Population Structure and Genetic Diversity of Castanea sativa Mill. Genotypes in the Republic of Croatia

Abstract

The European sweet chestnut (Castanea sativa Mill.) is an ecologically and culturally significant forest tree species in Croatia; however, its genetic diversity and population structure remain insufficiently characterized. This study aimed to evaluate the genetic diversity, structure, and connectivity of chestnut populations on Zrin Mountain, the country’s largest continuous chestnut area. Using seven nuclear SSR markers, we genotyped 153 individuals from three populations (PET, HRK, and BAC). All populations exhibited moderate genetic diversity (mean He = 0.571), with BAC showing the highest allelic richness and number of private alleles. AMOVA revealed that most genetic variance (67%) occurred among individuals, while population differentiation was low to moderate (FST = 0.064; PhiPT = 0.146), consistent with high inferred gene flow (Nm = 7.48). Both STRUCTURE and PCoA indicated that HRK was the most genetically distinct population, whereas PET and BAC were more similar. Overall, these findings demonstrate substantial gene flow and connectivity among Croatian chestnut populations, providing a foundation for sustainable management and conservation strategies in a broader European context.

1. Introduction

The European sweet chestnut (Castanea sativa Mill.) is a broadleaf tree species indigenous to Europe. It covers approximately 2.5 million hectares, stretching from the Atlantic coast of Portugal across central and southern Europe to the Black Sea, with approximately 1.7 million hectares concentrated in southern Europe [1,2]. It thrives in temperate climates with well-drained, acidic soils and is valued for its edible nuts, durable timber, and ornamental appeal [1]. Chestnut populations are shaped by both natural processes and centuries of human activity, resulting in a mosaic of wild and cultivated stands. Old groves and coppice forests are common in central and southern Europe, where regeneration occurs through a combination of natural seed dispersal and human influence. Coppiced stands, in particular, are dominated by younger trees due to periodic cutting and resprouting cycles [1,3,4]. Successful regeneration depends on competition, climate, and the prevalence of diseases [5,6,7,8,9].

Over the past two centuries, chestnut has faced severe challenges, including habitat fragmentation and the introduction of invasive pathogens and pests such as chestnut blight (Cryphonectria parasitica), chestnut gall wasp (Dryocosmus kuriphilus), and root and ink diseases caused by Phytophthora species. Additionally, the expansion of cultivated varieties into wild populations has raised concerns about genetic integrity. These pressures have intensified interest in the genetic diversity and resilience of natural chestnut populations [10,11,12,13,14,15].

In Croatia, the chestnut is an ecologically and culturally significant noble deciduous tree, covering approximately 136,000 hectares in mixed and pure stands—approximately 4.9% of the nation’s forest area [16]. It forms four main forest communities: Querco-castanetum sativae Horvat 1938 [17,18], Helleboro multifidi-Castanetum sativae Anić 1953 [19] nom. nov., Castaneo sativae-Fagetum Marinček & Župančič 1995 [20], and Aposeri foetidae-Castanetum sativae Medak 2011 [17,18,19,20]. Chestnut stands are scattered across continental Croatia, the Istrian Peninsula, and the islands of Krk and Cres, with the largest continuous stands found on Zrin and Petar Mountains [21]. The remarkable quality and health of these forests can be traced to historical factors. For centuries, the Zrin and Petar Mountains served as Europe’s military frontier against the Ottoman Empire, during which diverse ethnic groups, including Croats, Vlachs, Serbs, Germans, and Hungarians, introduced resilient agroforestry practices, including chestnut cultivation [22]. This factor influenced the development of diverse genetic variability in chestnuts within that area. In recent decades, chestnut forest area has declined due to widespread dieback, removal of infected trees, and reduced reliance on chestnut timber for traditional uses [23,24].

Genetic diversity and population structure in forest trees have commonly been investigated using microsatellite markers; simple sequence repeats (SSRs), valued for their high polymorphism; codominant inheritance; and reproducibility [25]. The earliest SSR applications in forest trees date back to the 1990s, with Pinus radiata [26], and were subsequently extended to multiple Quercus species [27,28,29,30,31,32]. In chestnut (Castanea sativa), SSR markers were first introduced by Buck et al. and Marinoni et al. in 2003 [33,34]. Two marker series, the EMCs loci (e.g., EMCs-4, EMCs-15, EMCs-25, EMCs-32, EMCs-38) and the CsCAT loci (e.g., CsCAT1, CsCAT3, CsCAT6, CsCAT16) have since found widespread use in chestnut genetics. These SSR loci are effective tools for evaluating genetic diversity, quantifying population differentiation, and identifying cultivars in both wild and cultivated chestnut populations across Europe [35,36,37,38,39,40].

Recent studies have revealed a rich genetic landscape within Croatian chestnut populations. Idžojtić et al. [41] identified 11 multilocus genotypes among 72 individuals of the traditional “Lovran Marron” variety. Prgomet et al. [42] reported 62 alleles across 10 SSR loci in 17 wild trees from Istria and Primorsko-Goranska County. Poljak et al. [43], studying populations from Croatia and the surrounding region, found three genetic clusters across central Europe and the western Balkans shaped by glacial refugia and human-mediated gene flow. These findings highlight the genetic uniqueness of Croatian chestnut germplasm and underscore the need for targeted conservation measures.

We hypothesize that the chestnut populations on Zrin Mountain have moderate to high genetic diversity, both naturally and due to previous human management, with the privately held BAC population having unique genetic signatures because of its divergent management history.

2. Materials and Methods

2.1. Study Area and Plant Material Collection

The study was conducted on Zrin Mountain in central Croatia, where three chestnut populations—PET, HRK, and BAC—were selected (Figure 1) based on their inclusion in previous research and monitoring projects. The PET and HRK sites host permanent monitoring plots established within the framework of the project “Protection of Sweet Chestnut Forests” [44]. The BAC locality has been part of the “Experimental Chestnut Grove” project [45]. All three sites are characterized by uneven-aged coppice stands dominated by naturally regenerating chestnut trees.

Leaf material in the PET, HRK, and BAC chestnut populations was collected using a simple random sampling approach in 2011, 2013, and 2016, respectively. A total of 153 chestnut leaf samples were collected: 52 in the PET population, 51 in the HRK population, and 50 in the BAC population (

Table 1. Sampling locations of Castanea sativa populations on Zrin Mountain, including geographic coordinates, altitude range, ownership type, year of sample collection, and sample size (N).

Code	Location Description	Latitude/ Longitude	Altitude (m)	Ownership	Sampling Year	N
PET	Department 47a, Management Unit Vučjak–Tješnjak, Forest Office Petrinja, Forest Administration Sisak, Croatian Forests Ltd.	45.416972° N, 16.255232° E	170–390	State forest	2011	52
HRK	Department 90a, Management Unit Šamarica I, Forest Office Hrvatska Kostajnica, Forest Administration Sisak, Croatian Forests Ltd.	45.229502° N, 16.493670° E	140–240	State forest	2013	51
BAC	Hrastovička Gora	45.386900° N, 16.273600° E	374	Private forest	2016	50

). Two to three healthy, fully expanded leaves were sampled from each tree at a minimum spacing of 20 m. Although a ≥20 m spacing may not fully prevent resampling of coppice ramets, 20 m was selected as a pragmatic upper bound—stools rarely exceed ~5 m in diameter and clustering occurs at ~10–20 m—and this limitation was mitigated by excluding stems from the same stool/root collar.

The collected leaf samples were immediately stored in plastic bags containing 12 g of silica gel (Kemika, Zagreb, Croatia) to ensure rapid desiccation and preservation of DNA integrity. Samples were then transported to the laboratory for molecular analysis. Chestnut maps and points—1. PET, 2. HRK and 3. BAC—were spatially displayed in ArcMap, ArcGIS Desktop 10.8.2 (Esri, Redlands, CA, USA).

2.2. DNA Isolation

Genomic DNA was extracted from approximately 0.1 g of dried leaf tissue using two different methods, depending on laboratory logistics. For samples collected in 2011 and 2013, DNA was isolated using a modified CTAB protocol [46], with the extraction buffer supplemented with 2% polyvinylpyrrolidone (PVP). DNA from samples collected in 2016 was extracted using the NucleoSpin Plant II kit (Macherey-Nagel, Düren, Germany) according to the manufacturer’s instructions. DNA concentration was quantified spectrophotometrically at 260 nm using a BioSpec-nano spectrophotometer (Shimadzu, Kyoto, Japan). DNA quality was evaluated by electrophoresis in a 0.8% agarose gel (Sigma-Aldrich, St. Louis, MO, USA) prepared in 1× TBE buffer and stained with SYBR™ Safe DNA Gel Stain (Thermo Fisher Scientific, Waltham, MA, USA). DNA aliquots (100 µL) were stored in 2 mL Eppendorf microcentrifuge tubes at −80 °C in an ultra-low-temperature freezer (ARCTIKO, Esbjerg, Denmark) to ensure long-term preservation of DNA integrity. Because extractions were performed with two protocols across years (CTAB in 2011; a silica-membrane commercial kit thereafter), extraction method constituted a potential source of batch effects. We mitigated this by using identical PCR mixes and cycling conditions, randomizing samples across plates/runs, and applying uniform allele binning/scoring. Nonetheless, minor method-related differences cannot be ruled out and were acknowledged as a limitation of this study.

2.3. SSR Primer Screening and PCR Amplification and Analysis

Fifteen fluorescently labeled SSR primer pairs (Table S1 and

Table 2. Characteristics of the SSR markers used for genotyping Castanea sativa populations.

Locus	Fluorescent Dye	5′-3′ Sequences (F/R)	Expected Length (bp)	Repeat Motif	References
EMCs2	NED	GCTGATATGGCAATGCTTTTCCTC/ GCCCTCCAGCCTCACTTCATCAG	172–178	(CGG)7	[30]
EMCs10	PET	GTCTCCCCCAATCATAAGTAGGTC/ TCAAGGGAACATTAGGTCATTTTT	218–230	(CA)8	[30]
EMCs13	VIC	TAGTCGGAGTACGGGCACAG/ TGATATGAGCATTTGACTTTGATT	158–164	(GCA)8	[30]
EMCs15	6-FAM	CTCTTAGACTCCTTCGCCAATC/ CAGAATCAAAGAAGAGAAAGGTC	089–095	(CAC)9	[30]
EMCs17	6-FAM	CGCCACGATTAGCTCATTTTCA/ GAGGTAGGGTCTTCTTCGGTCATC	210–222	(AGC)4(CCAA)5	[30]
EMCs25	6-FAM	ATGGGAAAATGGGTAAAGCAGTAA/ AACCGGAGATAGGATTGAACAGAA	140–158	(GA)12	[30]
CsCAT15	6-FAM	TTCTGCGACCTCGAAACCGA/ GCTAGGGTTTTCATTTCTAG	125–160	(TC)12	[31]

) were initially tested in single-primer PCR reactions using DNA from two chestnut samples. Out of the 15 tested loci, five (EMCs38, CsCAT6, CsCAT7, CsCAT34, and CsCAT41) were excluded due to the presence of three or more alleles per genotype. Two loci (CsCAT3 and CsCAT16) failed to amplify. One locus (CsCAT8) produced poorly resolved peaks, which made reliable allele scoring difficult. Loci that yielded well-resolved and reproducible electropherogram peaks, with clear distinction between homozygous and heterozygous genotypes, were selected for inclusion in the multiplex PCR assays. Multiplex 1 included five loci (EMCs2, EMCs10, EMCs13, EMCs15, and EMCs17). Multiplex 2 included two loci (EMCs25 and CsCAT15). Each 20 μL PCR mixture contained 1× PCR buffer, 200 μM of each dNTP, 1 U of Taq DNA polymerase (TAKARA Co. Ltd., Tokyo, Japan), and 1–2 ng of genomic DNA. Primer concentrations ranged from 0.1 to 0.4 μM depending on the locus. The PCR thermal profile consisted of an initial denaturation at 94 °C for 5 min, followed by 30 cycles of 94 °C for 30 s, 55 °C for 1 min, and 72 °C for 40 s. A final extension was performed at 72 °C for 2 min [31]. PCR products were separated by capillary electrophoresis on a 3500 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). Allele sizes were determined using GeneMapper™ IDX Software v1.5 (Applied Biosystems).

2.4. Genetic Analyses

Genetic diversity and population structure parameters were estimated using standard population genetic statistics implemented in GenAlEx version 6.5 [47]. For each population and locus, the following indices were calculated: number of alleles (Na), effective number of alleles (Ne), Shannon information index (I), rare alleles, private alleles (Pa), observed heterozygosity (Ho), expected heterozygosity (He), unbiased expected heterozygosity (uHe), fixation index (F), multilocus heterozygosity (MLH), standardized multilocus heterozygosity (sMLH) and internal relatedness (IR). Population differentiation was evaluated using Wright’s fixation index (FST), PhiPT (an AMOVA-based analog of FST), and R-statistics (RST, RIS, RIT), which account for allele size differences under a stepwise mutation model, making them particularly suitable for microsatellite data. Gene flow (Nm) was estimated from FST values as an indirect measure of the number of migrants per generation. Nei’s genetic distance (D) was calculated to quantify divergence among populations. Partitioning of genetic variance was assessed through analysis of molecular variance (AMOVA), which decomposed total variation into among- and within-population components, with significance tested using 999 permutations. Formulas, definitions, and references for all indices were provided in Table S2 [48,49,50,51,52,53,54,55,56,57,58,59,60]. Principal Coordinate Analysis (PCoA) was performed in GenAlEx to visualize genetic relationships among individuals based on multilocus genotype data.

To infer population structure, a model-based Bayesian clustering analysis was performed in STRUCTURE version 2.3.1 [61]. The analysis tested values of K (number of genetic clusters) from 1 to 10 using the admixture model with correlated allele frequencies. Each run consisted of a burn-in period of 1000 iterations followed by 10,000 Markov Chain Monte Carlo (MCMC) iterations, replicated 3 times for each K value. The most likely number of clusters was determined using the ΔK method of Evanno et al. [62], based on the rate of change in the probability of data between successive K values. StructureSelector [63] was used to summarize and visualize the results across replicate runs. Consensus bar plots for the most likely K values were generated with CLUMPAK [64].

3. Results

3.1. Genetic Diversity Analysis of Loci and Populations

Allele frequency analysis across the seven selected microsatellite loci revealed clear patterns of genetic variation and differentiation among the three chestnut populations (

Table 3. Genetic diversity and genetic differentiation values calculated using microsatellite (SSR) markers (FIT; Fixation index, FIS; inbreeding coefficient; FST; genetic differentiation coefficient between populations, Nm; Gene flow between populations).

Locus	Number of Alleles (bp)	FIS	FIT	FST	Nm
EMCs13	3 (155, 158, 161)	0.095	0.134	0.043	5.550
EMCs15	3 (79, 82, 88)	0.130	0.213	0.096	2.360
EMCs2	4 (156, 159, 162, 165)	–0.048	0.009	0.055	4.326
EMCs10	4 (214, 216, 222, 226)	0.037	0.045	0.009	27.342
EMCs17	4 (205, 209, 213, 217)	0.130	0.221	0.104	2.146
CsCAT15	8 (118, 120, 122, 124, 128, 132, 134, 138)	0.565	0.614	0.112	1.974
EMCs25	10 (138, 140, 144, 146, 148, 150, 154, 156, 158, 160)	−0.033	−0.004	0.028	8.689
Total/Mean ± SE	36 alleles	0.125 ± 0.078	0.176 ± 0.081	0.064 ± 0.015	7.484 ± 3.432

and Table S3). Sample sizes were consistent across loci and populations, ensuring the comparability of estimates. The number of alleles per locus ranged from three (EMCs13, EMCs15) to ten (EMCs25), with the highest allelic richness observed at EMCs25 (ten alleles) and the next highest at CsCAT15 (eight alleles). At EMCs2, allele 159 predominated in all populations, with minor alleles at low frequency. EMCs10 showed a shared predominance of allele 216 (especially in HRK), while allele 226 was elevated in PET. EMCs13 was dominated by allele 158, with allele 155 comparatively frequent in HRK. At EMCs15, allele 88 was the most common across populations, alongside population-specific variants (e.g., 82 in HRK and 79 in PET). EMCs17 contrasted among sites: allele 213 predominated in PET and BAC, whereas allele 209 was most frequent in HRK. EMCs25 exhibited considerable variability, including BAC-specific alleles (e.g., 158, 144), whereas PET and HRK contained common alleles, such as 156, at varying frequencies. CsCAT15 was highly polymorphic with multiple rare and private alleles; allele 132 was common in PET and BAC but rare in HRK, whereas allele 122 predominated in HRK.

Standard population genetic diversity analyses data were shown in Table S4. Across the three populations, the number of alleles was generally moderate, with the highest allelic richness observed at locus EMCs25. Effective allele numbers and the Shannon information index followed a similar pattern, indicating that EMCs25 contributed the most to overall allelic diversity, particularly in BAC. Observed heterozygosity varied widely among loci and populations. In PET, despite high allelic richness at some loci, several were dominated by homozygotes, while HRK showed a more consistent trend of heterozygote deficiency. In contrast, BAC maintained higher allelic richness and balanced heterozygosity at most loci, although some loci still showed evidence of homozygote excess. Fixation indices revealed that PET was largely in equilibrium, with a tendency toward heterozygote excess at several loci, except for a strong homozygote bias at EMCs25. HRK, however, showed heterozygote deficiency across multiple loci, while BAC displayed mixed patterns, combining high allelic richness with signals of inbreeding at certain loci.

Although private alleles in the HRK population (

Table 4. Private alleles with frequencies below 5% identified in Castanea sativa populations HRK and BAC.

Population	Locus	Allele	Frequency
HRK	CsCAT15	120	0.020
HRK	CsCAT15	134	0.010
HRK	EMCs25	154	0.029
BAC	CsCAT15	118	0.010
BAC	EMCs2	156	0.010
BAC	EMCs25	144	0.130
BAC	EMCs25	150	0.010

) were detected at low frequencies, their presence indicates a degree of genetic distinctiveness, potentially reflecting rare or recent variants. In contrast, the BAC population not only harbored a greater number of private alleles but also exhibited them at higher frequencies. Of particular note, three private alleles were detected at locus EMCs25, including allele 144, showing the highest private allele frequency. By contrast, the PET population exhibited no private alleles.

Genetic diversity indices from seven SSR loci were summarized in

Table 5. Genetic diversity parameters (Mean and SE) for Castanea sativa populations PET, HRK, and BAC. N: average number of alleles observed, Na: number of different alleles, Ne: effective allele number, I: Shannon index, Ho: observed heterozygosity, He: expected heterozygosity, F: Fixation index, Stat: statistic (Mean and SE), Pop: population.

Pop	Stat	N	Na	Ne	I	Ho	He	F
PET	Mean	52	4.143	2.39	1.015	0.511	0.569	0.074
	SE		0.553	0.173	0.081	0.07	0.03	0.127
HRK	Mean	51	4.143	2.313	0.979	0.448	0.553	0.179
	SE		0.508	0.157	0.071	0.043	0.036	0.071
BAC	Mean	50	4.429	2.631	1.096	0.506	0.59	0.127
	SE		0.571	0.321	0.113	0.044	0.042	0.085
Total	Mean	51	4.238	2.444	1.03	0.488	0.571	0.127
	SE		0.3	0.129	0.051	0.03	0.02	0.054

. Allelic richness (Na) was comparable in PET and HRK and slightly higher in BAC. The effective number of alleles (Ne) was also highest in BAC, indicating a more even allele-frequency distribution. Shannon’s index (I) was highest in BAC, intermediate in PET, and lowest in HRK. Ho was greatest in PET, similar in BAC, and lowest in HRK; He showed the same ranking (overall He = 0.571). F was lowest in PET and highest in HRK, with BAC intermediate, indicating minimal deviation from HWE in PET and a stronger heterozygote deficit in HRK. Overall, BAC showed the greatest genetic diversity across Na, Ne, I, and He, whereas HRK exhibited comparatively lower diversity alongside higher inbreeding. F was lowest in PET and highest in HRK, with BAC intermediate, indicating minimal deviation from HWE in PET and a stronger heterozygote deficit in HRK. Overall, BAC showed the greatest genetic diversity across Na, Ne, I, and He, whereas HRK exhibited comparatively lower diversity alongside higher inbreeding.

3.2. Genetic Differentiation Between Chestnut Populations: PET, HRK, and BAC

AMOVA based on individual genotypes indicated that most genetic variation occurred within individuals within populations (intra-populations/intra-individuals), followed by variation among individuals within populations (intra-populations/among individuals), and only a small fraction among populations (

Table 6. Analysis of molecular variance (AMOVA) showing the partitioning of genetic variation among Castanea sativa populations (PET, HRK, and BAC), among individuals, and within individuals.

Source	df	SS	MS	Est. Var.	%
Among Populations	2	1542.097	771.048	5.480	3%
Among Individuals	150	31,826.374	212.176	50.300	30%
Within Pops Within pops-Within individuals	153	17,071.000	111.575	111.575	67%
Total	305	50,439.471		167.355	100%

Note: df = degrees of freedom; SS = sum of squares; MS = mean square; Est. Var. = estimated variance; % = percentage of total variance.

). This distribution indicates that genetic diversity was largely maintained as heterozygosity within trees and as differences among trees within stands, whereas differentiation among PET, HRK, and BAC was modest, in line with the low–moderate population structure estimates reported elsewhere in this study.

Across all genotyped trees (n = 153), multilocus heterozygosity was moderate; sMLH centered on 1 (by definition), and internal relatedness was positive on average, indicating individuals tended to carry more common allele combinations than expected under random mating (Tables S5 and S6). Positive IR values were consistent with some mix of non-random mating, family structure in sampling, or restricted gene flow.

At the population level, sMLH was highest in PET, intermediate in BAC, and lowest in HRK. IR was positive in all populations and similar in magnitude; PET combined higher sMLH with relatively higher IR (Table S7).

R-statistics under the stepwise mutation model indicated low but significant among-population differentiation (small RST) and substantial within-individual components (RIS, RIT; both significant), with high indirect gene flow (Nm) (

Table 7. R-statistics describing the genetic structure among and within Castanea sativa populations (PET, HRK, and BAC) based on SSR marker data. RST was calculated based on the stepwise mutation model using SSR markers. p-values were based on 999 permutations. Gene flow (Nm) was estimated as Nm = (1 − RST)/(4 × RST).

Statistic	Value	p (Rand ≥ Data)	Interpretation
RST	0.033	0.001	Among-population differentiation (stepwise model)
RIS	0.311	0.001	Within-individual diversity relative to subpopulations
RIT	0.333	0.001	Within-individual diversity relative to total population
Nm	7.385	—	Estimated gene flow (number of migrants per generation)

).

A two-level AMOVA, partitioning variance only among and within populations, revealed that genetic variation among populations was relatively small compared to the much larger component of variation within populations (

Table 8. Two-level AMOVA for Castanea sativa populations (PET, HRK, and BAC).

Source	df	SS	MS	Est. Var.	%
Among Pops	2	90.659	45.330	0.798	15%
Within Pops	150	698.903	4.659	4.659	85%
Total	152	789.562		5.457	100%

Note: df = degrees of freedom; SS = sum of squares; MS = mean square; Est. Var. = estimated variance; % = percentage of total variance.

). This higher proportion of among-population variance reflects the aggregation of both among-individual and within-individual components into a single within-population category, thereby inflating the apparent contribution of population-level differences.

PhiPT, which incorporates repeat-length differences among SSR alleles, provides a refined measure of differentiation beyond allele frequencies. AMOVA/PhiPT indicated significant among-population structure, with the majority of variation within populations (p < 0.01;

Table 9. Summary statistics from AMOVA and gene flow analysis of Castanea sativa populations (PET, HRK, and BAC). PhiPT—fixation index from AMOVA; Nm—number of migrants per generation.

Stat	Value	p (Rand ≥ Data)
PhiPT	0.146	0.001
Nm	1.461	-

).

Pairwise comparisons revealed the highest genetic similarity between PET and BAC (0.939), consistent with the lowest FST value observed (0.018). The most differentiated populations were HRK and BAC, with the lowest genetic similarity (0.783) and the highest FST (0.067).

Comparing PhiPT, FST, and RST shows a metric-dependent pattern: PhiPT indicates moderate, significant among-population structure; pairwise FST suggests low–moderate differentiation based on allele frequencies; RST is small under the stepwise mutation model.

3.3. Population Genetic Structure Analysis

PCoA resolved clear structure: PET and BAC clustered closely, whereas HRK was more distant. The first two axes captured essentially all variation; along axis 2, BAC showed partial separation but remained closer to PET than to HRK (Figure 2).

STRUCTURE analysis (Figure 3), based on SSR loci, was used to evaluate the genetic structure of the three chestnut populations: PET (POP1), HRK (POP2), and BAC (POP3). The ΔK method identified K = 2 (Figure 3a) as the optimal value, based on the highest rate of change in log-likelihood between successive K values, while LnP(K) provided additional support for model fit across the tested range (Figure 3b). At K = 2, the STRUCTURE bar plot (Figure 3c) revealed a clear separation of individuals into two main genetic clusters, corresponding to the strongest hierarchical division among the studied chestnut populations. At K = 3, one of the major clusters identified at K = 2 (POP2 (HRK)) was further subdivided, revealing additional genetic differentiation among populations. At K = 4, further sub-structuring was detected, with individuals assigned to four distinct clusters, although considerable admixture was evident in several populations.

4. Discussion

4.1. Hypothesis and Objectives Evaluated

We hypothesized that Zrin Mountain chestnut populations would maintain moderate to high genetic diversity shaped by both natural processes and historical human management, with BAC bearing unique signatures due to its divergent management history. Our results support this overall: genetic diversity was moderate (overall He = 0.571; Table 5), BAC harbored the greatest number of private alleles, and PET and BAC were genetically close, whereas HRK was the most distinct (Figure 2 and Figure 3). Differentiation among populations was low to moderate, with high inferred connectivity (Table 3, Table 7, and Table 9). Together, these patterns are consistent with ongoing gene flow combined with localized demographic and management histories. In particular, BAC was enriched for private variants, while the distinctiveness of HRK suggests additional geographic or ecological influences [65,66].

4.2. Genetic Diversity and Allelic Patterns

All three populations showed moderate diversity, in line with European SSR studies of chestnut [15]. PET had the highest Ho, while BAC showed the greatest number/frequency of private alleles (Table 4), a combination compatible with local demographic history and management influences [66]. The overall heterozygote deficit (Ho < He) and positive fixation index indicate mild inbreeding or substructure typical of long-lived, outcrossing hardwoods [67]. Locus-specific contrasts (e.g., EMCs25) underline heterogeneous demographic or selective histories, potentially reinforced by age-class structure [68,69].

4.3. Population Differentiation and Gene Flow

AMOVA attributed most variance to within individuals, with a small among-population component, matching expectations for outcrossing, wind-pollinated trees (Table 9; [70,71,72,73]). Positive internal relatedness across stands suggests some parental relatedness/localized mating but at moderate levels (Tables S5–S7). Pairwise distances and FST rankings identified PET–BAC as most similar and HRK as most divergent, consistent with ordination and clustering (Figure 2 and Figure 3). High indirect gene flow estimates further support substantial connectivity.

The AMOVA results were predominantly advantageous for resilience. Heterozygous genotypes could buffer deleterious recessives and confer broader performance across variable environments, while abundant standing variation within populations supplied raw material for selection under climate, pathogen, or disturbance pressures. The results also signaled effective outcrossing and gene flow, processes that sustained adaptive potential. However, because these estimates came from neutral SSRs and a moderate heterozygote deficit (Ho < He) was present, the implications had to be viewed cautiously: long-term resilience depended on functional genetic diversity and demography, and localized inbreeding or substructure could still have eroded adaptive capacity if connectivity or effective population size declined.

4.4. Role of Mutation: Insights from R-Statistics

Under the stepwise mutation model, RST was small relative to PhiPT and FST, indicating that stepwise mutation contributes little to the observed structure compared with gene flow and drift (Table 7; [74,75,76,77]). High within-individual components (RIS, RIT) are consistent with an outcrossing mating system and maintenance of heterozygosity [69]. The modest mutational signal also implies that most private alleles were better explained by restricted dispersal, recent/local demography, or human influence rather than mutation alone.

4.5. Population Structure Patterns

PCoA and STRUCTURE yielded a concordant picture: PET and BAC cluster together, HRK forms the most distinct unit, and K = 2 was the best-supported partition with admixture in BAC (Figure 2 and Figure 3). This mirrors patterns seen across European datasets, where K = 2 and admixture were frequently recovered in forest and cultivated gene pools [14,39,78], situating Croatian stands within broader continental structure.

4.6. Methodological Context and Limitations

At the study scale, the curated 7-locus panel provided sufficient resolution to detect among-population signals while emphasizing within-individual variance. We noted two limitations discussed in Methods: (i) the ≥20 m spacing reduces—but may not eliminate—coppice ramet resampling (with post hoc clonality checks), and (ii) extractions spanned two protocols across years; downstream workflows were standardized, and this was acknowledged as a potential source of batch effects.

4.7. Implications for Conservation and Genetic Resource Management

Although overall differentiation among populations was low, the presence of private alleles and site-specific inbreeding patterns highlights the importance of localized conservation strategies. In particular, the BAC population, with its high allelic richness, should be prioritized as a reservoir of unique genetic diversity, while the HRK population, which showed elevated inbreeding, requires management to reduce further loss of diversity. These patterns may reflect historical bottlenecks or recent habitat fragmentation, underscoring the need for both restoration and monitoring.

Based on the genetic profiles of the three stands, we recommend the following, framed as clear, implementable actions. PET (higher Ho, well connected) should be designated as a seed stand: collect seed from ≥50 widely spaced (≥50 m) mother trees, prioritize seed-based over coppice regeneration, retain ≥20 veteran seed trees, and repeat genetic monitoring every 8–10 years. HRK (most distinct, lower diversity, higher inbreeding signal) should be treated as a genetic conservation unit: expand effective size by retaining ≥100 reproductive trees, reduce reliance on coppice, establish pollen/seed corridors toward PET, and, if recruitment remains low, use enrichment planting with local seed first and a capped ≤10%–20% admixture from the most ecologically matched nearby source; protect young cohorts from browsing. BAC (highest private alleles; private ownership) should conserve unique variants via a stewardship agreement: avoid non-local plant material, favor seed regeneration in mast years, limit stool harvesting to curb clonal dominance, maintain moderate connectivity with PET, and back up diversity in an ex situ archive of ~40–60 genotypes. Across all stands, maintain disease surveillance (blight/ink), promote mixed age structure and habitat connectivity, standardize seed collection (trees ≥ 50 m apart, many unrelated mothers), and implement decadal genetic monitoring (e.g., He, Ho, FIS, sMLH, IR, private alleles), with management triggers to increase effective breeders and connectivity if FIS persists >0.15 or diversity declines.

Ownership can shape genetics via management. State forests usually use regulated seed sources, larger contiguous areas, and natural regeneration—promoting gene flow and homogenizing allele frequencies. A private stand like BAC may have smaller breeding neighborhoods, more coppicing from a limited set of stools, distinct disturbance/age structures, and occasional non-local introductions (nursery stock, grafts). These factors can elevate low-frequency, stand-specific alleles in BAC while pollen flow keeps overall divergence low, matching BAC’s higher private alleles and PET–BAC similarity.

4.8. Conservation Implications for European Chestnut

These findings have broader implications for chestnut conservation at the European scale. The high within-population diversity (67%) emphasizes the importance of preserving multiple genetically rich stands to maintain adaptive potential. The presence of population-specific alleles in BAC and HRK aligns with the goals of the Pan-European strategy for forest genetic resource conservation. Given its allelic richness and unique gene pool, the BAC population is a strong candidate for both in situ and ex situ conservation. Despite low overall FST and high gene flow, STRUCTURE and R-statistics revealed moderate population structuring, particularly in HRK, warranting regionally adapted conservation approaches [13,14,43]. These genetic data can inform seed sourcing, restoration, and breeding efforts, supporting forest resilience under future environmental conditions, including disease pressure and climate change [79,80]. Ultimately, this study provides essential genetic insights that can guide evidence-based conservation and management of chestnut across Europe, reinforcing its ecological and genetic value in forest ecosystems.

5. Conclusions

This study presents a comprehensive evaluation of the genetic diversity, structure, and conservation relevance of three chestnut populations in Croatia using highly informative SSR markers. The selected loci demonstrated strong genotyping performance, enabling accurate assessment of population-level genetic variability. All populations showed moderate genetic diversity, with PET exhibiting the highest Ho, BAC the greatest number and frequency of private alleles, and HRK emerging as the most distinct unit. Variance partitioning indicated that most diversity resides within populations (AMOVA: 67% within individuals, 3% among populations), while differentiation was low–moderate (PhiPT 0.146, p = 0.001; pairwise FST 0.018–0.067). High gene flow was inferred (Nm ≈ 7.48), consistent with connectivity typical of outcrossing, wind-pollinated, long-lived trees. Mutation-scaled differentiation was limited (RST = 0.033), and within-individual variance was high (RIS = 0.311; RIT = 0.333), reinforcing that gene flow and drift, rather than stepwise mutation, primarily shape structure at this scale. Locus-specific signals (e.g., EMCs25) flagged possible microevolutionary heterogeneity. Clustering analyses converged on K = 2, with BAC admixed and HRK distinct, mirroring FST/PhiPT contrasts and PCoA. These results suggest a combination of historical gene flow and localized evolutionary pressures shaping current genetic structure. The presence of population-specific genetic signatures, private alleles, and evidence of inbreeding underscores the importance of implementing both in situ and ex situ conservation strategies. Effective conservation planning should aim to preserve gene flow while protecting the unique genetic composition of isolated or differentiated populations. Overall, this study highlights the value of molecular tools in supporting conservation efforts and provides a solid genetic foundation for the sustainable management of European chestnut. The prioritization of genetically diverse and distinct populations, such as BAC and HRK, is essential for maintaining the adaptive potential of chestnut under current and future environmental challenges.