey 1986, Rasmussen 1987, Siegel-Causey and Bromley, unpubl. data). L These and other details of natural history and IN ecology suggest that the Llanquihue Glaciation had great impact upon the present genetic struc- ture of Rock Shags. There are several altema- tive hypotheses relating to a population-level response to a glacial vicariance event in south- em South America. The Llanquihue Glaciation could have forced birds onto (a) a Pacific coast refuge, (b) an Atlantic coast refuge, or (c) both Pacific and Atlantic refuges. The null hypoth- esis is (d) there were no refuges or dispersals associated with the Llanquihue Glaciation. Post-glaciation dispersal in hypotheses a, b, and c was bicoastal, because that is the present lim- its of distribution. Choosing among alternative hypotheses will require scrutiny of the gene LI 500 km flow patterns among the present populations of Rock Shags. For example, hypothesis a predicts FIGURE 1. Distribution of Stictocarbo magellani- that gene flow patterns will be unidirectional cus in the Fuego-Patagonian regions of Argentina and from Isla ChiloC through Tierra de1 Fuego to Chile in southern South America. Sampled popula- Chub&, Argentina, whereas hypothesis b pre- tions are: Chubht (stippling), Santa Cruz (diagonal dicts the opposite. Hypothesis c predicts that lines), Isla Malvinas (Falkland Islands) (light shad- ing), Tierra de1 Fuego (dark shading). The breeding gene flow patterns will be centripetal and fo- distribution of Pacific populations is largely un- cused on Tierra de1 Fuego. If there is no sig- known, except for Isla Chiloe, and putative limits are nificant pattern of gene flow, or no differences indicated by dotted lines. Sampled colonies are indi- among populations, then the null hypothesis cated by closed circles. will not be falsified. In this study, I examine the geographic variation in allozymic loci in Rock Shags collected throughout the breeding range magellanicus), an endemic seabird of Fuego- and I use various genetic measures to assessthe Patagonian marine littoral habitats (Fig. 1). magnitude and direction of putative gene flow Rock Shags are primarily cliff-nesting birds among present day populations. feeding in shallow onshore waters and are de- pendent upon suitable rocky coastlines for MATERIALS AND METHODS breeding, feeding, and roosting habitats (Mur- Tissue samples from 90 Rock Shags were col- phy 1936, Siegel-Causey 1986, Punta and Sara- lected at coastal sites in Fuego-Patagonian Ar- via 1993). The present distribution of Rock gentina and Chile from 1985 to 1987. Locali- Shags includes the region that matches pre- ties in Argentina included Puerto Melo (Chubut cisely the extent of the Llanquihue Glaciation Province), Puerto Deseado and Monte Leon plus Atlantic coastal Patagonia and the Falkland (Santa Cruz Province), and Ushuaia (Tierra de1 Islands; although Rock Shags are essentially a Fuego); in Chile, Challahue and Isla ChiloC marine bird, small freshwater populations have (Region X). Tissues from three Rock Shags been discovered on Lago Fagnano in Tierra de1 were obtained from the Falkland Islands in Fuego within flying distance of rocky marine 1984. Morphological and natural history data shores (Chebez and Gomez 1988). Rocky habi- (Siegel-Causey, unpubl. data), and preliminary tats are patchy throughout Fuego-Patagonia and genetic data using RFLP analysis (Bromley and their disappearance as a result of a growing ice Siegel-Causey, unpubl. data) indicated that sheet might extirpate Rock Shags from large birds collected from the two localities in Santa parts of its former range. Previous studies have Cruz Province were not distinct, so these were indicated regional differences in plumage, mor- analyzed together. Liver, pectoral muscle, and phology, and behavior, which suggests that heart muscle were collected, frozen in liquid ni- population subdivision is important (Murphy trogen within three hours of collection, and MOLECULARBIOGEOGRAPHYOFROCKSHAGS 141 transported by air on dry ice at liquid nitrogen e,, = Cnij - Eij) temperatures, and thereafter stored in an ultra- LJ v$ cold freezer at -70°C. Tissues were homog- enized by mortar and pestle at 4°C in a grind- where nij is the observed occurrence of a given ing buffer with a 1% NADP-mercaptoethanol genotype and E, is the expected number. An es- solution and then centrifuged at 6,000X g for 5 timate of the variance of eij can be calculated minutes. The supematant was stored at -70°C as follows (Everitt 1977): until used for electrophoresis. All supematants were analyzed within 90 days of preparation. vii= (1-E)(l -z) I surveyed 39 enzyme systems comprising 65 presumptive allozymic loci. All loci were exam- where ni+ is the row total, n+j the column to- ined for all individuals at least twice. The rep- tal, and n++ the grand total of observed values. licate runs were designed to detect possible With the estimate of the variance, vii, the stan- cryptic differences in mobility of alleles, poten- dardized residual can be transformed to a stan- tial post-translational modifications of allo- dard normal deviate with mean of zero and unit zymes caused by extended freezing and re- variance. The transformation yields an adjusted peated freeze-thaw cycles. Presumptive loci residual defined as (Ever&t 1977): were visualized using techniques and recipes summarized in Murphy et al. (1990). Electro- d, = -& phoreses were carried out on horizontal 10% starch gels or on vertical polyacrylamide slab These values are distributed as a standardized gels (Johnson 1976, 1979). Five enzyme sys- normal variant z and significance and sign of tems comprising seven allozyme loci (aldehyde deviation can be assessed directly (e.g., reductase, carbonic anhydrase, malate dehydro- I z I > 1.96, P < 0.05; I z I > 2.56, P < 0.01; genase, succinate dehydrogenase, and xanthine I z I > 3.29, P < 0.001). dehydrogenase) gave unreliable, unreplicated, I used Slatkins’ (1985) formula for estimat- or ambiguous results and were not used in the ing levels of gene flow, namely: analysis. Five allozyme loci (CK-2, EST-3, ICD-3, PGK-2, PGM-3) showed evidence of In@(l)) = a ln(Nm) + b extensive post-translational modification, prob- To estimate direction of gene flow, I used ably from repeated freeze-thaw cycles of super- Hedricks’ (1971, 1975) U index, which mea- natants, and also were coded as ambiguous and sures the probability and asymmetry of a unique not used in the analysis. For the remaining 53 genotype relative to its absence in another popu- loci, the mobility of the most common allelic lation. Hedricks’ U is calculated using the for- product for each locus was used as the refer- mula: ence and designated as allele a with a mobility of 100. Other alleles of a locus were labeled al- phabetically beginning with the most anodal; u,, = ; 3 $ P,, P,, = 0 p&=1 J=l mobilities were assessed as percent migration relative to the reference allele. where i is the locus and L, the number of I used the BIOSYS-1 software program polymorphic loci, j is the genotype and gi the (Swofford and Selander 1989) for all gen- number of genotypes at locus i, PIlk is the etic computations including allelic and geno- frequency of genotype j at locus i in locality k. typic frequencies, observed (direct count) and u km is the mean probability of drawing a expected heterozygosities, heterogeneity chi- genotype from region k that is not from region square values for each locus, chi-square tests for m. If both sets of genotypes are drawn from departures from Hardy-Weinberg genotypic ex- the same population, Ukm and U,, will be the pectations within each sample, and Wrights’ same; if not, then unique genotypes will be (1978) measures of population subdivision (F- distributed asymmetrically and U,, and U,,,, statistics). The significance of deviation from will differ. Derivative populations will there- Hardy-Weinberg equilibrium values was as- fore have the least probability for unique sessed using the standardized residual of indi- genotypes and ancestral populations will have vidual chi-square cells. For cell (ij), the stan- greater probability values (Hedrick 197 1, dardized residual is defined as 1975). The sign of the difference between two 142 DOUGLAS SIEGEL-CAUSEY

TABLE 1. Allele frequencies of polymorphic loci for Rock Shags.

Falkland Ushuaia Chiloe YJ$ Loci ’ Allele2 Mobility3 n=3 n = 18 n = 17

ADA-l Z 1.041.00 0 0.868 0.818 0.684 0.765 0.833 0.105 0.106 0 0 C 1.11 0 0.026 0.045 0.158 0 d 1.19 0.167 0 0.030 0.158 0.235 ALD- 1 : 11.06 .oo 1.000 0.974 0.970 1 .ooo 1.000 0 0.026 0.030 0 0 EST-2 1.000 1 .ooo 1 .ooo 0.861 0.882 : 1.001.17 0 0 0 0.139 0.118 G6P-1 : 0.941.00 0.333 0.972 0.864 0.722 0.800 0.667 0.028 0.136 0.194 0.067 c 1.11 0 0 0 0.083 0.133 GDA-1 : 1.051.00 0.667 1.000 0.939 0.778 0.647 0 0 0 0.139 0 0 0 0 0.083 0.353 : 1.171.21 0.333 0 0.061 0 0 GLO- 1 : 0.851 .oo 1.000 1 .ooo 1.000 0.944 0.794 0 0 0 0.056 0.206 GPD-1 1.000 0.895 0.894 0.357 0 : 1.001.03 0 0.105 0.106 0.464 0.853 C 1.11 0 0 0 0.179 0.147 GPI-1 : 11.09 .oo 0.333 0.842 0.848 0.917 0.765 0.667 0.158 0.152 0.056 0 C 0.94 0 0 0 0.028 0.235 ICD- 1 : 11.16 .oo 1.000 0.868 1.000 1.000 1.000 0 0.132 0 0 0 NP- 1 : 11.22 .oo 1.000 0.895 0.924 0.889 0.588 0 0.105 0.076 0.083 0, C 1.28 0 0 0 0.028 0.412 PEP-A Z 0.911.00 1 .ooo 1.000 1.000 0.944 0.882 0 0 0 0.056 0.118 PEP-B 1.000 1.000 1.000 0.972 0.912 : 1.001.13 0 0 0 0.028 0.088 PEP-C 0 0.842 0.879 0.139 0 : 1.001.11 1.00 0.158 0.211 0.639 0.618 C 1.19 0 0 0 0.222 0.382 PGM- 1 1.000 0.868 0.970 0.889 0.941 : 1.001.04 0 0.132 0.030 0.111 0.059

’ ’ Loci abbreviations follow standard usage (e.g., Mu h et al. 1990). ’ ’ Allele a is the most cmuncm form; others are liste Flya phabetically relative to mobility and frequency. 3 Mobilities are relative and calculated as per cent distance of allele a.

populations will indicate the direction of the 39 were monomorphic among all five popula- difference, or in other words, indicate which of tions (ACN-1, ACN-2, ACP- 1, ACP-2, ADA-2, the two is likely the source and which is likely AK- 1, ADH-1, ADH-2, ALD-2, CK- 1, ENO- 1, the recipient of gene flow (Hedrick 1971, pers. EST-l, FUM-M, FUM-S, GDH-1, GLU-1, comm.). GOT-l, GOT-2, GAPD-1, G3PD-1, GPT-1, GSR-1, HK-1, ICD-2, LDH-1, LDH-2, MPI-1, RESULTS MPI-2, NP-2, PEP-D, PEP-S, PFK-1, PGD-1, GENETIC VARIATION WITHIN PGK-1, PGM-2, SOD-l, SOD-2, SDH-1, TPI- POPULATION SAMPLES 1). The percentage of polymorphic loci (p), Of the 53 loci scored, 14 (26.4%) were poly- mean number of alleles per locus (A) and mean morphic in at least one population (Table 1) and observed (?i,) and expected (BJ heterozygos- MOLECULAR BIOGEOGRAPHY OF ROCK SHAGS 143

TABLE 2. Jolymorphic’ loci (p at 0.99 criterion), mean alleles per locus (A ? SE), pooled heterozygosities (Ho ? SE, He 2 SE), Wrights’ F,, and estimated gene flow Nm for five populationsof Rock Shags.

Population n B x (SE) % (SE) Fe GE) FU Nm

Falkland 3 8.3 1.08 (0.04) 0.019 (0.007) 0.030 (0.015) 0.250 - ’ Puerto Melo 19 17.0 1.19 (0.06) 0.019 (0.007) 0.033 (0.011) 0.340 1.60 Santa Cruz 33 15.1 1.19 (0.07) 0.021 (0.008) 0.029 (0.010) 0.221 1.54 Ushuaia 18 22.6 1.36 (0.01) 0.055 (0.016) 0.068 (0.021) 0.069 0.55 Chiloe 17 22.6 1.25 (0.07) 0.016 (0.006) 0.070 (0.019) 0.785 0.23

’ ’ Not available due to small sample size.

ity estimates are given in Table 2. p-values tion from the expected number of heterozygous ranged from 8.3% (Falkland Islands) to 22.6% genotypes per locus and ranged from -0.161 (Tierra de1 Fuego and Isla ChiloC); the average to 1.000. The average F, of the samples, cal- over all individuals was 17.0%. A-values ranged culated over all polymorphic loci, ranged from from 1.08 (Falkland Islands) to 1.36 (Tierra de1 0.069 to 0.785 (Table 2). Chi-square tests for Fuego); the average over loci and samples was conformity to Hardy-Weinberg equilibrium 1,.19. Populations differed significantly in (Table 3) revealed 11 cases out of 45 for which P-values (Kolmogorov-Smimov two-sample observed and expected distribution of geno- test, D = 4.61, df = 2, p < 0.05) but not in types were significantly different (P < 0.05). A-values (Kruskal-Wallis test, P > 0.10). Nine of these cases were from the Isla Chilo- p-values ranged from 0.016 (Isla ChiloC) to epopulation of Rock Shags. F, values for these 0.055 (Tierra de1 Fuego); the mean across all loci were high and positive, normally indicat- populations of Rock Shags was 0.039. Residual ing a deficiency of heterozygotes. Residual analysis of mean observed heterozygosity val- analysis (Table 4), however, revealed that sig- ues indicated that all populations maintained nificant excess of rare homozygotes was much similar levels of heterozygosity except for the more important (77% of all deviating genotypic Fuegian population, which had a greater than frequencies). These values imply that all popu- expected proportion of heterozygous genotypes lations except for the Isla ChiloC population are (z = 2.16, P = 0.031). close to Hardy-Weinberg equilibrium. To assess The inbreeding coefficient, F, (Wright 1978) the effect of results from the Isla ChiloC popu- was calculated for each polymorphic locus in lation, I did subsequent analyses with and with- each population. Values of F, measure devia- out their inclusion.

TABLE 3. Chi-square values (2), degrees of freedom (in parentheses),and levels of significance (cor- rected by strict Bonferroni criteria) for the Hardy-Weinberg equilibrium test of genetic equilibrium at each locus.

Falkland Pto Melo Santa cruz Ushuaia Chiloe Loci n=3 n = 19 n = 33 n = 18 n = 17

ADA- 1 0.01 (1) 1.03 (3) 5.22 (6) 4.28 (3) 1.06 (1) ALD- 1 M ’ 0.01 (1) 0.06 (1) M EST-2 M M :01 (1) 0.92 (1) G6P- 1 1.28 (1) OMOl (1) 2.38 (1) 0.05 (3) 7.81 (3)* GDA- 1 0.01 (1) M 0.10 (1) 2.61 (3) 14.90 (l)*** GLO-1 M M 0.01 (1) 8.90 (l)** GPD-1 M 1.09 (1) !21 (1) 5.32 (3) 6.75 (l)** GPI- 1 1.28 (1) 4.44 (l)* 1.49 (1) 0.01 (3) 14.34 (l)*** ICD- 1 M 7.66 (l)** M M NP-1 M 1.08 (1) 1.09 (1) To1 (3) 8.26 (l)** PEP-A M M M 0.01 (1) 12.39 (l)*** PEP-B M M 0.01 (1) 2.47 (1) PEP-C M 211 (1) 3.68 (1) 1.25 (3) 11.25 (l)*** PGM- 1 M 0.41 (1) 0.01 (1) 0.01 (1) 8.36 (I)**

’ ’ Monomorphic locus. * P < 0.05; ** P < 0.01; *** P < 0.001 144 DOUGLAS SIEGEL-CAUSEY

TABLE 4. Genotype frequencies (and adjusted residuals) for non-equilibrium loci in Rock Shags from three populations in southern South America.

Loci aa ab ac ad bb bc bd cc cd dd

Pto Melo ’

ADA 0.789 0.105 0.053 0.053 (0.19)2 (-0.83) (0.11) (2.08) (-i.33) (i.00) GPD 0.842 0.105 0.053 (0.22) (-0.87) (2.08) GPI 0.789 0.105 0.105 (0.43) (-1.4) (2.50) ICD 0.842 0.053 0.105 (0.12) (- 1.64) (3.33) NP 0.842 0.105 0.053 (0.22) (-0.87) (2.08) PGM 0.789 0.158 0.053 (0.19) (-0.69) (2.08)

Santa Cruz

ADA 0.697 0.091 0.091 0.061 0.061 (0.05) (-1.17) (0.32) (0.26) (2.95) (-i.57) (-i.46) (-i.21) (-z.30) (-0.12) NP 0.879 0.091 0.030 (0.16) (-0.78) (2.16) PEP-C 10.818 0.121 0.030 (0.31) (-1.18) (2.39)

Chiloe

G6P 0.647 0.059 0.059 (0.48) (-y.29) (-::::; (5.24) (-z.53) (1.74) GDA 0.647 0.353 (1.51) (&3) (2.83) GLO 0.765 0.059 0.176 (0.73) (-1.98) (2.96) GPD 0.824 0.059 0.118 (0.14) (-1.62) (3.08) GPI 0.765 0.235 (1.00) (4.51) (3.42) NP 0.529 0.118 0.353 (1.35) (-2.27) (1.95) PEP- 0.882 0 0.118 A (0.50) (- 1.91) (4.26) PEP-C 0.588 0.059 0.353 (1.44) (-2.53) (2.37) PGM 0.941 0 0.059 (0.25) (- 1.39) (5.60)

’ ’ All loci and genotypes at Ushuaia were in equilibrium; the small number of samples from the Falkland Islands precluded analysis. ’ Significance of genoty ic deviation was assessed using the adjusted residual of a contingency chi-square test on genotypic frequencies. Sign indicates direction o P dewatmn,” bold face indicates significant deviations (Izl > 1.96, P < 0.05; Izl > 2.56, P < 0.01; lzl > 3.29, P i 0.001). MOLECULAR BIOGEOGRAPHY OF ROCK SHAGS 145

TABLE 5. Mean fixation values over all populations for polymorphic loci. Loci showing significant among- sample heterogeneities in allelic frequency distribution are indicated by asterix (G-tests corrected using strict Bonferroni criteria).

All localities Excluding lsla Chilok

F,, F,, F,, F,,

ADA- 1 0.168 0.492 0.389*** 0.119 0.493 0.425*** ALD-1 -0.027 -0.005 0.021 -0.027 - 0.007 0.020 EST-2 0.110 0.188 0.082 -0.161 -0.036 0.108 G6P- 1 0.479 0.613 0.257** 0.422 0.580 0.273** GDA- 1 0.270 0.398 0.175* -0.089 0.075 0.151*** GLO-1 0.607 0.657 0.128 -0.059 -0.014 0.042 GPD-1 0.520 0.763 0.505*** 0.459 0.614 0.286*** GPI-1 0.694 0.776 0.268** 0.596 0.711 0.285 ICD-1 0.770 0.795 0.108 0.770 0.793 0.102* NP-1 0.472 0.573 0.192* 0.211 0.233 0.027 PEP-A 0.645 0.668 0.065 -0.059 -0.014 0.042 PEP-B 0.468 0.495 0.052 -0.029 -0.007 0.021 PEP-C 0.471 0.736 0.500*** 0.281 0.673 0.545*** PGM- 1 0.260 0.289 0.039 0.296 0.322 0.037 MEAN 0.429 0.610 0.317 0.283 0.504 0.307

* P < 0.05, **p < 0.01, ***p < 0.001

GENETIC VARIATION AMONG cus (Table 5). Population subdivision as mea- POPULATION SAMPLES sured by F,, (Wright 1978) ranged from 0.021 Three patterns predominated among continen- (ALD-1) to 0.505 (GPD-1). Across localities tal populations of Rock Shags (Table 1). For and loci, the significant heterozygosity chi- some alleles, fixation levels increased from the square (x292 = 628.3, P < 0.001) indicated the Pacific side to the Atlantic side of Fuego- existence of geographic subdivision. Seven loci Patagonia (e.g., ADA-l”, GDA-I”, GLO-l”, (ADA- 1, G6P- 1, GDA- 1, GPD-I , GPI-1, NP- 1, PEP-A”). For others, this pattern was reversed PEP-C) were more variable among populations and the level of fixation increased from the At- than they were within a given population of lantic side to the Pacific side (e.g., ADA-Id, Rock Shags. The mean of these F,, values was GPD-lb, PEP-C”). The third pattern was where 0.317, a very high value for birds. When birds allelic frequencies in the Fuegian population from the population resident on Isla ChiloC were differed significantly (Kolmogorov-Smimov excluded from this analysis, the mean F,, value two-sample test, P < 0.05) from neighboring (Table 5) was reduced slightly to 0.307, still populations (e.g., ADA-I”, GDA-l”, G6P-I”, well outside the limits expected for an equilib- GPI-1”). Rock Shags from the Falkland Islands rium population of birds (Barrowclough 1983). appeared fixed for all but four loci, but this is The pattern of fixation values among loci, how- likely an artifact of sample size. ever, remained similar to fixation values when Only two populations showed any evidence all populations were considered. Values were of unique autapomorphic alleles. Rock Shags reduced further when birds from Isla Chiloe and from Puerto Melo had in moderate frequency Falkland Islands were excluded from analysis, cf = 0.132) a fast migrating allele of ICD-1. A but the mean F,, still was high (0.135; data not moderate proportion (f = 0.139) of birds from shown). Tierra de1 Fuego were found to have an allele Analysis of the levels of gene flow indicated slightly cathodal (1.05) of the common allele of that the estimated number of immigrants per GDA-1. A much more common pattern seen in generation (Nm) received by the Atlantic coastal Rock Shags was synapomorphic alleles shared populations (Puerto Melo and Santa Cruz) was by adjacent populations (e.g., GDA-l”, NP-I”, approximately the same (Table 2). By contrast, PEP-Ab, PEP-C” for Tierra de1 Fuego and Isla Rock Shags resident in Tierra de1 Fuego and Isla ChiloC). ChiloC had much smaller levels of gene flow Overall genetic differentiation among popu- (e.g., 0.55 and 0.23, respectively). The general lations was calculated for each polymorphic lo- estimate of Nm considering all individuals was 146 DOUGLAS SIEGEL-CAUSEY

TABLE 6. Hedrick’s probability (rr,,) of a unique genotype averaged over 14 polymorphic loci for five localities of Stictocarbo magellanicus.

Mean Probability of Drawing a Genotype That Does Not Occur In: From: Pto Melo %a Cruz Ushuaia Chilo6 Falkland

Pto Melo - 0.021 0.03 1 0.066 0.316 Sta Cruz 0.048 0.033 0.061 0.368 Ushuaia 0.181 0.153 0.074 0.463 ChiloC 0.162 0.103 0.059 - 0.43 1 Falkland 0.333 0.048 0.333 0.333

1.12. Jackknifed estimates of Nm were made by from the Santa Cruz population had a probabil- serially removing samples and assessing the ef- ity of 0.368 of being absent from the Falkland fect; this process, however, did not reveal any Islands population; the reverse situation of a significant outliers. unique genotype found in the Falklands Islands Hedrick’s probabilities of a unique genotype population but not in the Santa Cruz population (U,,) were largest for Falkland Island birds was only 0.048. This indicates that the geno- (Table 6) and to a lesser extent, Puerto Melo; a types found in the Falkland Islands birds was a large value of U,, is characteristic of an ances- subset of those sampled from Santa Cruz, and tral region (Hedrick 1971, 1975). U,, also is that the Santa Cruz population comprised a useful in identifying regions likely to yield greater diversity of genotypes. genotypes not found elsewhere. For example, The difference between paired U-values (i.e., the probability of drawing a genotype from Isla u net = u,, - U,,) provides an estimate of the ChiloC that was not present in Puerto Melo was magnitude and direction of gene flow among 0.162, but was only 0.059 compared to Tierra populations (e.g., Hedrick 1971, 1975, Lubbers de1 Fuego. In other words, the probability that 1991). From the example above, the difference Rock Shags from Isla ChiloC and Puerto Melo in U-values between Santa Cruz and Falkland were genetically distinct was more than twice Island populations was +0.320 relative to Santa that between Isla ChiloC and Fuegian popula- Cruz, indicating a strong gene flow towards tions. By contrast, a unique genotype drawn Santa Cruz from the Falkland Islands. The difference in U-values between Puerto Melo and Tierra de1 Fuego relative to Puerto Melo was - 0.150, or a moderate gene flow from Puer- to Melo towards Tierra de1 Fuego. These results are shown graphically in Figure 2, where direc- tions and magnitudes are indicated by arrow. Two patterns are immediately obvious. First, some populations such as Puerto Melo act as a genetic source and some serve as a genetic sink such as Tierra de1 Fuego. Populations at these localities have U,,, values with the same sign. In other words, the arrows all point the same direction: either all out, indicating flow of unique genotypes away from the population (centrifugal movement), or all in, indicating the reverse (centripetal movement). Second, some FIGURE 2. Diagram of the putative gene flows populations including those resident in the Falk- (measured by Hedrick’s U-value) among populations land Islands, Santa Cruz, and Isla ChiloC serve of Rock Shags. Arrows represent magnitude of the both as source and recipient. In all cases, how- difference between paired U-values between popula- ever, the net probability of finding a genotype tions (thin lines, lU,,,l < 0.20; thick lines, lU,,,l > 0.30). Circles with values represent localities and at a given locality that is not found anywhere sum of all &values associated with the Rock Shag else is the sum of all U,,, values for that local- population. ity (see Fig. 2). Thus, the total net probability MOLECULARBIOGEOGRAPHYOFROCKSHAGS 147 of finding unique genotypes within the Falkland obtain when this population is excluded from Islands population that were shared among analysis. Mean F,, values over all loci and lo- other Rock Shag populations was -0.084, calities are quite high for this population, almost which indicates a centrifugal movement of an order of magnitude greater than the average genotypes primarily towards the Santa Cruz for all birds (viz., 31.79% vs. 3.5%). The esti- population. The total net probability of draw- mates of gene flow are low (between 0.55 and ing a unique genotype for the Fuegian popula- 1.60 immigrants per generation in equilibrium tion was +0.143, indicating strong centripetal populations), but it is yet unclear whether Nm flow. By contrast, the total net probability values as calculated using private alleles repre- for the Isla ChiloC population was near zero sent current or historical patterns, or more (U,,, = 0.025), suggesting that net exchange of likely, an integration of values over an unknown genotypes is static. It is tempting to apply sig- period of time (Slatkin 1985a, Zink et al. 1987). nificance values to Hedricks’ probabilities, but Nevertheless, the mean estimate of gene flow the underlying probability distribution and sta- among populations is low, only 1.12 immigrants tistical properties of U,, have yet to be deter- per generation. mined (Hedrick, pers. comm.), so quantitative The pattern of directional movement of geno- assessment of these results must await further types within the current range of Rock Shags study. (Fig. 2) reveals the strongly asymmetric nature of gene flow throughout Atlantic and Pacific coasts of southern South America. On the At- DISCUSSION lantic coast, genetic diversity is highest in the GENETIC DIVERSITY AMONG POPULATIONS northernmost populations near Puerto Melo and In this study of Rock Shags, the percentage of decreases southward towards Tierra de1 Fuego polymorphic loci (p), the mean number of al- and westward from the Falkland Islands to- leles per locus (A), and observed levels of mean wards the mainland. Moreover, high U,,,, val- heterozygosity (BO) are consistent with the re- ues here are indicative of ancestral populations sults of other allozyme surveys of non-passerine and support the notion that these historically birds (Barrowclough 1983). Observed heterozy- non-glaciated regions served as refugia during gosity levels for all but one population were the Llanquihue Glaciation. similar; the exception was found in birds resi- The patterns are obscured by the small dent in Tierra de1 Fuego. There, the proportion sample of individuals from the Falkland Islands, of heterozygotes in the population was nearly but evidence presented here suggests that it was three times that found in neighboring popula- a minor refugium. By contrast, the present-day tions and somewhat higher than the average het- northern range of Rock Shags in Chubiit Prov- erozygosity reported for most birds (H = 0.053, ince, Argentina, appears to have served as the Barrowclough 1983). In a related study on Im- primary Atlantic refuge during glacia$on. Al- perial Shags (Leucocarbo atriceps), which are lelic diversity (measured by A- and P-values) sympatric with Rock Shags, Rasmussen (1991, here matches mean species values (G-test, P > 1994) found a similar pattern, but of smaller 0.50), and genotypic diversity (measured by magnitude. For that species, the population on U,,,) indicates that Puerto Melo (and presum- Tierra de1 Fuego had between 1.5 and 2 times ably coastal Chub&) served as a primary the heterozygosity found in neighboring popu- genetic source pool for the population complex. lations. F, values are high, something normally associ- Further evidence suggests that the Rock ated with heterozygosity deficiency (Crow and Shags analyzed here were not drawn from a Kimura 1970, Kimura and Ohta 1971). Residual single panmictic population, The significant de- analysis of genotypes, however, indicates that viations from genotype frequencies expected F,, values were affected instead by excess of under Hardy-Weinberg equilibrium for seven rare homozygotes (e.g., ADA-lbb, ICD-lbb, and loci, the highly significant chi-square values, PGM-lbb). At moderate levels, this is expected and the high F,, values all indicate strong ge- in avian populations (Barrowclough et al. 1985) netic structuring within this species. The non- and it further suggests that the Puerto Melo equilibrium status of the Isla ChiloC population population has always been large. clearly affects these findings, but similar results Populations resident in Santa Cruz and Tierra 148 DOUGLAS SIEGEL-CAUSEY de1 Fuego show every evidence that they were abundant as in any other locality (Humphrey, recently colonized from regions to the north and pers. comm.). Furthermore, I found no evidence east. U,,, values are positive and U,, values to suggest that assortative or disassortative mat- show significant centrifugal exchange of geno- ing was important at any locality. Intensive types from Puerto Melo and Falkland Islands study of courtship behavior suggested that al- populations. Furthermore, it appears that the though there were slight differences in form and population currently resident in Tierra de1 Fu- sequence of displays between Atlantic coastal ego has a different genetic population structure populations, sexes did not assort by morpho- than Rock Shags in Santa Cruz. Allelic diver- logical types (Siegel-Causey 1986). Migration sity is significantly elevated and observed het- by individuals from adjacent populations is less erozygosity levels are higher here than any- easy to rule out, although the associated values where else; F, values are low and all genotype of Slatkin’s Nm value suggest that such migra- frequencies match expected values. These find- tion is low (Table 2). ings plus the pattern of U,,,, values strongly sug- Residual analysis of genotype frequencies gest that the Fuegian population has received within the ChiloC population (Table 4), revealed genetic material from northern glacial refugia a significant underrepresentation of heterozy- on both coasts. gotes common in the range of the species (e.g., GDA”“, GPIab), but a more frequent occurrence GENETIC DISEQUILIBRIUM WITHIN of unique, rare genotypes (e.g., GPD”“, PEP- PACIFIC POPULATIONS Cbc). Moreover, Rasmussen (1987) identified The status of the population on Isla Chiloe is several juveniles with white abdomens (i.e., de- unclear and the genetic structure here is enig- rived plumage states) from Chile that are con- matic. Allelic diversity is high. F, is very high, sistent with autapomorphic origin in isolation. but heterozygosity levels and genotypic diver- Intermediate states were noted in birds collected sity are low. Heterozygote deficiency is bal- in Chile and from Tierra de1 Fuego, but none anced by an excess of homozygotes, and simi- were identified from Atlantic populations. lar to the Falkland Island population, Rock These data suggest either low level introgres- Shags from Isla ChiloC lost several allelic forms sion in Isla ChiloC by an unsampled population found otherwise in this species. The finding that (Wahlund effect), or that birds resident on Isla this population is in Hardy-Weinberg disequi- ChiloC were recently established and represent librium for some polymorphic loci clouds easy the expanding front of a population located else- interpretation of these results. This unexpected where. finding of nonequilibrium is difficult to recon- Considering that the distance separating Isla cile given that all other populations of Rock Chiloe and Ushuaia is roughly the length of the Shags are genetically stable. Although I am un- west coast of the United States, the existence able to determine the nature of the disequilib- of breeding colonies between them is not un- rium at this time, clues to possible destablizing likely. If the Rock Shags collected on Isla forces may be found in examining how Hardy- ChiloC are not representative of a discrete Weinberg equilibrium might be violated. breeding population, but instead are at the mar- Of the eight main assumptions of the Hardy- gins of an expanding colonization front, or a Weinberg model (Wright 1978, Hart1 1988), zone of introgression, then the findings reported possibly three were not met by birds breeding here are consistent with a population in non- on Isla Chiloe. The population size may have equilibrium admixture (Maruyama and Fuerst been smaller than at other localities which al- 1984, 1985). If so, then the present population lowed a disproportionate effect from individual genetic structure of Rock Shags in Fuego- genotypes, mating may have been nonrandom Patagonia reveals not only evidence of a past and individuals were assortatively (or dissorta- response to a widespread biogeographic vicari- tively) pairing, or migration may have played a ante event, but suggests in addition that the dy- greater role in this population than elsewhere. namics of change have continued. Accurate estimates of population size, or for that matter most other parameters of basic biol- ACKNOWLEDGMENTS ogy, are unavailable for this species; however, I thank P. S. Humphrey and P. C. Rasmussenfor as- Rock Shags on Isla Chiloe appeared to be as sistance in the field work. I benefited greatly from MOLECULAR BIOGEOGRAPHY OF ROCK SHAGS 149

comments by B. C. Livezey, P. C. Rasmussen, F. HUMPHREY, P. S., D. BRIDGE, P. W. REYNOLDS, AND Vuilleumier, and several anonymous reviewers. Per- R. T. PETERSON. 1970. Birds of Isla Grande mits were issued by many agencies in Argentina, (Tierra de1 Fuego). Preliminary Smithson- Chile, and the United States. Assistance of many ian Manual. Smithson. Inst., Washing- kinds was given by the following: in Argentina, by ton, DC. B. Maver and F. V. Faurina of Pataaonia Comercial JOHNSON, G. B. 1976. Hidden alleles at the a- S. R. L.; J. Navas, J. M. Gillardo, aid R. A. Bockel, glycerophosphate locus in Colias butterflies. Ge- Museo Argentino de Ciencias Naturales, Buenos netics 83:149-167. Aires; E. I. Portela, Prefectura Naval Argentina, Puer- JOHNSON,G. B. 1979. Increasing the resolution of to Deseado; J. C. Victoria, Subprefectura Naval Ar- PAGE by varying the degree of cross-linking. gentina, Puerto Santa Cruz; and in Chile, by H. Nie- Biochem. Genet. 17:499-516. meyer and colleagues of the Museo National de His- KIMURA, M., AND T. OHTA. 197 1. Theoretical aspects toria Natural, Santiago; M. Sallaberry A., T. G. Iri- of population genetics. Princeton Univ. Press, arte, and A. M. Veloso, Universidad de Chile; Princeton, NJ. J. E. Llaiquen, Pargua, and 0. Cardena de Soto, An- LUBBERS,E. L., K. S. GILL, T. S. Cox, AND B. S. GILL. cud. I greatly appreciate Polly Denham’s efforts and 1991. Variation of molecular markers among enthusiasm while preparing the map. Financial sup- geographically diverse accessions of Triticum port came from grants-to the author from the KU Mu- tunschii. Genome 34:354-361. seum of Natural History, and NSF grant BSR MARUYAMA, T., AND F.’ A. FUERST. 1984. Population 84-07365. bottlenecks and nonequilibrium models in popu- lation genetics. I. Allele numbers when popula- tions evolve from zero variability. Genetics LITERATURE CITED 108:745-763. BA~z, A. M., AND G. J. SCILLATO YANE. 1979. Late MARUYAMA, T., AND P. A. FUERST. 1985. Population Cenozoic environmental changes in temperate bottlenecks and nonequilibrium models in popu- Argentina, p. 141-156. In W. E. Duellman [ed.], lation genetics. II. Number of alleles in a small The South American herpetofauna. Univ. Kan- population that was formed by a recent bottle- sas Mus. Nat. Hist., Lawrence, KS. neck. Genetics 111:675-689. BARROWCLOUGH,G. F. 1983. Biochemical studies of MERCER,J. H. 1976. Glacial history of southernmost microevolutionary studies, p. 223-261. In A. H. South America. Quat. Res. 6: 125-166. Brush and G. A. Clark, Jr. [eds.], Perspectives MURPHY, R. W., J. W. SITES, JR., D. G. BUTH, AND in ornithology. Harvard Univ. Press, Cambridge, C. H. HAUFIXER. 1990. Isozyme electrophoresis, MA. p. 45-126. In D. M. Hillis and C. Moritz [eds.], CHEBEZ, J. C., AND D. G~MEZ. 1988. Notas zoo- Molecular systematics. Sinauer, Sunderland, geograficas sobre aves de Tierra de1 Fuego. El MA. Homer0 13:75-78. MURPHY, R. C. 1936. Oceanic birds of South CROW. J. F.. AND M. KIMURA. 1970. An introduction America. 2 volumes. Am. Mus. Nat. Hist., New to population genetics theory. Burgess Publica- York. tions, Minneapolis, MN. PLATNICK,N. I., AND G. NELSON. 1978. A method of EVERI~T, B. S. 1977. The analysis of contingency analysis for historical biogeography. Syst. Zool. tables. Chapman and Hall, London. 27:1-16. FJELDS~,J. 1985. Origin, evolution, and status of the PUNTA, G., AND J. SARAVIA. 1993. Distribution, abun- avifauna of Andean wetlands. Omithol. Monogr. dancia y aspectos de la biologia reproductiva de1 36:85-112. cormoran cue110 negro (Phalacrocorux magel- HAFFER,J. 1985. Avian zoogeography of the Neotrop- lanicus) en la provincia de1 Chublit, Argentina. ical lowlands. Omithol. Monogr. 36:113-146. Homer0 13:295-298. HARTL, D. L. 1988. A primer of population genet- RASMUSSEN,P. C. 1987. Molts of the Rock Shag and ics, 2nd ed. Sinauer, Sunderland, MA. new interpretations of the plumage sequence. HEDRICK, P. W. 1971. A new approach to measuring Condor 89:760-766. genetic similarity. Evolution 25:276-280. RASMUSSEN,P. C. 1991. Relationships between coast- HEDRICK, P. W. 1975. Genetic similarity and dis- al South American King and Blue-eyed Shags. tance: comments and comparisons. Evolution Condor 93:825-839. 29:362-366. SIEGEL-CAUSEY,D. 1986. The behaviour and affini- HOFFMANN,R. S. 1976. An ecological and zoogeo- ties of the Magellanic Cormorant (Phalucuo- graphical analysis of animal mygration acrosi the corax magellani&s). Notornis 33x249-257. Bering Land Bridge during the Quaternary Pe- SIEGEL-CAUSEY,D. 1988. Phylogeny of the Phalacro- riod, p. 100-109. In V. L. Kontromavichus [ed.], coracidae. Condor 90:885-905. Beringia in the Cenozoic. Nauka, St. Petersburg, SIMPSON,B. S. 1979. Quatemary biogeography of Russia. the high montane regions of South America, p. HOFFMANN,R. S. 1981. Different voles for different 141-156. In W. E. Duellman [ed.], The South holes: environmental restrictions on refugial sur- American herpetofauna. Univ. Kansas Mus. Nat. viva1 of mammals, p. 25a5. In G. G. E. Scud- Hist., Lawrence, KS. der and J. L. Reveal [eds.], Evolution today. SLATKIN, M. 1985a. Gene flow in natural popula- Univ. Chicago Press, Chicago. tions. Annu. Rev. Ecol. Syst. 16:393430. 150 DOUGLAS SIEGEL-CAUSEY

SLATKIN, M. 1985b. Rare alleles as indicators of VUILLEUMIER, F. 1985. The forest birds of Patagonia: gene flow. Evolution 39:53-65. ecology, geography, speciation, endemism, and SWOFFORD. D. L.. AND R. B. SELANDER. 1989. fauna1 historv. Omithol. Monoer. 36:255-304. BIOSYS-1. A computer program for the analy- WRIGHT, S. 197i3. Evolution and- the genetics of sis of allelic variation in population genetics and natural populations. Univ. Chicago Press, Chi- biochemical systematics. Illinois Nat. Hist. cage. Surv., Champaign, IL. ZINK, R. M., D. F. LOTT, AND D. W. ANDERSON. 1987. VUILLEUMIER,B. s. 1971. Pleistocene changes in the Genetic variation, population structure, and evo- fauna and flora of South America. Science lution of California Quail. Condor 89:395405. 173:771-780.