Genetic Maternity and Paternity in a Local Population of Armadillos Assessed by Microsatellite DNA Markers and Field Data

Genetic data from polymorphic microsatellite loci were employed to estimate paternity and maternity in a local population of nine‐banded armadillos (IDasypus novemcinctus) in northern Florida. The parentage assessments took advantage of maximum likelihood procedures developed expressly for situations when individuals of neither gender can be excluded a priori as candidate parents. The molecular data for 290 individuals, interpreted alone and in conjunction with detailed biological and spatial information for the population, demonstrate high exclusion probabilities and reasonably strong likelihoods of genetic parentage assignment in many cases; low mean probabilities of successful reproductive contribution to the local population by individual armadillo adults in a given year; and statistically significant microspatial associations of parents and their offspring. Results suggest that molecular assays of highly polymorphic genetic systems can add considerable power to assessments of biological parentage in natural populations even when neither parent is otherwise known.

ing genetic parentage in a local natural population in mum likelihood procedures.For cases such as the present when neither parent is known, these statistical proce-northern Florida.The first paper in this series described the development of microsatellite DNA markers for ar-dures were initially considered and then further explored in several earlier works (Thompson 1976(Thompson , 1986; Meagher madillos, documented the polyembryony phenomenon genetically, and detailed the spatial dispersion of clonal and Thompson 1986Thompson , 1987;;Thompson and Meagher 1987).First, genetic exclusions of biological parentage for sibships in this population (Prodo ¨hl et al. 1996).We now extend the genetic analyses across single generations by armadillos were evident when, for one or more of the assayed loci, neither allele in a juvenile's diploid genotype using microsatellite genotypes to estimate maternity and paternity for 99 juveniles sampled over a 4-yr period.matched an allele of the adult in question.However, several or many adults often remained nonexcluded as par-The genetic data are considered alone as well as in conjunction with natural history information such as the lac-ents by genetic evidence.Then, genetic likelihoods for nonexcluded parents were calculated (using as a baseline tating status of females and the exact geographic positions of offspring relative to their inferred parents. the allele frequencies in the local population sample) from the joint genotypic frequencies observed in particular combinations of adults and juveniles.These likeli-Material and Methods hoods, summarized as LOD scores (the likelihood ratio Collections and Genetic Markers between parent-offspring status and unrelatedness), were calculated for each individual parental gender separately Animals were caught (and then physically marked and (potential mother and father of a juvenile) and for parent released) at the Tall Timbers Reseach Station near Tallapairs simultaneously.Thus, the adult male or female havhassee, Florida, a locale embedded within the broader ing the highest LOD score is the maximum likelihood farange of nine-banded armadillos in the southeastern ther or mother of a given juvenile, and the parent pair United States.Thus, the population cannot be considered with the highest LOD score is the maximum likelihood closed.Collection sites were specified precisely using a parent pair.Mathematical as well as empirical investigasatellite global positioning system (Trimble Pathfinder tions indicate that positive correlations normally exist Basic ϩ).Intensive field sampling was conducted during between single-parent and parent-pair LOD scores the summers of 1992-1995. (Meagher and Thompson 1986).Nonetheless, for reasons Detailed procedures for the molecular assay of armadescribed later, absolute discrepancies between some of dillos are described elsewhere (Prodo ¨hl et al. 1996).
the individual parents estimated by these two approaches Briefly, an armadillo genomic library was constructed do arise in particular instances.In any event, for each juand screened with a hybridization cocktail consisting of venile who had multiple genetically possible parents in several different oligonucleotide probes.Positive recomour data, the four highest LOD scores for candidate binants were isolated and sequenced for development of mothers, fathers, and each parent pair were calculated PCR primers flanking di-and tetranucleotide repeat moand recorded for further analysis.tifs.For the current population survey, small notches of For any juvenile armadillo (defined as an individual ear tissue were preserved in ethanol and employed as a less than 1 yr old), all relevant individuals in the populasource of nuclear DNA for PCR amplifications of each of tion initially were considered as candidate parents.For seven of these polymorphic microsatellite regions, one of example, because armadillos usually do not begin breedwhich proved to be X-linked.The PCR products were ing before their third summer (McDonough 1992), deseparated through polyacrylamide sequencing gels, and spite being physiologically capable of breeding at year autoradiographs were developed and scored to reveal inone (McCusker 1977), juveniles in the 1992 sample were dividual genotypes at each locus.Sample sizes in the curincluded in the pool of candidate parents for juveniles rent study are slightly smaller than those reported previcollected in 1994 and 1995 but not 1993.For any given ously (Prodo ¨hl et al. 1996) because a few specimens for year-class of juveniles, the initial candidate pool also inwhom not all loci were scored were excluded from the cluded all other adults collected at any time during the current analysis.The earlier report provides information 4-yr study.The only exceptions involved individuals on levels of genetic variation and absence of detectable found dead; these were excluded from subsequent years population structure (as gauged by nonsignificant deparof parentage analysis.tures from Hardy-Weinberg equilibrium) for pooled col-One additional computational complication involved lections from the Tall Timbers site (Prodo ¨hl et al. 1996).
the X-linked microsatellite locus.At such a locus, the single allele in a male offspring derives from the mother and Statistical Analyses the allele in the father is irrelevant.The sex of the offspring and of the putative parent enters into population Microsatellite genotypes were employed to estimate genealogical relationships among individuals based on maxi-genotype frequencies, into exclusion probabilities, and into expected and realized LOD scores.Programs for au-Results tosomal loci were modified to accommodate the popula-Genetic Parentage Estimates tion genotype frequencies and segregation probabilities for any specified sex-linked loci and, hence, to compute All microsatellite loci resolved with exceptional clarity (examples in fig.1).Three to seven alleles per locus were the relevant exclusion probabilities and LOD scores.
As elaborated in the ''Results'' section, three stages of observed in this armadillo population, and their frequencies together with the calculated exclusion probabilities analysis were employed to summarize the geographical arrangements of juvenile armadillos and their probable for various parent-offspring combinations are presented in table 1.For all autosomal loci considered jointly, these biological parents.In the first stage, spatial distances were determined between juveniles and their putative parents, probabilities are reasonably high (about 97% for the exclusion of parent pairs, and 89% for exclusions when one who displayed the highest genetic LOD scores.Thus, these parentage estimates were based strictly on statistical parent has been identified).Table 2 provides numerical counts by year of the assayed numbers of juveniles, pos-analyses of the genotypes themselves, without reference to further life-history information.(However, in 28 cases sible parents, and genetically excluded maternal and paternal parents per offspring.In various years, between two or more potential parent pairs shared an identical highest LOD score, and the tie was broken by assuming 71% and 81% of the candidate mothers and fathers for a typical juvenile could be excluded as biological parents that the true parents were spatially closest.) In the second stage of analysis, spatial distances were on the basis of the genotypic data alone.Altogether, 69 genetically different juvenile sibships measured between juveniles and their putative parents as estimated from ''total evidence.''In addition to LOD (18 of which were represented by two to four polyembryonic clonemates, the remainder by singletons) were as-scores, total evidence involved considerations of lactating status of females (lactation provides a solid indication of sessed for parentage in this study.Table 3 (cols.1-3) lists the estimated mothers, fathers, and parent pairs for each recent litter production by candidate mothers; see table 6 below); prior genetic assignments (a candidate mother of these offspring based on the highest LOD scores.For 51% of the assayed sibships, the candidate mother with normally was excluded if she already had been assigned with higher likelihood to another genetically distinct ju-the highest LOD score was the same individual as the most likely mother identified in the parent-pair LOD venile of similar age; however, see below for a caveat to this rule, which stems from armadillo polyembryony and scores; for fathers, the same individual was identified by single-parent and parent-pair analysis in 30% of the sib-the possibility that clonal adults sometimes contributed to the same year-class of progeny); and the geographical ships.
Seven cases existed (e.g., offspring 40506F in year-class distance between parent and offspring (a consideration used to break ties or near ties when multiple possible 1992) in which no parent pair was compatible with the juvenile's genotypes yet individual adults of both genders parents remained after examination of all other evidence).This latter consideration adds a potential element were nonexcluded as parents (table 3).This can happen as for example when a juvenile displays genotype a 2 a 3 , of circularity to the spatial analyses, but this effect should be minimal because only six such cases were involved.adults of each gender exist with genotype a 1 a 2 , and no other a 3 alleles are observed in the population.One evi-Also, this consideration probably is realistic given the biology of the species.The third stage of the parentage dent possibility apart from de novo mutation is that some true biological parents were not included in the analysis involved comparisons of spatial distances between putative parents and offspring as had been esti-assays either because they were not captured or because they died or emigrated before being sampled.Also, the mated separately in stages one and two described above.
To assess the statistical significance of any spatial clus-estimated parent pairs often included a best single parent (male or female) not represented among the top four tering between putative parents and their offspring, randomization tests (Manly 1991) were conducted.These LOD score parents of a given gender.This can happen because a juvenile may carry in heterozygous condition a involved compiling artificial geographic distance matrices (1,000 matrices per test) for individuals spatially ran-rare allele that points to likely single parents of either gender but that need not be present in both parents domized across the Tall Timbers collection sites and then comparing these (Mantel 1967) against the empirical ma-when they are considered in pairs.Such instances demonstrate clearly that high LOD scores from assayed geno-trices of spatial distances between purported relatives.The null model under examination is that spatial associa-types do not in all cases assign parentage correctly or unambiguously.tions of estimated parents and offspring in the Tall Timbers population are no tighter than those expected for in-Columns 4 and 5 of table 3 list the most likely mothers and fathers of each offspring based on total evidence.In dividuals of unknown genetic relationship sampled at random from this population.
42% of all cases for fathers, and 30% for mothers, the as- signments of parentage based on total evidence were associated young.Based on genetic evidence, two of these putative mothers were pinpointed exactly by displaying identical to the assignments based strictly on the genetic data (i.e., on the highest LOD score either for a single the highest LOD scores both in the single-parent and parent-pair estimates.The third putative mother was parent or for a member of a parent pair).Indeed, in 13 cases (19%) the exact parent pair estimated from the ge-pinpointed in the parent-pair LOD score and appeared among the top four candidates in the single-parent LOD netic data agreed perfectly with the most likely parents identified by total evidence (table 3).Various other sum-estimates.The fourth putative mother by field evidence was not among the top four LOD-score candidates.mary statistics on parentage assignments are presented in table 4.
The biological parents inferred either from genetic evidence alone or from total evidence also permit estimates In four cases, field observations suggested strongly that particular females were the probable mothers of spatially of the fraction of the adult population that has bred suc- * Calculated exclusion probabilities are based on the genetic data.U is the probability of excluding an unrelated pair as parent and offspring; UU is the probability of excluding a parent pair when the trio are unrelated; QU is the exclusion probability for an unrelated parent when one parent has been identified correctly (Meagher and Thompson 1986).
† For case UU for sex-linked loci, the putative parents are two individuals of opposite sex, unrelated to each other or to the offspring.For case QU for sex-linked loci, the putative parent is an unrelated adult of opposite sex to the correctly identified parent.For example, if the putative parent is female, then the father is assumed correctly identified for case QU.
‡ Taken from the expression 1 Ϫ ∏ (1 Ϫ PE i ), where PE i is the exclusion probability at the ith locus and the product is taken across the loci examined.
§ The exclusion probabilities contributed by the sex-linked locus are tabulated separately because these depend on the sex of the offspring and of the putative parent (as described in the ''Material and Methods'' section).Note that at a sex-linked locus the genotype of a putative male parent is irrelevant to identifying a male offspring, regardless of whether the mother is identified correctly.Note also that there is generally greater exclusion power for a female offspring since she must receive an allele from each parent.The exception is the pairwise parental relationship with a putative mother; here, the potential heterozygosity of the female offspring leads to exclusion probabilities approximately half of those for the hemizygous male offspring, if in fact the individuals are unrelated (case U).Note: Each five-digit number is an individual's (or sibship's) identification tag.Boldface entries denote individuals identified as the best candidate parents both on the basis of one or more of the top LOD scores themselves and on the basis of total evidence, which included other ecological and biological information.Ellipses reflect unresolved ambiguities that precluded a ''best'' estimate of parentage.Numbers in parentheses are the distances (in meters) between putative parents and offspring.
* No compatible parental pair.† An identical maternal genotype is most likely for these to nonsibling juveniles (30021F and 30022F) collected in the same year.In this case, full-sibling (i.e., clonal) mothers have been responsible.cessfully at the Tall Timbers location (table 5).De-comparing the number of observed lactating females to the number of these females who were identified as likely pending on the parental gender and the method by which most likely parents were identified, approximately mothers of the assayed litters (table 6).On average across the 4 yr, only about 25% of the females observed lactat-36%-46% of the adult armadillos appear to have reproduced successfully one or more times during the 4 yr of ing in a given year were identified genetically as most likely mothers of an assayed armadillo brood of that year, our study.Most of these individuals reproduced only once, and fewer than 3% of the adults appeared to have and this value increased only to about 38% when the most likely mothers were estimated on the basis of total produced surviving offspring in all 4 yr (table 5).These values are based on offspring captured and, therefore, evidence (which included lactation status as a consideration).Thus, even for the pool of adult lactating females may underestimate reproductive activity to the extent that some juveniles may have died or left the study area (unweighted mean of 58% of the adult female population in a given year), demonstrable reproductive success ap-without being sampled.In any event, a large proportion of adults (Ͼ53%) appears to have left no surviving off-peared to be modest.Among the potential explanations are that juvenile mortality is high or that many juveniles spring to the Tall Timbers population between the years 1992 and 1995.The frequency distribution of numbers of remained unsampled (e.g., because they emigrated from the area or otherwise remained uncaptured).brood years per parent (as estimated by total evidence) does not differ significantly from Poisson expectations (Sneath and Sokal 1973) and, thus, gives no indication Spatial Relationships of Parents and Offspring that most of the successful reproduction was concentrated in particular parents (fig.2).
Geographic distances between each offspring and its inferred parents are shown in table 3.As might be ex-Reproductive success can be looked at another way by   pected, the spatial separations were significantly smaller [Manly 1991]).The genetic data revealed no significant on average for parentage estimates based on total evitendencies for offspring to occur closer to their most dence (which included geographic considerations) than likely mothers than to their fathers (table 7), although a for parentage estimates based on genetic LOD scores hint of such tendencies existed for parentage assessments alone (table 7).However, in nearly all categories of cases based on total evidence (group comparison C ϫ F in ta-(the only exception being ''best LOD mothers ϫ offble 7).spring''), mean geographic distances between inferred The actual spatial arrangements of estimated parents parents and offspring were highly significantly smaller and their offspring are shown on a schematic map of Tall than those for random pairs of individuals (Mantel test Timbers in figure 3. Note the tendency for spatial clustering as evidenced by the large numbers of short arrows connecting parents and offspring and the relative paucity Mother-to-father spatial distances were also considered (table 7).These tended to be larger on average than par-Mean ent-to-offspring spatial distances, though strongly sigdistance in nificantly so only in the comparison of ''best total evi- Second, successful reproduction by some of these adult In ANOVA tests for differences between group means, the following comparisons were highly significant (P Ͻ .where no suitable parents were found for a particular litter.

Discussion
rows through some of their first summer (Taber 1945;Galbreath 1982;McDonough and Loughry 1995;Prodo ¨hl et al. 1996).Litters appear to break up (from dispersal or Nine-banded armadillos are relatively asocial, burrowing mammals that are active primarily at night (Newman mortality) by the fall (McDonough and Loughry 1997).
The above account provides several reasons why the 1913; Galbreath 1982;McBee and Baker 1982).Physiological data (Enders 1966;McCusker 1977), as well as type of analysis reported here is necessary.First, because armadillos are asocial and nocturnal and occur in thickly field observations of males and females associating in close proximity (i.e., paired;McDonough 1992McDonough , 1997)), vegetated habitats, observations of reproductive behavior are rare (but see McDonough 1992McDonough , 1997) ) and cannot be suggest that the breeding season lasts from early summer through early fall.Males may be observed paired with used reliably to assign paternity and maternity for many adults in the population.Second, after litters emerge more than one female during a breeding season, but females typically pair with just one male (McDonough from their natal burrows, they appear to have little contact with the mother (McDonough and Loughry 1997Loughry ), 1992Loughry , 1997)).Implantation of the fertilized embryo is delayed but usually occurs by late fall or early winter so behavioral associations between a female and a set of juveniles that could lead to inferences about maternity (Storrs et al. 1988), with females giving birth to litters of genetically identical quadruplets the following spring are rare.Finally, home-range overlap between a female and a set of young is also not a reliable indicator of ma- (Newman and Patterson 1910;Patterson 1913;Storrs and Williams 1968;Prodo ¨hl et al. 1996).Litters first emerge ternity.Females may share burrows (Herbst and Redford 1991) and often have overlapping home ranges (Clark from their natal burrows from early May through August (Loughry and McDonough 1994) and remain in close 1951; Layne and Glover 1977;Breece and Dusi 1985;Herbst and Redford 1991;McDonough 1992McDonough , 1997)), thus proximity, foraging together and sharing the same bur-

Figure 1 :
Figure 1: Gel autoradiographs exemplifying alleles (numbered at the right) at four of the seven microsatellite loci assayed in armadillos.The Dnov-65 locus is X-linked with all (hemizygous) males displaying a single allele (Prodo ¨hl et al. 1996); the other loci are autosomal.
001): A ϫ C, A ϫ F, B ϫ C, pairs might have remained unrecorded in the genetic B ϫ F, C ϫ D, C ϫ E, D ϫ F, E ϫ F, C ϫ Z, X ϫ Z, and Y ϫ Z.The assays because of juvenile emigration or death.In any group comparisons C ϫ F and D ϫ X were marginally significant at event, these results suggest caution in extrapolating from P ϭ .04.No other group means were demonstrably different.Sample observations of pairing in nature to inferences about resizes for each group were corrected to reduce bias from situations productive success.

Figure 3 :
Figure 3: Map of the Tall Timbers study area showing all individuals (adult females, squares; adult males, triangles; juveniles, circles) collected over the 4 yr of the study.Symbols are superimposed when multiple individuals were collected at close locations.Lines with arrows connect all armadillo offspring to their best ''total evidence'' mothers and fathers (or individual parent on those few occasions when only one was deduced).These genetically linked individuals are indicated by filled symbols.Some arrows are drawn with kinks to simplify this presentation by minimizing overlaps.

Table 1 :
Allele frequencies at six autosomal microsatellite loci and one sex-linked microsatellite locus in the Tall Timbers populations of armadillos

Table 2 :
Mean (Ϯ SE) number per offspring of adult armadillos genetically excluded as parents for each year of the survey

Table 3 :
Parentage estimates for the total of 69 armadillo clonal sibships (F ϭ female, M ϭ male) at the Tall Timbers location

Table 4 :
Summary statistics regarding parentage assessments for the 69 total sibships of armadillos assayed from the Tall Timbers site

Table 5 :
Numbers and percentages of individuals identified as composing the parental gene pool of the Tall Timbers armadillo population Numbers in parentheses are percentages, which are based on the total of 112 potential male parents and 99 potential female parents over the 4 yr of this investigation. Note:

Table 6 :
Comparison of observed numbers of lactating females (field data) in the Tall Timbers armadillo population against the numbers of lactating females actually assigned as mothers of particular offspring litters : Numbers in parentheses are percentages.All percentages refer to the proportions of lactating females assigned as likely mothers of assayed litters in a given year. Note

Table 7 :
Summary statistics for geographic distances between offspring and their most likely parents (above), and between of such arrows traversing larger distances.
most likely mother and father pairs (below)