Genetic diversity of cultivated Gayo Arabica Coffee (Coffea arabica L.) based on morphological and microsatellite markers

Abstract

Background

Indonesia is the fourth largest global producer of coffee (Coffea spp.). The primary production centre for Coffea arabica in the country is the Gayo Highlands of Aceh. The Gayo Highlands in Aceh serve as a key arabica coffee-producing region, covering over 100,000 hectares across three districts: Aceh Tengah, Bener Meriah, and Gayo Lues, situated at elevations ranging from 900 to 2,000 m above sea level (asl). Despite the economic significance of Gayo Arabica coffee, studies on its genetic diversity remain limited. The objective of this study was to evaluate the morphological characteristics, particularly those associated with yield, and their genetic diversity.

Results

We measured two types of analyses: a) morphological characteristics (N = 32 characteristics) were analysed from 51 accessions, with the resulting data categorised into qualitative and quantitative traits. The assessment of qualitative traits was conducted through visual observation, while quantitative traits were measured. The mean values of these traits were then subjected to Principal Component Analysis (PCA) in order to ascertain the principal contributing variables. In order to provide a comprehensive evaluation, genetic diversity was analysed using molecular markers in addition to the morphological assessments; b) genomic DNA was extracted from young leaf tissues [N = 52 samples; -1 robusta as outgroup was sampled], and a polymerase chain reaction (PCR) was conducted using eight microsatellite (SSR) markers. Subsequent analyses were then performed on the resulting DNA profiles, utilising clusterisation methodologies. The results indicated significant morphological variation among the accessions, yet comparatively low genetic diversity, as evidenced by a Nei's genetic diversity index of 0.36. Each SSR marker detected 2 to 4 polymorphic bands. Markers Sat227 and Sat240 produced 2 to 3 bands, while M24, Sat32, and A8847 revealed an average of 4.5 bands per marker. Sat207, Sat255, and AY2434 each produced three bands, with Sat255 classified as moderate in polymorphism based on prior literature. This finding suggests that a genetic bottleneck may have occurred in the Gayo Highlands arabica population.

Conclusion

This work will be of particular relevance to breeding programs and conservation initiatives aimed at ensuring the sustainability of arabica coffee cultivation in Indonesia and other tropical highland regions. We found a relative high morphological diversities in terms of phenotypic trait (N = 32 traits) esp. in these yield related-traits e.g. leaf number per axil, number of leaves per branch, number of cherries per branch, etc.. However, the genetic diversity measured in this study by employing 51 arabica coffee accessions was found to be relatively low, namely 0.34. Furthermore, the use of Principal Component Analysis (PCA) helped to distil complex trait data, with the first principal component (PC1) accounting for 35.6% of the total variation, mainly driven by leaf- and yield-related characteristics. The findings of this study on the present genetic diversity of C. arabica will doubtless underpin future endeavours to enhance the species genetically, to develop future breeding programmes, and to improve the quality of the various varieties.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-025-06768-4.

Introduction

Coffea spp., which belongs to the Rubiaceae family, comprises 130 species within the genus [1]. The plant is predominantly found in tropical Africa, South America, and Asia, including Indonesia. Coffee is the world's second most important beverage and the second largest export commodity, exceeded in this regard only by petroleum products. It is a multi-billion-dollar industry that supports millions of people around the world [2, 3]. Arabica coffee, in particular, is vital for foreign exchange in developing countries, with global sales amounting to between $10 and $74 billion per year. Globally, there has been a consistent increase in the area of plantations, with the most recent figures indicating a growth from 10 to 10.2 million hectares [3]. This expansion is evident in approximately 80 tropical and subtropical countries, which are engaged in the trade- and supply-chain for this commodity. It is estimated that approximately 12.5 million households worldwide derive income from the coffee industry [4].

Despite the extensive genetic diversity exhibited by the Coffea genus, the global coffee trade is dominated by only two species: C. arabica L. (arabica) and C. canephora (commonly known as robusta or conilon) [5, 6]. Two other species, C. liberica Bull. ex Hiern and C. excelsa A. Chev., are cultivated to a much lesser extent and are better adapted to tropical lowland environments, typically at altitudes between 400 and 600 m above sea level (asl) [6]. Of these, arabica is distinctive in that it is the sole allotetraploid species (2n = 4x = 44), arising from a natural hybridisation between two diploid ancestors, C. eugenioides S. Moore and C. canephora. Arabica is predominantly self-fertile, although its wild ancestors are genetically self-incompatible [5, 7]. Despite its commercial importance, several studies have reported that arabica varieties have a narrow genetic base and relatively low genetic diversity [8]. With regard to environmental requirements, Arabica is most suited to tropical highland regions, with an altitude range of 800 to 2,200 m. The optimal temperature range for its growth is between 15–24 °C, and its growth is optimized under annual rainfall levels ranging from approximately 1,643 to 2,000 mm [6].

The southwestern montane forests of Ethiopia and the Boma Plateau in Sudan are recognized as the primary centres of origin and diversification for Coffea arabica, due to the high levels of genetic diversity existing in wild populations from these regions [2, 4, 9]. Although the species originated in Africa, its first known cultivation is believed to have taken place in Yemen, possibly as early as the sixth century [4, 10, 11]. From there, arabica coffee was introduced to India, Southeast Asia, Latin America, and East Africa during the 17th to 19th centuries [4, 12].

Modern arabica coffee varieties largely trace back to ancient cultivars selected for their superior cup quality. Among these, the"Mocha Yemen"coffee gave rise to two major botanical types: i) C. arabica var. bourbon, which reached South America via the island of La Réunion (formerly known as Bourbon island), and ii) C. arabica var. typica, the earliest cultivated type in Latin America and Asia, including in India and subsequently Indonesia (Java and Sumatra) [11, 13]. In the case of typica-derived varieties, many modern cultivars—particularly those in Brazil and other coffee-producing countries; descend from a single C. arabica plant originally brought from Indonesia to Europe and then spread to the Americas [7].

Currently, Brazil holds the position as the world’s largest coffee producer, contributing approximately 37% of the global supply [3]. Indonesia ranks fourth, accounting for over 8% of global coffee production [3, 6]. The cultivation of Coffea arabica in Indonesia was first introduced by the Dutch colonial administration, with plantations established across major islands such as Sumatra, Sulawesi (formerly known Celebes), and, to a lesser extent, Java. The total area under coffee cultivation exceeds 1.3 million hectares [6]. Nationally, the average productivity of arabica coffee is estimated at 0.55 tons per hectare, with an annual production volume of approximately 709,000 tons [6, 14].

Within Indonesia, the province of Aceh, located on the island of Sumatra, ranks as the third-largest coffee-producing region after South Sumatra and North Sumatra, with a total production of approximately 794.8 thousand tons in 2022 [15]. The Gayo Highlands in Aceh (Fig. 1) serve as a key arabica coffee-producing region, covering over 100,000 hectares across three districts: Aceh Tengah, Bener Meriah, and Gayo Lues, situated at elevations ranging from 900 to 2,000 m above sea level. Uniquely, about 78% of the land in this region is owned by smallholder farmers, who play a central role in cultivation and production. Such ownership pattern contrasts with the general trend in other parts of Indonesia, where coffee plantations are typically owned by private enterprises or state-owned companies [14]. The region's average yield ranges between 700 and 800 kg per hectare [15, 16]. These figures highlight the significant role of coffee as a primary source of livelihood for smallholder farmers in Indonesia, particularly on the islands of Sumatra, Sulawesi, and Java [6]. In highland areas, C. arabica dominates coffee production, accounting for nearly 90% of the total output, underscoring its economic importance in supporting rural communities.

Fig. 1.

Geographical location of Aceh Province on Sumatra Island. The dashed red oval line indicates the position of Sumatra within the Indonesian archipelago, meanwhile the red star marks the location of the Gayo Highlands

Genetic diversity study in Coffea

Baltazar et al. [17] conducted a genetic diversity analysis on 27 Coffea arabica accessions cultivated at high elevations in the Philippines, representing three varietal groups: typica, bourbon, and a typica–bourbon combination. Their study identified a total of 56 alleles, with an average of 3.7 alleles per locus. Among the 19 microsatellites or Simple Sequence Repeat (SSR) markers used, 80–83% were found to be polymorphic. Although the overall genetic diversity of the Philippine C. arabica collection was classified as low, the relatively high number of alleles and polymorphism information content (PIC) values provided valuable insights. These genetic parameters enhance the potential for selecting promising parental lines for future hybridization and breeding programs [17].

Simple Sequence Repeat (SSR) markers, also known as microsatellites, are tandemly repeated DNA motifs consisting of one to six nucleotides [18]. These markers are widely used in plant genetic diversity studies due to their favourable characteristics, including multi-allelic nature, high variability, codominant inheritance, chromosome-specific distribution, and broad genome coverage—including organellar genomes. Additionally, SSR markers are amenable to automation and exhibit a higher capacity to detect polymorphisms compared to other marker systems [19].

A recent study on Coffea arabica in Indonesia by Dianingsih and Mawardi [20], who had utilized five SSR markers to assess the genetic diversity of 32 accessions collected from Wamena, located in Papua, the easternmost region of Indonesia. Their findings indicated a high level of heterozygosity, likely due to gene flow or genetic exchange among coffee genotypes in the field. A total of 30 alleles were successfully amplified, with DNA fragment sizes ranging from 175 to 553 base pairs. Among the markers, CarM101 exhibited the highest polymorphism (0.82), while the lowest one (0.29) was performed by CarM052. Cluster analysis using bootstrap methods revealed six distinct genetic groups among the accessions. These findings underscore the genetic richness of C. arabica in Papua, which holds significant value for conservation and natural breeding, especially given the increasing threat of deforestation in the region.

Cubry et al. [21] conducted a comprehensive investigation into the genetic structure of the Coffea genus using 60 highly polymorphic Simple Sequence Repeat (SSR) markers. The study analyzed 42 individuals representing 15 Coffea species, primarily from Africa, including C. canephora, C. eugenioides, C. heterocalyx, C. liberica, C. anthonyi, and C. pseudozanguebariae. Their results highlighted the effectiveness of SSR markers in assessing genetic diversity and demonstrated their high cross-species transferability within the genus. Notably, the study confirmed a close genetic relationship between C. anthonyi and C. eugenioides, two morphologically similar species of distinct geographic origins; Cameroon and East Africa, respectively. Furthermore, the authors identified the potential utility of SSR markers in targeting genomic regions associated with sucrose metabolism, which tend to exhibit low genetic variability, thus opening new avenues for functional genomic research in coffee.

In a separate study, Lashermes et al. [12] employed Random Amplified Polymorphic DNA (RAPD) markers to assess the genetic diversity of C. arabica accessions (N = 16 accessions) collected from Ethiopia, Kenya, and Yemen under the ORSTOM–UN FAO program. Their findings revealed two major genetic groups with moderate to high genetic distances (ranging from 0.70 to 1.00). In a subsequent study, the same research group expanded their analysis to 20 cultivated and wild C. arabica accessions [13]. They reported (a) low levels of molecular polymorphism within C. arabica; (b) substantial differentiation between germplasm from the southwestern highlands and southeastern regions of Ethiopia; and (c) phenotypic divergence in F₁ hybrids, likely due to heterosis effects.

Indonesia being known as the fourth largest coffee producer globally, anticipates a threefold increase in production by 2050 [14], with domestic consumption projected to grow at an annual rate of 4.5–9% [3]. Coffee remains a critical source of livelihood for smallholder farmers, with over 60% of the national yield directed toward export markets [14]. Nevertheless, national average yields remain relatively low, and the sector faces ongoing challenges related to climate change, pests, and diseases [22].

In the Gayo Highlands of Aceh, C. arabica cultivation has a long-standing history dating back to the nineteenth century. However, comprehensive studies on the genetic diversity of local arabica populations remain scarce [6]. The only available report, by Harnelly et al. [23], focused exclusively on morphological variation between arabica and robusta accessions, examining traits such as stem, leaf, flower, fruit, and bean morphology. No molecular-level assessments have yet been conducted. Given the dominance of a few elite cultivars and the region’s propagation practices, it is hypothesized that a genetic bottleneck has reduced overall genetic diversity.

Prior to the Coffee Leaf Rust (CLR) outbreak in the early twentieth century, between 10 and 19 exotic C. arabica varieties were still cultivated in the Gayo Highlands, including the well-known “S-lines” (S288, S795, S1934) [24]. These cultivars were renowned for their superior cup quality and adaptation to high-altitude environments (> 1,200 m asl), but were highly susceptible to CLR [25]. Efforts to enhance their disease resistance through introgressive breeding were later initiated in India [11]. Following the CLR pandemic, a significant shift in varietal composition occurred. Presently, arabica cultivation in the Gayo Highlands is largely based on three Catimor (Caturra × HdT) derived cultivars which are usually known to be compact, disease-resistant types similar to C. canephora itself, namely the Gayo I and II. Both was officially released by the Indonesian Ministry of Agriculture. Meanwhile, the so called Gayo III- being recently released in year 2020 was also still related with the previous two varieties (pers. comm., Ishar, 2023 who serves as the Head of the Coffee Genetic Resources and Local Gene Bank (IP2TP), Middle Aceh). They are still the descendants of Hybrido de Timor (HdT) [26], which is known as the only interspecific hybrid derived from C. arabica and C. canephora, and this also includes the Catimor [24, 25, 27].

In light of the limited molecular data and concerns over declining genetic variability, this study aims to evaluate the genetic diversity of C. arabica in the Gayo Highlands using a two-tiered approach: agro-morphological characterization and molecular analysis based on SSR markers. While morphological traits provide useful phenotypic insights, they are often influenced by environmental factors and gene interactions, such as pleiotropy and epistasis [5]. In contrast, molecular markers offer greater specificity, reproducibility, and resolution [17]. By elucidating the extent of existing genetic variation and identifying promising parental lines, this research seeks to inform future breeding strategies, support the development of robust F₁ hybrids, and contribute to long-term germplasm conservation and genetic resource management. This work will be of particular relevance to breeding programs and conservation initiatives aimed at ensuring the sustainability of arabica coffee cultivation in Indonesia and other tropical highland regions.

Materials and methods

Materials

A total of 52 Coffea accessions were used in this study (Supplement 1), of which 98% were identified as C. arabica. These samples were collected from various altitudes in Middle Aceh, Indonesia, ranging from 700 to 1,500 m above sea level (Fig. 1). The majority of the accessions (65%) were obtained directly from farmers'fields and community-managed garden coffee systems [28], while the remaining 35% were sourced from the Agricultural Technology Assessment and Research Installation [Instalasi Penelitian dan Pengkajian Teknologi Pertanian (IP2 TP)] or also known as the local coffee gene bank at Bener Meriah.

Methods

Morphological analyses

A total of 32 morphological traits; covering overall plant performance, branching and fruit characteristics, leaf color, and other descriptors (Supplement 2)—were assessed to characterize C. arabica accessions from Middle Aceh. Trait selection was based on the standardized coffee plant descriptor manual [29]. Morphological data were collected directly from field-grown plants to complement the molecular characterization conducted using Simple Sequence Repeat (SSR) markers. Of the 52 accessions initially sampled, one C. canephora (robusta) accession was excluded, resulting in 51 C. arabica accessions used for the final statistical analyses.

The data were categorized into two main groups:

1. Qualitative traits (Supplement 2, Section A.), which included descriptors e.g. shape, form, and color. These traits (No. 1 up to 13) were evaluated through field observations and scored on an ordinal scale (1 to 4), with each score corresponding to specific descriptor definitions. The values were then used for subsequent statistical analyses (see "Statistical Analyses" Section S).

2. Quantitative traits, accounting for approximately 58% of the total traits (Supplement 2, Section B.) were measured directly in the field. These included variables such as plant height, leaf dimensions, stem and fruit thickness, (No. 1 up to 19) and were statistically analyzed under the assumption of normal distribution (Supplement 4).

Field equipment and measurement tools included cotton thread and a 20-m measuring tape for assessing tree diameter and plant height. A vernier caliper (New Deland, 150 × 0.02 mm, China) was used for precise measurements of leaf, stem, and fruit thickness, as well as leaf length and width. A mechanical tally counter was employed to count leaves, fruits, and branches. Field data were recorded using standard writing tools and subsequently digitized in Microsoft Excel for further statistical processing, including descriptive and multivariate analyses.

Genetic diversity analyses via microsatellite marker

Fifty-one (N= 51) Coffea arabica L. accessions were selected for molecular analysis using young leaf tissues as the DNA source. Approximately ten leaves were collected per accession, cleaned with tissue moistened with 70% (v/v) ethanol, wrapped in newspaper, and stored in paper envelopes containing 10–15 g of silica gel for desiccation. Samples were then shipped to the Plant Biology Laboratory of the National Research and Innovation Agency (BRIN), located in Cibinong, West Java, Indonesia (6.54° N).

B.1. DNA Extraction

Genomic DNA was extracted following the Doyle and Doyle [30] protocol with slight modifications. After adding 1–2 mL of chilled isopropanol, the samples were incubated at –20 °C for one hour. Due to the high antioxidant content in coffee leaves, the resulting pellets were washed twice with 70% ethanol (v/v). After drying, the pellets were suspended in 190 µL of TE buffer containing RNase (10 mg/mL) and incubated at 37 °C for 30 min, followed by two heat shocks at 65 °C for five minutes each. To enhance DNA purity, 100 µL each of TE buffer, distilled water, 5 M NaCl, and 0.5 M EDTA were added. Samples were homogenized, incubated on ice for 30 min, and centrifuged at maximum speed for 5 min. The supernatant was transferred to a new tube, and DNA was precipitated using 100 µL of cold isopropanol. After drying at room temperature, the DNA was resuspended in 60 µL of TE buffer and stored at 4–5 °C until further use.

B.2. Polymerase Chain Reaction (PCR) Amplification

The quality of extracted DNA was assessed using a NanoPhotometer® P-Class P-360 (Implen GmbH, Germany). DNA purity, meanwhile, was confirmed using the A260/A230 absorbance ratio, with values ≥ 1.80 considered acceptable [31]. PCR amplification was conducted in a 25 µL reaction volume containing 100 ng of genomic DNA, 0.4 µM each of forward and reverse SSR primers (Genetika Science, Indonesia; Table 1), 75 µM of each dNTP, 2.5 mM MgCl₂, 1 × TBE buffer, and 1 unit of Taq DNA polymerase (Invitrogen, USA).

Table 1.

Microsatellite markers information (F and R) applied in this study based on previous studies [32–34]

No.	Primer ID	Forward Primer (F) (51 > 31)	Reverse Primer (R) (51 > 31)	Reference
1	Sat 32	AAC TCT CCA TTC CCG CAT TC	CTG GGT TTT CTG TGT TCT CG	Gimase et al. [32]
2	Sat 207	GAA GCC GTT TCA AGC C	CAA TCT CTT TCC GAY GCT CT
3	Sat 227	TGC TTG GTA TCC TCA CAT TCA	ATC CAA TGG AGT GTG TTG CT
4	Sat 240	TGC ACC CTT CAA GAT ACA TTCA	GGT AAA TCA CCG AGC ATC CA
5	Sat 255	AAA ACC ACA CAA CTCT CCT CA	GGG AAA GGG AGA AAA GCTC
6	M 24	GGC TCG AGA TAT CTG TTT AG	TTA ATG GGC ATA GGG TCC
7	A8847	GCA CAC ATG AAA AAG ATG CT	GAT GGA CAG GAG TTG ATG G	Rovelli et al. [34]
8	AY2434	CGC AAA TGT TTA TGT CAA TC	GCA ACT TAT GAG CCT AAT CC	Cristancho et al. [33]

Thirteen SSR primer sets were initially screened, but only eight showed consistent and reproducible amplification. Preliminary trials with six markers (M24, Sat32, Sat207, Sat227, Sat240, and Sat225) (Supplement 3) were conducted using two randomly selected C. arabica samples. Amplification was performed in a T100 Thermal Cycler (Bio-Rad) under the following conditions: initial denaturation at 94 °C for 5 min; followed by 35 cycles of denaturation at 94 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 90 s; with a final extension at 72 °C for 10 min.

B.3. Gel electrophoreses, genotyping and data analysis

Amplified products were resolved on 6% agarose gels using a TBE buffer system and visualized via silver nitrate (AgNO₃) staining. Electrophoresis was performed using a vertical tank system (Cleaver Scientific) with CS-300 V power supply. Band presence (1) or absence (0) was scored manually for each accession. Meanwhile, the allelic peaks were analysed using PEAK Scanner v1.0 (Applied Biosystems) with reference to the GeneScan 500 LIZ size standard. Each peak was considered a codominant allele, and genotypes were recorded accordingly. The resulting binary matrix was used for cluster analysis in NTSYSpc v2.2 (Exeter Software, USA). A genotype accumulation curve (Fig. 2) was also generated to assess the adequacy of marker coverage. Genetic relationships among accessions were visualized via an Unweighted Pair Group Method with Arithmetic Mean (UPGMA) dendrogram using the Simple Matching coefficient (Fig. 4).

Fig. 2.

Genotype accumulation curve based on eight SSR markers, illustrating the relationship between the number of alleles sampled (x-axis) and the number of unique Coffea arabica genotypes identified (y-axis)

Fig. 4.

Cluster dendrogram of 51 Coffea arabica accessions from the Gayo highlands based on the Nei’s genetic distance and UPGMA (unweighted pair-group method) clustering using eight SSR markers

Statistical analyses

Two types of data were analysed in this study: morphological and molecular. For the morphological data, quantitative traits were measured in replicates and averaged before statistical processing. Descriptive statistics, including minimum, maximum, and mean values, were calculated, and the normality of the data distribution was assessed using histograms (Supplement 4). To explore patterns of phenotypic variation and reduce data dimensionality, Principal Component Analysis (PCA) was conducted using JMP Pro 16 (SAS Institute Inc., university-licensed version). PCA facilitated the identification of key traits contributing to variation among accessions, supporting the selection of promising individuals for breeding programs [31, 35] as well as by the sensory analyses [26]. The results are presented in a two-dimensional PCA biplot (Fig. 3).

Fig. 3.

Principal Component Analyses (PCA) based on correlations (Statistical Software JMP ver. 16) of the morphological characteristics (N= 51 samples); A Summary plots of eigenvalue; B A biplot diagram, where 96% of the total accessions was clustered in pink (left); while only three accessions (acc. no. 40,41,42) were separated on different cluster; C the major quantitative characteristics identified by the PCA

For the molecular data, a binary scoring matrix based on the presence (1) or absence (0) of SSR marker bands was used to assess genetic relationships among accessions. This matrix was subjected to cluster analysis, and a dendrogram illustrating genetic similarity was constructed using the Unweighted Pair Group Method with Arithmetic Mean (UPGMA) and the Simple Matching coefficient (Fig. 4). Further methodological details are provided in Section B.3.

Results

To assess the genetic diversity of Coffea arabica cultivated in the Gayo Highlands, we analysed 51 accessions collected from various altitudes in Aceh, at the northern Sumatra. A dual approach integrating both morphological and molecular data allowed for a comprehensive evaluation of phenotypic variability and genetic structure among the accessions. This approach is particularly valuable, given that morphological characterization of perennial crops such as Coffea spp. can be time-consuming, labour-intensive, and costly across vegetative and reproductive phases.

Field observations revealed that the first onset of flowering (S.4.1) occurred, on average, at 5.7 months after planting, nevertheless some of them showed quite late flowering period, namely after 9 months. Farmer interviews and questionnaires provided additional context: (i) the majority of farmers lacked information on the original source of their planting material, often acquiring seedlings through government assistance programs or informal farmer-to-farmer exchanges; (ii) many farmers reported negative impacts of rising temperatures over the past 20–30 years, including increased vulnerability to abiotic stress and heightened incidence of pests and root-borne fungal diseases.

Morphological variation in Gayo Arabica Coffee

Morphological evaluation of the 51 C. arabica accessions was based on 32 traits, categorized into qualitative and quantitative types (Table 2). The further explanation of Table 2, however, should be aligned with those histograms presented on Supplement 4 (stated in the right column belongs to Table 2). Notably, C. canephora (robusta) accessions showed clear morphological differences from C. arabica, particularly in leaf morphology, where the length-to-width ratio in robusta leaves was up to 60 times greater than that in arabica.

Table 2.

Summary of the qualitative and quantitative traits of Gayo arabica coffee (N = 51 accessions)

No.	Morphological Characteristics		Symbol vs. Definition
A	Qualitative	1	2	3	4	5
1	Plant Appearance: Growth habit	bushes	Ø dwarf but having many branches	tree with big & already extended branches
2	Plant Appearance: Overall performance	conical shape	Ø pyramid shape	undefined bushes
3	Vegetative growth	monopodial	Ø sympodial
4	Branching characteristics	few major primary branches exist	many primary and secondary branches;	Ø primary, secondary, and tertiary branches exist
5	Angle between the primary & secondary-branches	Ø drooping up to 90o	horizontal 180o	semi erect 45o
6	Stipule shape	circle	oval	Ø triangle	Ø equal triangle	trapezium
7	Leaf color	Ø green	brown	reddish	bronze	other
8	Leaf shape	type 1	type 2	Ø type 3	type 4	type 5
9	Petiole color	Ø green	brown	reddish	bronze	other
10	Inflorescence position	axillary	Ø terminal
11	Cherry color	Ø green	brown	Ø red	bronze	other
12	Cherry (fruit) shape	type 1	type 2	type 3	type 4	Ø oval thinned
13	Disease vulnerability against	Ø root fungi	bacteria	virus	insects	others
B	Quantitative	Unit	Min	Mean (μ)	Max	Supplement File
1	The first onset of flowering	months	5.6	5.7	8.8	see: S.4.1
2	Plant height	cm	125.0	169.0	350.0	see: S.4.2
3	No. of secondary branches per primary branch	unit	7.0	12.0	42.0	see: S.4.3
4	Stem diameter	cm	3.5	5.7	11.5	see: S.4.4
5	Internode length on the primary branch	cm	2.0	4.4	12.0	see: S.4.5
6	The distance between the first upper branch and the top of plant	cm	0.0	7.7	13.5	see: S.4.6
7	Canopy diameter	cm	85.0	223.8	300.0	see: S.4.7
8	No. of leaves per branch	unit	17.0	359.0	1,467.0	see: S.4.8
9	Leaf length	cm	2.0	12.6	18.5	see: S.4.9
10	Leaf width	cm	2.3	5.8	9.7	see: S.4.10
11	No. of petals per flower	unit	2.0	4.9	6.2	see: S.4.11
12	No. of cherries per side branch	unit	11	291	1,453.0	see: S.4.12
13	Cherry length	cm	1.0	1.5	2.3	see: S.4.13
14	Cherry width	cm	0.6	1.2	3.5	see: S.4.14
15	Cherry thickness	cm	0.6	1.1	3.5	see: S.4.15
16	Petiole length	cm	0.5	1.2	3.7	see: S.4.16
17	Flowering period	months	8.0	15.0	18.0	see: S.4.17
18	Leaf number per axil	unit	2.0	11.0	23.0	see: S.4.18

Ø = the majority one (more than 60% occurrence based on observation)

Principle component analyses based on correlations

Principal Component Analysis (PCA) was employed to reduce the dimensionality of the morphological dataset and identify major traits contributing to variation (Fig. 3). The first two principal components (PC1 and PC2) explained 32.6% and 13.4% of the total variation, respectively (Fig. 3A). PC1 was primarily associated with leaf-related traits, including leaf number per axil, leaf length, and leaf width, while PC2 was strongly correlated with yield-related traits such as side branching, flowering period, and fruit yield per tree.

The PCA biplot, moreover, revealed two distinct clusters among the accessions. Approximately 94% of the accessions were grouped in a single cluster (colour line: pink) (Fig. 3.A), while only three accessions: Gayo I, II, and III from the IP2TP gene bank had formed a distinct cluster (blue). These elite lines were characterized by superior agronomic performance, including high yields (up to 6 kg per tree), larger leaf numbers (1,468; 629; 682 leaves), and high cherry production (397, 340, 530 cherries per side branch). Conversely, the accessions grouped in the main cluster exhibited broader morphological diversity, particularly in traits such as leaf size, flowering time, and cherry number.

Genetic diversity generated by microsatellite marker

Genomic DNA was successfully extracted from 51 arabica accessions, with DNA purity values exceeding the acceptable threshold (A₂₆₀/A₂₃₀ > 1.7). Eight SSR markers were selected based on their clear and reproducible amplification profiles, generating DNA fragments ranging from 100 to 285 bp. A genotype accumulation curve (Fig. 2) demonstrated that genetic diversity was effectively captured with the selected markers, as the number of unique genotypes plateaued after the second allele, indicating sufficient marker coverage.

Each SSR marker detected two to four polymorphic bands. Markers Sat227 and Sat240 produced 2–3 bands, while M24, Sat32, and A8847 revealed an average of 4.5 bands per marker. Sat207, Sat255, and AY2434 each produced three bands, with Sat255 classified as moderate in polymorphism based on prior literature [36]. The polymorphism information content (PIC) values ranged from 0.157 (AY2434) to 0.61 (M24), with an overall mean of 0.41 (Table 3). Observed heterozygosity (Ho) ranged from 0.00 (AY2434) to 1.00 (M24, A8847), reflecting varying levels of allelic diversity across loci.

Table 3.

H_e expected heterozygosity, H_o observed heterozygosity, Polymorphic Information Content (PIC)

No.	Marker	No. of Alleles	No. of Individuals	Observed heterozygosity	Expected heterozygosity	Polymorphic Information Content (PIC)
1	M24	4	45	1.000	0.667	0.610
2	Sat32	4	43	0.977	0.658	0.591
3	Sat207	3	38	0.947	0.540	0.427
4	Sat227	2	39	0.000	0.450	0.346
5	Sat240	2	42	0.381	0.569	0.519
6	Sat255	3	32	0.406	0.335	0.285
7	A8847	4	42	1.000	0.573	0.471
8	AY2434	3	34	0.000	0.167	0.157

Genetic distance and cluster analyses

A UPGMA dendrogram (Fig. 4) constructed from SSR-derived data revealed three major genetic clusters, each subdivided into two subgroups. Cluster 1 contained 37% of the accessions, including elite lines such as Gayo I, II, and III, grown at high elevations (> 1,200 m asl), and traditional landraces like Orange Bourbon and OR_PCM (locally known as Orange Pucuk Merah). Accessions in this group generally exhibited dwarf growth habits (< 180 cm), high leaf numbers (mean: 260 per branch), and elevated cherry counts (up to 231 cherries per branch).

Cluster 2, comprising 58% of the accessions, was divided into sub-clusters 2.1 and 2.2. This cluster included farmer-preferred varieties such as Ateng Janda, Abessinia, Caturra (C49, C50), CH 360, CTT, Margogype (types 1 and 2), USDA 62, Arabica Mocca Big, Yellow Caturra, and Bourbon. These accessions were generally characterized by high productivity, exhibiting 72% more cherries and 69% more leaves per branch compared to Cluster 1 accessions, while also maintaining compact plant architecture. Only two Mocca Yaman accessions (large and small bean types) were being clustered in 3 (Cluster 3), and they were genetically distinct with the rest. By comparing their morphological characteristics with the others, these two showed exceptional traits, including extra ordinary high cherry production per branch (31–47 cherries), high leaf numbers (46–53 per branch), and greater plant stature, namely up to 84% taller than the average height of accessions belong to other clusters.

The overall genetic distance among the 51 arabica accessions ranged from 0.00 to 0.36. The shortest genetic distance (0.20) was recorded between Clusters 1 and 2, indicating moderate similarity. In contrast, unpublished data suggest a much wider genetic divergence (0.52) between C. arabica and C. canephora (robusta), consistent with the mean PIC value of 42.57% (Table 3). These findings confirm the genetic distinctiveness between the two species and underscore the limited but structured diversity within Gayo C. arabica germplasm.

Discussion

To the best of our knowledge, this is the first comprehensive study to evaluate and characterize the genetic diversity of Coffea arabica cultivated in the Gayo Highlands of Aceh since its introduction over a century ago to Indonesia, that emphasized the molecular data. This research integrates both morphological and molecular approaches to assess the magnitude of genetic variation, offering valuable insights for germplasm management, cultivar identification, and future breeding strategies [16, 37].

The morphological evaluation of 51 accessions revealed significant diversity across key traits, including plant height (ranging from 100 to 250 cm), secondary branch number (> 40), and leaf shape and size. This variation reflects the diversity of local cultivars, such as Ateng Super (Ateng S), P88, and Caturra-derived lines (e.g. C41, C47, C48), which are highly favoured by local farmers. Ateng S, in particular, is recognized for its high yield potential (0.91–1.76 tons/ha), larger bean size, and early fruiting [36]. However, its productivity is known to fluctuate across growing seasons, raising concerns about its long-term viability [25].

The observed phenotypic variation may be attributed to several factors. First, C. arabica is an allotetraploid (2n = 4x = 44) species that likely originated through natural hybridization involving unreduced gametes from C. canephora, resulting in gene flow and structural genome changes [38, 39]. Polyploidy is often associated with enhanced fitness, morphological variability, and environmental adaptability due to genome-wide interactions and gene duplication. Second, spontaneous mutations in major genes may contribute to local adaptation, particularly in highland environments [38, 39].

Our findings align with previous reports highlighting the narrow genetic base of commercial C. arabica populations worldwide [4, 7]. It is believed that most modern arabica cultivars derive from a limited number of trees, including descendants of C. arabica var. typica and C. arabica var. bourbon, which spread from Ethiopia and Yemen to Latin America, Asia, and Indonesia. This restricted genetic base necessitates targeted efforts to conserve and expand genetic resources.

In this study, although we employed a relatively limited number of SSR markers (N = 8), the analysis was supported by a diverse panel of 51 accessions and a comprehensive morphological assessment involving 32 traits. Genotype accumulation analysis indicated that the selected SSR markers were sufficient to capture the genetic diversity present, as the accumulation curve reached a plateau, suggesting a saturation point had been achieved [40, 41]. A similar approach was used by Montagnon et al. [4], who applied 8 SSR markers across 137 accessions, successfully clustering global arabica germplasm into well-defined genetic groups.

Over the past three decades, DNA markers have played a pivotal role in genetic diversity studies, phylogenetics, and marker-assisted selection (MAS) [37]. While previous studies have utilized between 8 and 60 SSR markers [4, 8, 21, 42], the number of markers required may vary depending on the diversity of the germplasm and the resolution needed. Importantly, SSR markers offer several advantages: they are co-dominant, highly polymorphic, robust, cost-effective, and suitable for differentiating wild and cultivated germplasm across diverse regions—from Colombia and Ethiopia [7, 28, 42], to Brazil [7], Yemen [4, 7], Kenya [32], and China [43]. However, to date, no such analysis has been reported for accessions from the Gayo Highlands.

The SSR markers used in this study yielded PIC values ranging from 0.157 to 0.61, with M24 showing the highest PIC and observed heterozygosity (Ho = 1.00), consistent with previous studies [36]. M24 was particularly effective in distinguishing between C. arabica var. typica and C. arabica var. bourbon, which followed different introduction routes in the early seventeenth century [7]. When Ho exceeds He (expected heterozygosity), it may indicate ongoing outcrossing or gene flow within populations [44], though further exploration with additional markers would enhance resolution.

Despite its high productivity and adaptation to local conditions, the genetic base of commercial Gayo arabica cultivars appears narrow, as evidenced by the SSR and PCA results. This is consistent with previous findings on arabica populations in Ethiopia and elsewhere [13, 34]. The clustering of accessions into two major groups and a maximum genetic distance of 0.36 underscore this limited diversity. Narrow genetic bases can hinder breeding progress and increase vulnerability to biotic and abiotic stressors.

Reliable identification of germplasm and tracing of geographic origins are essential for breeding, conservation, and commercialization [45]. Several locally valued and potentially sub-spontaneous cultivars, such as Ateng Super and Ateng Janda, were identified in this study. These cultivars may trace their ancestry to Hybrido de Timor (HdT) and could have evolved through natural selection or human intervention [12]. Despite showing desirable traits such as yield improvements up to 85%, they remain unstable across seasons [25]. Given the growing challenges of climate change, deforestation, and disease pressure, the identification and utilization of such well-adapted local germplasms e.g., accessions 40, 41, and 42—should be prioritized in future breeding programs.

In conclusion, this study confirms the presence of phenotypic and genetic variation among C. arabica accessions from the Gayo Highlands, though the overall genetic base remains narrow. Expanding and conserving this genetic diversity is critical for the sustainability of coffee production in Aceh and beyond. Future efforts should focus on incorporating underutilized or locally adapted germplasm into breeding pipelines to enhance resilience, maintain quality, and support smallholder livelihoods in the face of global agricultural challenges.

Conclusion

This study provides the first comprehensive scientific evidence of the genetic diversity present in Coffea arabica L. germplasm from the Gayo Highlands, Middle Aceh, Sumatra. Our integrated approach—combining agro-morphological characterization with molecular analysis using eight microsatellite (SSR) markers revealed a high degree of morphological variation across 33 traits. However, this was contrasted by a relatively narrow genetic base, as indicated by the limited allelic diversity and low polymorphism values observed in the molecular data.

The discrepancy between phenotypic and genetic findings may be attributed to environmental influences and genotype-by-environment (G × E) interactions, which can obscure underlying genetic relationships and create phenotypic biases, particularly in long-lived perennial crops such as coffee. The use of Principal Component Analysis (PCA) helped to distil complex trait data, with the first principal component (PC1) accounting for 35.6% of the total variation, mainly driven by leaf and yield-related traits.

The low level of genetic diversity detected in this study raises critical concerns for the long-term sustainability of arabica coffee cultivation in the Gayo Highlands. In light of increasing threats such as deforestation, climate change, and habitat degradation, immediate attention is needed to strengthen in situ and ex situ conservation efforts. Enhancing the genetic base through the identification and incorporation of underutilized local and exotic germplasm into breeding programs will be essential to secure the future resilience and productivity of Indonesia’s highland coffee systems.

Authors’ contributions

Rita Andini (R.A.): material preparation, morphological analysis, marker selection, wrote majorly the main manuscript text. Ahmad Zaelani (A.Z.): the main responsible person in the molecular analyses and the marker selection. Muhammad Ikhsan Sulaiman (M.I.S.): concept and design of the project, field collection and wrote partially the main manuscript text. Enny Rimita Sembiring (E.R.S.): genetic analysis and marker selection. Deepak Bhanot (D.B.), Meenakshi Gusain (M.G.), Roma Sarnaik Morghade (R.S.M), Abdel-Rhman Z Gaafar (A.Z.G.): molecular marker analysis, editing and revise and proof-read manuscript. Rachman Jaya (R.J.): assisting part of the field collection and morphological analysis. Abdul Aziz (A.A.): assisting part of the field collection. Heru Prono Widayat (H.P.W.): head of the project, corresponding author, designed the project.

Authors’ information

Rita Andini, Muhammmad Ikhsan Sulaiman, Heru Prono Widayat, Rachman Jaya, and Ahmad Zaelani in February 2020 were granted a research grant from the USK. She was started to be employed at USK in June 2018 and was conducting the coffee research at that institution until December 2021. Since January 2022, she has been employed as a researcher in National Agency of Research and Innovation (BRIN) and continued the coffee research with the assistance of Ahmad Zaelani at BRIN, who has been earlier employed, there.

Funding

Funding for the research was from Universitas Syiah Kuala (Title: Peningkatan kualitas kopi Gayo Aceh melalui karakterisasi genetik dan sensori No. 529/UN11/KPT/2022).

Data availability

The leaf sample collection for the DNA studies in our manuscript can be accessed publicly at the IP2TP (the local coffee gene bank at Bener Meriah, Middle Aceh, Aceh province in Indonesia). The supporting letter from its leader has been also included here at the journal's website during our submission on 20th May 2025.

Declarations

Ethics approval and consent to participate

The authors declare that experimental research works on the plants described in this paper comply with institutional, national and international guidelines. The use of plant material (coffee leaf samples) has been officially permitted by the Head of IP2TP Gayo in Bener Meriah (Mr. Ishar, SST). Moreover, voucher specimens have been deposited in public collection institution (IP2TP) and it has also provided access to deposited material. The official permission was also granted when we conducted the sampling on public land and those mentioned villages (official letter signed by Mr. Ishar, SST has been submitted, too).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.