Genetic variation in an endangered cedar (Widdringtonia cedarbergensis) versus two congeneric species

Widdringlonia cedarbergensis is a southern African conifer species under threat at extinction. This study explored the genetic status of the species to assess whether loss of genetic variation may be contributing to population declines. Widdringtonia nodiflora and W schwarzii were used as benchmarks against which to compare genetic diversity in W cedarbergensis. Isozyme electrophoresis was used to resolve seventeen isozyme loci in seedlings of the three species. Genetic diversity was greatest in W nodiffora, followed by W cedarbergensis and W schwarzii. There is no evidence that W cedarbergensis has undergone a genetic bottleneck relative to its sister species. Patterns of genetic variation varied between species with most of the variation occurring within populations of W cedarbergensis, between populations of W schwarzii, and within population 'neighbourhoods' of the more widespread W nodiflora. The isozyme data indicate inbreeding, probably due to self~pollination, in W cedarbergensis and W nodiflora. Populations of W schwarz;; were outbreeding. Selting in W cedarbergensis may be caused by a change in population density from dense to sparse stands with potentially deleterious genetic consequences.


Introduction
Estimating levels and patterns o f genet ic diversity within thre:llenell spedes has bccoml.! an important aspect of conservation biology. As sp(:c ies beco me increasingly frag men ted , sub-populations beco me smaller ami more isolated and as a result. begin to face demographi c and genetic ri sks (Barrett & Ka hn 1991 ;Lacy 1992) resu lting in a loss o f genetic variati on. A high level of genL!tic variation can provide insurance agai nst ex tincti on, while a low leve l. someti mes caused by population boltlenccks, is likely to lead [0 a greater risk of extinction. InbreL:d ing depressio n in smull , isolated populations can result in loss of fitness which. in lurn, influences popu lati on demography (Ellstrand & Elam 1991). The theorelical imparlance of genetic princip les in conse rvati on has been widely accepted. However empirical evi dence for the import ance of genetics in conserving species, especially plant species, is still very limited (You ng et al. 1996). We studied the possib le importance of ge netic processes in Ihe conservation of the L!ndangered conifer, WiddrillRlOllia cedar-hergclIsis.
WiddriJlMtoflifl cedarbergcllsis (the Clanwilliam cedar) is a tree species endemic to the Ce(krberg in the western Cape. The speci es is r~lIcd as e ndangered in the most recent South Afri can Red Data Book (H ilt on-Taylor 1996). Though 'he cedar OCC lirs in a landscape dom in ared by tire-prone fynbos shrublands, it see ms poorly adapt!.!d to prevailing fire regimes. Mature trees are easily killed by tire, there is no canopy-stored secdbank, sapling growth rates are relatively slow and the trees take longer th an any co-occurring fy nbos shrub to reach malUrity (Manders 1985). This set of traits has led to a marked decline in the population in the prevailing fynbos fin.! regime (Manders 1986;Brown et at. 1991). Special tire managemen t procedures have had to be devised for the cedar in a 'cedar reserve'. A replanting scheme has also been initiated us ing seeds which have been col lected predominantly fro m a cedar plantation (Mustarl et al. 1995).
One explan ation for thl! apparent poor adaptation o f cedars to contem porary lire regimes may be tbat it has passed through a population bottleneck resu lt ing in a lack of genetic variation on whi ch natural sL! lect ion might act. Anot her possibility is that the popu lation structure has ~h anged fro m a rdati ve ly dense fire -proof woodland to the current structure of scattered trees in a matri x of Fynbos. The species is thought to have been heavi ly utili sed by intensive logg in g in t~';! last two centuri es and thi s may have contri buted to it s current Fragmented population structu r(: (Ma nders 1986).
We studied genetic variation anu stru~turc in W cedarbergensis, (I) to te st whet her it had parti cularly low levels of geneti c vari ation re latiw to two congeneric species, W llodij7om and W schwarzii ind icative or a geneti c bott lened ; (2) to de termine pattern s o f genetic variation in the three specil!s in relati on to their life history aurihutes and 3) to ddefln ine the genetic relationships of the Cederberg populations and their bearing on selection of seeds for the replanting programme.
W cedarbergclIsis, W llodij7om and W sc/nvarzii, differ in their bio logy and distribution. These di fferences are ou tlined in Table I. Differences in reproductive biology and li fe hi story attributes are known to inlluem.:e patterns of genetic variation (e.g. Hamrick & Godt 1990). We therefore expected differe nt popu lation genetic structures regard less of possi ble bottleneck effects. If W. cedC/rhcrgclIsis is ex peri encing a bott leneck effec t, then It!vel s of genetic diversi ty should be lower than those of the ot her two spe~i es.
All three species arc wind-pollinated and shou ld all . therdore, be panmictic to some extent. Gene flow could po ss ibly differ among species hecause of their different seed di spersal syndromes, ti re survival strategies and geographic distribution.
W cedarhergellsis:-Gene flow betwee n populations may be reduced hecause of th e very poorly dispersed heavy seeds (Manders 1987 ). W. cedflrbergellsis is a geograph icall y rest ricted species and therefore relatively low levels of polymorphism were expected . A general tende ncy has been s hown for widespread species to hav!.! a hi gher ut.!gree of geneti c polymorphism than narrowly restricted species (Karron et al. 1988), although ot her W I/odi/lom: We. ex pected gene fl ow bctween populations to he high since. seeds arc light and winged, e nabling long-distance dispersal. However. W Ilodiflora is distributed along most mountain ranges up the east coast to central Africa as far as Malawi and distum.:es between these mountain ranges cou ld present a barrier to gene flow. We therefore expected patterns of population di fferentiation to be influenced by these distances. Since it is a widespread species and d isplays a high deg ree of morph o lo g i~ cal variati o n (Marsh 1966;Pauw 1992), we expected high levels of genetic differentiation among d ifferent populations in W lIodi~ flom . W Ilodif/or(l resprouts after fire and therefore adults persist for m any generation s. This shou ld lead to an accumul ation of heterozygotcs wit hin popu lati ons. Resprouting shoul d affect population di fferentiation in that the establishment of individuals and popul:ltions would be a rare evcnt since the prohability of a new cohort o f seedling s being eliminated by lire is high .
W schwarzii: Gene now between populations is pote ntiall y high owi ng to l o n g~d ista n cc di spersal of light, winged seeds. Outcross ing between popu lations could be hindered by the co n~ linement o r populations to deep, narrow, relatively tire-proof kloofs in the Baviaansk loof and Kouga mountains. We therefore expected relativc ly high leve ls of genetic differentiation amo ng populations since they arc di sc rete and wide ly separated by int e r~ vening ri dges. W schwlIrzi; is a lso a geograp hica ll y restricted species and we therefore expected relatively low leve ls of genetic di ve rsity.
W. cec/lIrbergensis and W. schwarzii should show similar levels of genetic divers ity. notwithstanding the effects of recent rrag~ mentation. s incc both have the sa me limited d istribution and s im~ il ar bio logies barring differences in seed dispersa l.

Materials and methods Sampling
Populations of W crdlllbergellsis. W IIodiflora and W sdll,-'orzii were visi ted and sa mpled for cones. Popu lat ions from Ill ost of the mnge of W cl'darbergclIsis was sampled. W sc/nvnrzii grows on cliff faces and stCl!p ravines and, because or diffi cu lties of acccss. only three popUlations were sampled. W Ilod~flora is a widespread species. Sampling was concentrated on popUlations occurring in the Cape Floristic Region. whe re condi tions are most similar to its congeners. but a single popu lat ion from the Natal Drakensbcrg was also included.  Figure  1). The plantation of W a darbergensis used as the main seed source for the replant ing programml.! is denoted as MS , not to bl.! con fu sed with an additi onal W cedarbergellsis pl antation, KD . Four to live cones from 30 to 40 randomly positioned trl.!es per Knuga population were sampled. Cones were ove n dried at 6(J°C to release the seeds which were then germinated in a germination chamha wi th tem peratu res alte rnating hetween 10°C and 20°C on a 12 hour cycle. The germlings were allowed to develop unt il the secondary shoO[ bud wa~ detectable. Isozymes for one seedling per tree wcre analysed. The seedlings were then crushcd in a chillcd mortar and pestle using vegetative ex traction buffcr I from Cheliak and Pitel (1 984). Five filter pape r wicks. which were cut to the size of J mm x 12 mm from no. 4 Whatrna nn tilter pape r. were satu rated in the extract and placed in an Eppindorll lUbe which was directl y trans~ ferred to a _20°C frt!ezer. The isozymcs were separated 011 12% starch gels which were 10 mm thi ck usi ng the enzyme/gel bulTer combinations outli ned in Conkle e/ al. (19R2). Before each gel run . the wicks were inserted betwc~n the anodal <. ! nu cathoda l sect ions or the gels. The gel s were left to run for 5 hours iJI 60 mA and the n sliced horizontally 2 nun thick. Th e followi ng cnzyme systems were stained using staining procedures outl ined in Conkle el af. (1982): malate dehydrogenase (Mdh), shikimic acid dehydrogenase (Sd h). isocitrate dehydrogenase (Idh) , glucose-6-phusphaic dehydrogenuse (G6pdh). glutamate dehydrogenase (GcJh). su perox ide di smut ase (Sud), dia~ phorose (Dia) . aspartote aminotransferasc (A at), acid phosphatase (Aph), menadione reduc tase (Mnr), peroxidase (Per), malic ~nzyme (Me), phosphoglucose isomerase (Pgi), phosphoglucomutase (Pgm), leucine amino peptidase (Lap). (:(~esterase (a~Est). f3-eslerase C P-Es l), nuorescent estl!rase (Fiest) and aconitase (A eon).
The gels were interpreted and scored. Loci were labelled in ascending orcJer from I to 3 from fas test (m igrating furth est in the gel towards the cathodal end) to slowest (migrating least in the gel towards the cathodal end). Alleles were lahelled from A to Z in ascending order from fastest to slowest.

Data analysis
For each population for which allozyme data were collected, genolype arrays were analy sed using the Biosys-l package (Swofford & Sdander 1989). Allt! le frequencies were calculated and used in con~ juncti on with the genotype data to calculate mean geneti c d iversity estimates including: number o r alldes per locus (A). % o r po l ymor~ phic loci (P) , mean proport ion ohservcl..i heterozygosity ( H,, ) and mean proportion expected panmictic heterozygosity {HJ. Levels of inbreeding within popu lations were estimated through co mparing H" and H~. Hl is an expected genotypic frequency calculated by using a binom ial ex pans ion o r the allele frequ encies. This binom ial expan~ sian is call ed the H ard y~Weinberg principle (Nei 1978). If observed levels of heterozygosity conform with expected levels of heterozy~ gosity. the popUlation is said to he in Hardy -Weinberg equilibriu m. The inbreeding coeflicienr, F. can be used to compare levels of inbreed ing between sUh populations. F is de tined as the probability that two olle les at a gene tic locus in the inbred indiv idu al are descended from a si ngh! gene in a single ancestor shared by the parents (Wright 19(9) <lnd is characterised by the formula F = (H~ -H,,)/ H~ where H~ is the expected heterozygosity at a locus and H" is the observed heterozygosity at a locus.
The levels and di st ri bution of gl!net ic d ive rsity we re ca lculated using Wri ght's tixa tion indices (Wright 1965) These stat islics were used to describe threl! levels of genetic interact ion. The basic formula used in Biosys-l is: the level of genetic interact ion between individua ls withi n the sa me derne or tree cluster, A positive F l , is associated wil h deficiencies of heterozygotes and suggests inbreeding wh il e a negative F,_ , suggests too many ht:lerozygote s relative to HiJrdy·Weinberg equilibrium (Linhart et al. 198 1). F"l represents the correlation between random game te~ within a g iven deme relative to gametes wi thin the whole populatiun. This value is used to determine the amount of differenti· ation between sub-popu lations and is an indirect measure of gene 11 0w (hi gh di fferen tiatio n among sub· popu lat ions implies low ge ne tlow). Fit represents the correlation between uniti ng gametes, and Iherdorc the fixation index:, with in the whole popuhltion.
Genetic distance and cluster analYSis Nei 's ( 1978) genetic identity (I ) was used to describe Ihe ex tent of gene tic relationships between populations of each s pecies. Pairwise 135 30km com parisons we re made and the identity values (I ) obtai ned and translated into a UPGMA (unweighll!u pair group method with arithmetic averaging) cluster analysis. The UPGMA algorithm used in Biosys· 1 is describl!d in Sneath & Soka l (1971). Mantel 's /-lest was used with the aid of NTSYS · pc (Rohlf 1993), \0 compare cophe~ netic genetic idenlilY and geographic dist:ln(:e matrices.

Gene frequencies
A ll ele frequencies for 19 enzyme loci Wt're calculated. Five of lhese loci wen~ found to he pol ymorphit.: in W. ccdarbergellsis (Table 2a), seven were found to Ql! polymorphic fo r w: Ilodiflora (Table 2b ) amI four were found to bl:! polymo rphic fo r ~v. schworzii (Table 2c). AlIeh.: frequ encies among diffe re nt populations or W     Mean observed and expec ted levels of heterozygosity varied among populations of alllhree species although on ly popUlations of HI: schwarzi; conformed to Hartly· Weinberg equilihri u m (Table 3) In W cedarbl.'rgellsis. the most inbred populations seem to hI.! WB and SB where tixation indi ces were fai rly high. The plantation. MB, was the only 'population' which seems (0 be outbrced ing. KK wou ld also appea r to be ollthred although this result may be a function of sampling e rro r si nce KK is monomorphic for an additional two loci. In W lJodij7ora, the most inbred popUlations seem to he OK, BKR and SBR where fixation indices \Vcre particularly high. BKY was Ihe onl y population with <llow fixation ind!..!x.

F-statistics
F j , values arc comparabk to the tixalio ll index and measure thl.! degree of st: lting. F r , was much lower in HI. schwarzii than in the other two species (Tabl!. . ! 4). F j , was high in W ced{/rbe l:~ellsis, but especially high in HI: IlOciijlo1'(l . Sim ilar trends can be sec n for Fit and F.'!. The F,( for W IlOd{flonl is so hi gh that it approaches values for selling plants (G'l = 0.5 10 in Hamrick & Godt 1990) implying ext remely low levels of gl! nc ll ow hetween populatio ns. HI: sc/llt'{1rzii was the only species within the genus that has an F'l va lue close to that found for most gymnosperms which ge nt!rally have high levels of gene flow and li llIe population dilTert!llliation (mean G,\ for gymnosperms = 0. 068. Hamrick & Godt 1990) .

Genetic distance and cluster analysis
The results of the pairwise comparisons of genetic distance bl!tweell popUlations using Nei's unbiasl.!d genetic dista nce are presented graph ically using a UPGMA duster analysis ( Figure  2). Mantel 's Hcst shows then:: was no relationship betwct!1l geographic prox imity of pop ulati ons (sec Figure I) and geneti c  In W. schwarzii, a ll three popu lati ons were genet icall y w ry sim ilar alt ho ugh NK and OK for m a sepa rate cluste r. With suc h a sma ll sample size it was impossi ble to test gcnl!t ic Cl nd gl!Og raph ic di stance.s for statisti cal sig ni fca nce.

Levels of genetic polymorphism
Narrowly emlcmil: species arc gene rally known to he g!.!nctically depaupl!rate. This has been shown in case studies for plants such as thi! narrow ly ende mic Bel/sol/iella oregoll(/ (Saxifragaceae) (So lti s el af. 1992), and the rare ElIca !.vpflIs pJllven:nlllla (Pl! ters el al. 1990). Two extreme endemics, Pedicul(/ris jilrbishae (Walh!r c:t al. 1987) andPinus torreyaJl(l (Ledig &Conkle 1983) have heen shown to have no variation at the loci examined (a lthough one population of Pinlls torreyalla was po lymorphic at ).4% of lhe loci). Conversely, widespread species are. genera ll y S. Afr. J. Hot. 1997. 61Cl) known to he more variable than their lIlore ende mic congeners (Lovell!ss & Hamrick 1984: Ham ri ck & God t 1990Karre n et aJ. 1988). This genera l trend holds true for Widdril1glOlIia. W Ilodiflora was the most widespread spec ies of all and has lhl! highest allel ic di versi ty and genetic po lymorphism. T he t1itlerenccs in h!vels of genetic t1iversity betwee n the two geographically restrictetl specics wcre slight. Th is suggl!sts that the poor adaptation of the Clanwilliam cedar to current ecological cond itions cann ot he attributed to lack o r gl!nctic variation due to a genetic hottlencck. Al though W cedarhergel1sis may have suffered severe recent reductions in numbers, genctic crosion would be delayed by residual heterozygosilY in adult trl!es, many of which havl! survived for centuries. In compariso n with other narrowly n!stricted conifers. W Ct't/arhergellsis and W. Jchwtlr;.ii certainly have higher levels of polymorphism and a lleli c diversi ty than has been found for the rare PillllS torreyww (Lt::di g & Conk le 19R1).
There is suffic ien t evi de nce to suggest that these two co mponents or d ivl!rsity: po lymorph ism and a ll e lic d ive rsi ty: arc morc re liable ind icaw rs of the e ffects of populatio n bou le necks than arc lIleall ohserved and c.!xpl!ctcd proport ions of heterozygosi ty. This has heen shown mathcl11,tti ca lI y hy Nd et 01. (1975). in computei' sim ulations by Lacy ( 19X7) anti by Lc. ! berg ( 1992Lc. ! berg ( . 1993 mosqu itolish which has a generation time of 56 days (Lcberg 1993). H owever, average heterozygosit y depends not only 011 the size of the bottlenec k hut also on the rale of population growth as the population recovers (Nei er ai, 1975). Average nurnbcr or alle les per locus, on the other hand , is profou ndl y affected by hottlenL:ck size but not so much by population growth (Nei et af. 1(75). The tind ing that W /lodij1ora, the most abund am and widespread species within the genus has the lowest h.:vels of heterozygos ity, suggests that, in thi s case at least. heterozygosity is an unrdiahle indicator of the effec ts of bottle nel:k s in a hel wee n-species compari son.

Organisation of genetic variation
Palterns of ge netic variation in the th ree species of WiJdringtoll;a were intiil:Hted by Wright's F·sraristics. observed and expected hete rozygosi ty. and by patterns of ge netic re latedn ess hetween populations. Hi gh F;, values and heterozygote deticiem:ies in \¥. cedflrbergel1sis and W nodif/()m suggest the possihility of eit her inh reeding or the Wahlund effect due to high levels of population sub·structuring. Given the biology and ecology of thl!se two spec ies. these two fa ctors need not mutually exclude each 01 her in determining patlerns of sub-structuring within popUlations.
The relatively high It!vels of inbreeding in the C lanwilliam cedar Illay he due to se lf pollination. Lack of pollen moveme nt between trees may be caused by popul ation level fragmentation with the majority of trees restricted to rocky outcrops in a sea of tlammabl e fy nhos and with large gaps between trees. Evidence for lowe red outcrossing rates have been found in low density stands of ponderosa pine, compared with high dens ity stands, as a result of inefficicnt pollen movement (Farris & Millon 1984 flora had a mctapopulati on that was structured into population neighhourhoods where popul ations from th e same mou ntain range wert' highl y rdated to popu latio ns from distant mountain ran ges. W cedarbagellsis, on the othcr hand. had an un structured melapop ulati on where neighbouring popu lati ons wen! not necessarily more relatcd than more distant populations. Thl.! high lewl of popu larion diffe rentiation in W llodij1ora may to some extent he attrihuted to the t:ffects of tire as pn!dicted at the ou tset of the s[m.l y but the c lu ster analysis shows that distance is a major harri er to ge ne now since neighbouring populations of W lloc/ij10nl arc as related as populations of the high ly panmictic W scJ/lt'{/rzii. Gene /low between popul atio ns of W lIoc/ij1ora is therefore efficient at close range. Population differentiation in W cedarbergellsis can be att ri buted to lack of gene tl ow within populations caused by poor seed di spersal and la<.:k of pollen movement. Th is indicates the vulnerability of the Clanwi lliam cedar's popUlation ge ne ti c stmc lUre to fragme nt ation. The genc tic rdationships between popUlations strengthe ns ev ide nce for a histori call y more continuous distrihuti on of HZ ccliarbergellsis. Geographic proximit y of populati ons was flat correlated with genetic relatedness which implies that all the populations of W cedarhergellsis wc re once parL of a greater panmictic popu lation with a hi gh degree of relatedness. As the meta population became in creasin gly frag mented. low levels of gene now hetwee n the various sub-populations may have Jed to the genetic diffcrl! ntiation of the smaller suh-populations as a result of selting. The seed sou rce plantation at Middelberg (MB) has a ffinities with a small northern population (KK) and with a 139 population inc1udcd in the Cedar pre servati on area (WE).

Conclusion
Our eieclrophoretil: comparison of genetic variation and structure in three Widdrillgtollia s pecies showe<..l littl e differe nce between th e threatened W ced(lrhagel/sis and the geographically rest ri cted \V sc/Jll'{m:ii. In long-li ved tn:es such as these. a populati on hottleneck \\lould he unlikely to have gt.!l1etic dfects except on timespans of millennia. OUi" data dot.:s provide somc supporr for a more continuous di st ributio n or the C lanw iJli am cedar in the past. The current fragmented population structure may be . l eading to increased levels of se. itin g as trees become lIlore di stant from one an other. The relatively high leve ls of inbreedi ng may be caus ing in breed in g (kpression. In conifers. this can he expressed in reduced st:l!d production. poorer germination rates and reduced seed lin g grow th (Charlt;!sworth & Charlesworth 1987). Inbreedi ng de pression due to frag me ntatio n 01" cedar pop~ ulat ions co uld tlll.:rdon.: pote ntially contribute to population decline.
It is int en.:sting to notc the large differences in popUlation genetic structure hetween the Iwo nOll-s prouting species and the sprout ing W llOdifiom. Ge netically. this spec ies behaves al most like an apomict wit h very liltl!.!: gene flow be tween populations. Genetic differences between sprouting and non-sp routing li fe hi stori es have bee n ignored in previolls reviews of life hi story effects on the di stribution of genetic va ri ation (e.g. Hamrick & Godt 1990). Further comparati ve studi es are needed to es tablish the gene rality of the patterns reported here.