“He ‘uala ka ‘ai ho‘ōla koke i ka wī”,

The sweet potato is the food that ends famine quickly.

-‘Ōlelo no‘eau (Pukui 1983)

Introduction

Sweet potato (Ipomoea batatas [L.] Lam., ‘uala in Hawaiian) is a member of the plant family Convolvulaceae and is an important crop worldwide, with more than 105 million metric tons produced yearly (CIP, n.d.). Sweet potato originated in a region spanning Central and South America, where it diverged from its closest wild relative I. trifida ~ 800,000 years before present (Yang et al. 2017), followed by multiple hybridization events with I. triloba to form the hexaploid I. batatas (Muñoz-Rodríguez et al. 2018). From there, the crop found its way to all parts of the world including the Hawaiian archipelago. The most well-supported explanation for its dispersal across Oceania is the “tripartite hypothesis” (Barrau 1957; Green 2005; Yen 1974), which posits that an early Kumara sweet potato line was transferred from the South American coast (Scaglion 2005) to central Polynesia around 1000–1100 C.E. and diffused outward with Polynesian voyagers (Athens et al. 2014). Recent findings by Ioannidis et al. (2020) of native South American gene flow into Polynesia around this time provide added support for this exchange of people and plants. Two other lines were transported ~1500 C.E. by Spanish and Portuguese traders, the Camote line from Mesoamerica and the Batata line from the Caribbean. The geographical patterning suggested by this hypothesis is supported by genetic analysis of extant global collections (Roullier et al. 2013). For the Hawaiian Islands, the earliest archeological record of sweet potato dates to the fourteenth century C.E. in the rain-fed Kohala Field System (Ladefoged et al. 2005), where it served as the key staple in the large dryland field systems of both Hawai‘i Island and Maui (Coil and Kirch 2005; Kirch et al. 2009; Vitousek et al. 2004) and supported populations in arid and coastal areas across the archipelago (Handy 1940).

Tremendous crop diversity is suggested by the ~300 documented Hawaiian names for sweet potato cultivars, suggesting the development of numerous locally-adapted cultivars (Handy 1940; Kagawa-Viviani 2016). This diversity also includes European introductions of sweet potato cultivars new to Hawai‘i, which likely started arriving shortly after Cook’s 1778 arrival (Nagata 1985). By the time the first extensive lists and descriptions were printed ~80 years later (e.g., Kaaie 1860; Napihelua 1857; Rooke 1855), introduced cultivars were already established. Old varieties continued to be valued for medicine and ceremony (Fornander 1919; Kaaiakamanu and Akina 1922), but recognition of the loss of ethnobotanical knowledge motivated collections in the early twentieth century (Handy 1940). The twentieth century also saw an acceleration of foreign introductions, breeding, and clonal selection of both natural and artificial sports to support commercial sweet potato production (Chung 1923; Poole 1952, 1955a, 1955b, 1959; Takahashi 1937). We currently know little about the extent to which traditional varieties still persist across the Hawaiian Islands. At present, more than 30 accessions identified as traditional cultivars are maintained across a half dozen conservation-oriented botanical collections on the islands of O‘ahu, Maui, and Hawai‘i. We use the term “traditional” in this text as a synonym for “heirloom” and “heritage.” We reserve the term “Hawaiian” or “traditional Hawaiian” for cultivars we assume were developed and named by indigenous Hawaiian farmers from germplasm present in the islands prior to 1778, and we use “Hawai‘i” for cultivars presently found in the islands, regardless of origin.

To address this gap in knowledge, nuclear and chloroplast markers were utilized to analyze the genetic diversity among Hawai‘i, USDA NPGS accessions, and herbarium samples. The chloroplast genome encodes a variety of organelle-specific components and is highly conserved (Jansen et al. 2005). As a result, intracellular genetic markers can help illuminate complex evolutionary dynamics and reveal phylogenetic relationships in sweet potato through chloroplast loci analysis (Roullier et al. 2011; Wang et al. 2019; Yan et al. 2015). Additionally, single nucleotide polymorphism (SNP) markers have exceptional utility in many genetic applications, including germplasm characterization which encompasses population structure, genetic diversity and relationships (Ertiro et al. 2015). The diversity of sweet potato can thus be identified based upon SNP genotyping and phylogenetic relation analysis (Feng et al. 2018).

We also evaluated phenotypic diversity among traditional Hawaiian sweet potato cultivars by quantifying leaf morphology from herbarium voucher specimens collected in the early 1900s from Hawaiian farmers across the archipelago in an ethnobotanical survey. Although vine-type plants are known to demonstrate morphological variation throughout their development (heteroblasty) (Lee and Richards 1991), this has not been documented for Ipomoea batatas. Instead, Gupta et al. (2020) and Jackson et al. (2020) [and references therein] established that variation in sweet potato leaf morphology is driven by genetics, the growing environment, or both. While our use of herbarium voucher specimens did not allow us to control for either age or environment as in Gupta et al. (2020) and Jackson et al. (2020), the voucher specimens we analyzed provide an important snapshot of diversity.

Renewed interest in traditional cultivars has spurred efforts to bring indigenous crops and cultivars to commercial markets, in large part due to a potential 400% price premium (Oberholtzer et al. 2005). Indigenous cultural and language resurgence and in-situ conservation motivate recovery of these crops not only for the sake of preserving germplasm, but also to preserve names and stories. In turn, these reflect intrinsic value of traditional cultivars for consumers, adding value for growers.

Sweet potato germplasm present in Hawai‘i today is a confusing array of ancient cultivars, introductions from over 200 years of transfers, and an unknown degree of hybridization. Yet clarification and confidence in the identification of traditional cultivars can lead to improved production and value for local producers and practitioners (Kagawa-Viviani et al. 2018). The considerable work already done in another culturally important crop, taro (Helmkampf et al. 2017), largely aligns with traditional groupings. In this study, we sought to elucidate relationships among extant genotypes and historic collections by examining available genetic, phenotypic, and ethnographic data on sweet potato in Hawai‘i.

Materials and Methods

Plant Material

Ipomoea batatas DNA was obtained from various repositories: fresh plant tissues from gardens, markets, and collections across the Hawaiian Islands, germplasm originally sourced from the United States Department of Agriculture National Plant Germplasm System (USDA-NPGS), DNA sourced from a Hawaiian plant DNA library (Randell and Morden 1999), and herbarium voucher specimens. The fresh Hawai‘i samples included putatively Hawaiian traditional varieties maintained as botanical collections at Waimea Valley Arboretum (O‘ahu), UH Mānoa’s Harold L. Lyon Arboretum (O‘ahu), and Amy Greenwell Ethnobotanical Garden (Hawai‘i Island, also represented in the DNA library) and several local varieties from markets on O‘ahu. USDA-NPGS germplasm maintained in local facilities represented broad geographic diversity with varieties originating from the Americas, Asia, and Oceania/Indo-Pacific. Finally, pressed tissue samples from the UH Mānoa Joseph F. Rock Herbarium included I. batatas from Chinatown markets (downtown Honolulu, O‘ahu), Vietnam, and several Pacific Island groups. For a list of all samples see Electronic Supplementary Material [ESM], Appendix 1. DNA from fresh material was extracted with the DNeasy Plant Mini Kit (Qiagen, Valencia, California) and DNA from herbarium material with the QIAamp DNA Stool Mini Kit (Qiagen, Valencia, California.

Chloroplast Loci Sampling: Hawai‘i Accessions

DNA was obtained from 50 I. batatas samples including fresh plant tissues (28 Hawai‘i botanical gardens), DNA library (9, Amy Greenwell Ethnobotanical Garden), and dried material (13, UHM Joseph F. Rock Herbarium). For each of these 50 samples (see ESM, Appendix 1), two phylogenetically informative standard chloroplast loci (Graham and Olmstead 2000) trnL-trnF and ycf4 were amplified using PCR with primers designed using Geneious software (Kearse et al. 2012). The subsequent PCR products were sequenced on an ABI 3730 (Applied Biosystems, Foster City, California, USA) and aligned using Geneious software. The trnL-trnF and ycf4 sequences were used to develop phylogenetic trees through two approaches: 1) maximum likelihood phylogeny with the generalized time-reversible (GTR) substitution model (Tavaré 1986), and 1000 bootstrap replications and a transition/transversion rate of four using PhyML 3.0 (Guindon et al. 2010); and 2) Bayesian phylogeny using MrBayes version 3.2.6 (Huelsenbeck and Ronquist 2001) using a GTR substitution model with Ipomoea indica as the outgroup. For this, we used the following specifications: gamma rate variation with four categories, Markov Chain Monte Carlo (MCMC) chain length of 100,000 using four heated chains with a subsampling frequency of 500 and burn-in length of 10,000, and a starting random seed of 26,695 with an unconstrained branch length prior. Tree topologies of the two methods were compared for concordance.

Nuclear Loci Sampling: Hawai‘i and USDA Accessions

We analyzed nuclear genomic variation for 77 sweet potato samples representing both local and global diversity. Material was sourced from Hawai‘i botanical gardens (29), other local gardens and markets (10), the USDA-NPGS (30), and the UH Mānoa Joseph F. Rock Herbarium (8) (see ESM, Appendix 1). Following extraction, DNA was sent to Rapid Genomics (Gainesville, Florida, USA) for sequencing. Target probe regions were designed for 152 genes which were subsequently sequenced for each of the 77 samples and analyzed through a standard bioinformatics pipeline. To clean sequences, the 3′ end was trimmed by removing low quality bases (quality score < 20). Reads were discarded when more than 10% of the read had a quality score of less than 20. Cleaned reads were aligned with MOSAIK, and Freebayes (Garrison and Marth 2012) was used for identification of SNPs. Raw VCF and Filter1 VCF with vcftools (Danecek et al. 2011) were general using Filter 1: minQ 10, max-missing-count 3, min-alleles 2, max-alleles 2, min-meanDP 3, max-meanDP 750, maf 0.01, mac 1). This resulted in 27,195 SNP markers.

Genetic assignment was conducted on the sweet potato samples. First, a principal component analysis (PCA) was performed using SNPrelate (Zheng et al. 2012). Next, a discriminant analysis of principal components (DAPC) was conducted using the adegenet package in R (Jombart 2008; Jombart and Ahmed 2011), where genetic groups (K) of 2–6 were tested. Based on genetic distance, individuals were considered clones if they were 97% identical. This value was empirically derived by using individuals sourced from different places with the same names (ESM, Appendix 2).

Hawaiian ‘Uala Voucher Specimens: Leaf Shape, Naming, and Geography

We analyzed leaf morphology of 128 pressed voucher specimens of Hawaiian ‘uala (sweet potato) housed in the Bernice Pauahi Bishop Museum Herbarium (Honolulu, Hawai‘i). Most of the samples were collected between 1912 and 1931 from Hawaiian farmers on the islands of Ni‘ihau, O‘ahu, Moloka‘i, Maui, and Hawai‘i. Digital photographs of the voucher specimens were analyzed in ImageJ (Abràmoff et al. 2004; Schneider et al. 2012). For each sample, a single representative leaf was selected and evaluated for leaf area, perimeter, length from petiole to tip, width at widest point, leaf lobe count, and leaf shape following the definitions in Descriptors for Sweet Potato (Huamán, 1991, p. 56, Figs. 2 & 3). We derived aspect ratio as length:width and circularity as 4π x area / perimeter2 based on their usefulness for capturing variation in I. batatas leaf shape (Gupta et al. 2020; Rosero et al. 2019).

We note that Gupta et al. (2020) and Jackson et al. (2020) controlled for both the possibility of heteroblasty (ontogenetic effects) and phenotypic plasticity (environmental effects) on leaf morphology by sampling foliar material at comparable stages of growth under common garden conditions. Because we could not control for either age or environment with the voucher specimens, we prioritized aspect ratio and circularity as they are less sensitive to environmental influences. We make the necessary assumption that the voucher specimens were sampled from mature, healthy, and representative plants grown under the typical high light conditions ideal for sweet potato cultivation. We, in turn, analyzed undamaged, mature, and representative leaves from each voucher specimen.

To better understand the connection between varietal names and phenotypes, we focused on a subset of traditional Hawaiian sweet potato names. Varieties or variety clusters associated with the name Mohihi, Huamoa, and Piko are relatively well described in written records from the past century (Kagawa-Viviani 2016) and were represented by multiple herbarium voucher specimens. We examined leaf morphology and sampling locales to better understand associations between names, phenotypes, and geographic distribution of these once popular varieties.

Results

Chloroplast Loci-Based Analyses

The results of our chloroplast loci (trnL-trnF and ycf4) analysis indicate strong separation of the Hawai‘i (fresh tissue) samples from UHM Joseph Rock Herbarium samples representing I. batatas collected outside of Hawai‘i, regardless of the method used for tree construction (ML or Bayesian). Furthermore, cultivars from Amy Greenwell Ethnobotanical Garden (Kona, Hawai‘i) formed a cluster distinct from cultivars from all other Hawai‘i botanical gardens (Fig. 1). While purportedly traditional Hawaiian varieties Kala, Kalia, and Pala‘ai sourced from fresh tissues and the Amy Greenwell DNA library samples were separated on both trees, the analysis suggests consistency for Papa‘a kowahi, where fresh tissue indicate similarity to the older accession from Amy Greenwell.

Fig. 1
figure 1

Reconstruction of chloroplast trnL-trnF-ycf4 phylogeny using A) Bayesian and B) maximum likelihood methods.

Full size image

Nuclear Loci Analyses

Clear sub-populations emerged at K = 3, 4, 5, and 6 clusters within our samples of I. batatas (Fig. 2). At K = 3, cluster 1 (purple) included varieties from Hawai‘i gardens and markets: the purple-fleshed, white-skinned Rapoza, Iliahi, and Ogasawara, the undescribed Ko‘oko‘au [sic] and Hui 18b, and HAW12952 and HAW12950 from Honolulu Chinatown. This cluster also included samples sourced from USDA-NPGS: yellow-fleshed Acc108 from the Solomon Islands, CUBA 2 from Cuba, and Camote blanco from Guatemala. K = 3/cluster 2 (green) included a combination of varieties sourced from USDA-NPGS’s more global collection and several locally popular varieties such as the yellow-fleshed Nancy Hall and Japanese-type white-fleshed, pink skin varieties (Yama, Satsuma). This cluster also included several varieties assumed to be Hawaiian (Pa‘ū o Hi‘iaka, Huamoa, Kalia, 89p250, Mahina) tending to be pale-fleshed from white to cream to yellow. K = 3/cluster 3, on the other hand, was dominated by varieties maintained in Hawai‘i botanical gardens (Fig. 2, red).

Fig. 2
figure 2

Discriminant analysis of principal components exploring K2–K6 groups, where each column indicates how varieties clustered depending on the number of groups specified. The specific ordering and proximity of two individuals within a cluster is not related to genetic distance. With clustering optimized at K = 3, Hawai‘i/USDA/Herbarium varieties, Hawai‘i botanical garden varieties, and USDA germplasm varieties grouped together. As group number increased (K = 5 and 6), Hawai‘i botanical garden varieties grew increasingly genetically distinct.

Full size image

At K = 4, a fourth grouping emerged from within the K = 3/cluster 2 that consisted almost entirely of USDA-NPGS samples (Fig. 2, blue). Accession records and notes revealed seven of the twelve varieties to be moist, orange-fleshed, red-skinned cultivars developed in the US Southeast (Louisiana, South Carolina, and Georgia) often marketed as “yams.” Other varieties included cream-fleshed varieties (Sumor from South Carolina and Hawaii from Tonga) and two undescribed accessions from Fiji (Anny and Finlayt K = 5, the K = 3/cluster 3 split into two groups, one consisting of suspected local adoptions [Mana pohaku, Okinawa] and several varieties with the same name assigned to multiple known phenotypes including both white-fleshed and purple-fleshed cultivars [Mohihi, Kaneohe] (Fig. 2, yellow). Unfortunately, we were unable to confirm the specific phenotypes associated with these samples.

The remainder of this cluster (column 4, red) included what has been documented in ethnographic and botanical garden accession records or suspected by growers to be traditional, “old Hawaiian” varieties. These include the very deep purple fleshed/purple-skinned Papa‘a kowahi and ‘Ele‘ele, the light yellow-fleshed/purple-skinned Kala and Piko, and the yellow-fleshed/ light-skinned Huamoa, Pala‘ai, Lanikeha, and Manamana lima. The unnamed varieties 89p246 and 89p249 recovered at a Lāna‘i island roadside were also included in this group. We note that this group also included some USDA-NPGS varieties; ones we have descriptions for indicate they are cream to yellow-fleshed. We note that the sixth cluster emerging at K = 6 (orange) are all characterized by varieties with cream to yellow flesh.

We observed in this analysis that multiple samples labeled as Huamoa ended up in different clusters at K = 6, even when drawn from the same botanical garden. Huamoa_1438 was drawn from a DNA library developed in the 1990s, while Huamoa_2003.0254 was sampled at a later date from the same garden, Amy Greenwell Ethnobotanical Garden. Huamoa (Lyon) thus appears to be more closely associated with the older accession, and it is possible that collection mix-up could have occurred at Amy Greenwell between the two sampling times.

Reassuringly, a number of identically-named cultivars from different botanical gardens clustered together. Besides Huamoa (Amy Greenwell, Lyon), this applied to cultivars labeled as Kala (Amy Greenwell, Lyon) and cultivars labeled as Palau (Waimea, Lyon) for each K-means 3, 4, 5, and 6. The exception here is for Lanikeha, where the samples from Lyon and Waimea clustered into different groups at K = 6, suggesting a need to assess collection fidelity through phenotyping, genotyping, or common garden experiment.

Analysis of Hawaiian ‘Uala Voucher Specimens and Names

In our survey of the 128 voucher specimens housed at the Bishop Museum Herbarium, we observed several names occurred with higher frequency across the Hawaiian Islands. These included Mohihi, which was documented as mohihi a mohaluhalu (Molokai), mohihi keokeo and mohihi ulaula (collected at Laupāhoehoe, Hawai‘i, as transfers from Waihe‘e, Maui), mohihi keokeo and mohihi ulaula (Kaupō, Maui), or simply mohihi (Kona, Hawai‘i; Mokulē‘ia and Mākiki, Oahu). Piko was documented as piko hao, piko manamana, and piko nui (Mokulē‘ia, O‘ahu). Huamoa was huamoa ulaula and huamoa keokeo (Kaupō, Maui), or simply huamoa (Ni‘ihau; Mākiki, O‘ahu; Kona, Hawai‘i). We note that these cultivars sharing common epithets piko, huamoa, and mohihi do not appear to be island-specific names based on the voucher specimen records (ESM, Appendix 3).

The morphological data indicate that Hawaiian cultivars also span a wide range of leaf morphologies, and—drawing again on our examples of mohihi, huamoa, and piko—do not necessarily share a single leaf type (Fig. 3). In general, however, huamoa varieties tend to have more heart-shaped leaves with higher values for circularity and lower length:width. Conversely, piko varieties tend to have lower circularity values, indicating increased lobed-ness. Mohihi varieties span a broader range of circularity values. While voucher specimens provide excellent records of leaf morphology, leaf shape alone is, at best, only loosely associated with cultivar naming or geographic distribution.

Fig. 3
figure 3

Hawaiian ‘uala span a wide range of leaf shapes captured by circularity and aspect, and the names Mohihi (highlighted in orange), Piko (gray), and Huamoa (yellow) are not tightly associated with any particular leaf shape. Photo insets: clockwise from top left: palahai [sic], pia, manamaka/ka‘e, piko ha‘o, mohihi, oheohe, manamana/piko nui, and maihui ulaula

Full size image

Discussion

Our analysis provides insights into the current status of sweet potato cultivar diversity present in Hawai‘i and needs for the continued conservation of traditional Hawaiian cultivars. Emergent lessons include: 1) collections of traditional Hawaiian cultivars tend to be more related to one another than to germplasm from other geographic regions; 2) the Amy Greenwell collection as represented by the DNA library stands out as distinct within this Hawaiian pool; and 3) considerable leaf phenotypic variation characterizes the traditional Hawaiian varieties, even for cultivars carrying similar names. In the following sections, we discuss the value of these findings for current efforts to conserve and increase cultivation of traditional, heirloom Hawaiian sweet potato. We also discuss study limitations and opportunities for future work, including a need to compare extant Hawaiian varieties with the historical collections housed at the Bernice P. Bishop Museum Herbarium.

Status of Hawaiian Sweet Potato Cultivar Diversity

Our chloroplast and nuclear loci analyses both indicate that many traditional sweet potato varieties maintained in Hawai‘i botanical garden collections (ex-situ conservation) are distinct from cultivars sourced from elsewhere in the Pacific region and the world. We do not see evidence of strong widespread admixture, or hybridization, of assumed old Hawaiian cultivars and newer introductions. Furthermore, the distinct clustering of several Amy Greenwell cultivars among those currently grown in Hawai‘i in the chloroplast analysis suggests these accessions are exceptional among the botanical gardens. Our nuclear loci analysis confirmed that extant Hawaiian sweet potato genotypes are different from moist, orange-fleshed “yams” of U.S. provenance as well as the locally popular and drier purple-fleshed, white-skinned “Okinawan” types. Some overlap with white-fleshed, pink-skinned Japanese types, however, is apparent. We conclude that distinct sweet potato lineages have been maintained in Hawai‘i even through at least a century of introductions and breeding of sweet potatoes, including with old Hawaiian varieties.

As we increased the number of clusters from K = 5 to K = 6, Hawaiian cultivars separated into two, then three groupings. From these groupings, we infer that more genetic diversity is present among the Hawaiian samples than among herbarium and USDA germplasm samples we evaluated. Hawaii’s environmental heterogeneity and pre-contact geopolitical complexity provided a context where native farmers of old might have rapidly developed and maintained distinct varieties fit for local climate, soils, and social preferences. This highlights the importance of maintaining traditional Hawaiian sweet potato diversity and preserving the current gene pool. Our work serves as a step toward identifying the unique alleles present across Hawaiian sweet potatoes that could help present-day farmers better understand and value the diversity of traditional varieties.

Varietal Confusion Highlights Need to Verify Collections and Invest in Ex-Situ Conservation

Understanding population genetic structure is a common objective of exploring biodiversity. It is also important when considering conservation and plant breeding. Population structuring occurs when there is a shared history of mutation, migration, selection, and/or drift (Bossart and Prowell 1998; Foll and Gaggiotti 2006; Loveless and Hamrick 1984). Specific phenomena that frequently cause structure in crop plants are mating systems (St. Onge et al. 2011), human use (Taitano et al. 2018), distance between breeding populations (Platt et al. 2010), and environmental gradients that lead to local adaptation within populations (Frichot et al. 2013). Overall levels of diversity can be greatly increased with very little sexual recombination (Roselius et al. 2005). Roullier et al. (2013) identified that the bright yellow color of Kumara cultivars was heavily selected but that background was largely heterogeneous. This was due to the mixed clonal/sexual mating system in sweet potato. In our data set, we identified that while this occasionally occurred, more often than not Hawaiian germplasm was more closely related to other Hawaiian material than germplasm sourced from different locations.

The differentiation of the Amy Greenwell Ethnobotanical Garden collection from other Hawai‘i botanical gardens in the chloroplast analysis indicates garden-specific differences in Hawaiian sweet potato ex-situ conservation. We suggest that garden’s location in fertile remnants of the Kona field system (Kelly 1983), explicit focus on Hawaiian ethnobotany/culturally-relevant plants, staff/community commitment to Hawaiian crop plants, and stable leadership enabled continuity of its Hawaiian ‘uala collection through the 1990s construction of the DNA library (Randell and Morden 1999).

Our chloroplast and nuclear loci analyses indicate inconsistencies between identically-named germplasm, highlighting gaps in the conservation of old cultivars. While the similarity between the Papa‘a kowahi sampled from Amy Greenwell in the 1990s (DNA library) and Papa‘a kowahi in the current collection indicate successful conservation of this genotype (Fig. 1), the distant placement of Kala, Pala‘ai, and Kalia sourced from the DNA library vs. fresh tissues suggest genetic differentiation over time between the 1990s and 2014. Follow-up work is required to identify the reason: sample or accession mix-up, genetic drift, or hybridization.

Similarly, the nuclear analyses of fresh plant tissues confirmed both redundancy and mix-ups across extant sweet potato collections. In some cases, identically-named samples from different gardens grouped together into the same clusters (redundancies). Mix-ups are implied when such samples fell into different groups, indicating a need to revisit and verify the accessions. Thus, molecular tools are useful for identifying coverage and gaps in ‘uala cultivar conservation. We note that collection managers and grounds staff across Hawai‘i’s conservation-oriented botanical gardens have done tremendous work to collect and preserve Hawaiian crop plants, exchanging planting material and accession records with each other. Yet without an integrated approach for tracking these exchanges and verifying cultivars, important information on a fuller set of Hawaiian ‘uala diversity remains fragmented across garden accession records and collection manager memories (T. Sherrill, D. Orr, P. Van Dyke, pers. comm.). Our concern is that cultivars can thus be easily lost without anyone even noticing.

Phenotypes and Genotypes: Opportunities and Challenges of Herbarium Collections

The analysis of Bishop Museum Herbarium voucher specimens indicates that varieties with similar names and collected from the same island are often phenotypically—and we suspect genotypically—distinct. Our comparison of leaf shape and accession notes lead us to conclude that leaf shape does not serve as the central basis for Hawaiian varietal naming and/or some shuffling of names occurred. That said, we acknowledge that sweet potato leaf traits exhibit strong environmental plasticity (Gupta et al. 2020) and mutate even on the same stem (Yen 1963). While we will never be able to fully characterize their phenotypes, these herbarium samples, as the earliest historical collection of Hawaiian sweet potato cultivars, are a crucial resource for assessing their diversity. We were unable to access the Bishop Museum’s voucher specimens for DNA analysis nor match samples common to our study and Roullier et al. (2013), but future work should seek to compare historic and current-day collections in terms of both phenotypes and genotypes.

Toward Connecting Historical and Living Collections

Roullier et al. (2013) point out that a multi-pronged approach is required to interpret historic plant movements in light of modern plant movements and recombination. Similarly, a multi-pronged approach is also required to conserve indigenous crop diversity. We suggest a framework for coordinating efforts to comprehensively research, phenotype, and genotype Hawaiian ‘uala through time by working with oral traditions to current day farms (Fig. 4) which would help increase understanding, cultivation, and enjoyment of this indigenous crop.

Fig. 4
figure 4

Framework in which ethnographic and archival research, herbarium specimens, conservation-oriented (ethno)botanic gardens and collections, and farmers interested in heirloom cultivars might work more synergistically to collectively reinvigorate Hawaiian ‘uala in current-day economic landscapes. Phenotyping and genotyping Hawaiian cultivars will support in-situ, on-farm cultivation, while the added value from these efforts should contribute to the often financially- strapped operations and staff working in gardens, museums, and archives that maintain the names and histories creating that value.

Full size image

Implications for In-Situ Conservation, Breeding, and Selection of Traditional Crops

By confirming the historic status of cultivars, we suggest farmers and cultural practitioners wanting to perpetuate “heirloom” varieties can add value to their operations. This would incentivize the conservation and active propagation of these museum cultivars for growers and crop biodiversity conservation efforts alike, not only in gardens (ex- situ) but on farm (in-situ). Increasing the profit margin for crops may mitigate risk for farmers growing on smaller land areas. The diversity of sweet potato cultivars currently being maintained in gardens can make their way to tables and increase public awareness of crop biodiversity. Genetic assignment analysis is just one component of an effort to understand the diversity and uniqueness of culturally important crops. Data presented here provide a better understanding of the diversity of distinctly Hawaiian sweet potatoes which could lead to potential for niche production on local farms.

When working with indigenous crops, however, care must be taken to engage indigenous communities to ensure open and equitable communication and address concerns regarding bio-prospecting or misappropriation of culturally important germplasm. In Hawai‘i, previous attempts to patent taro (Colocasia esculenta) hybrids and explore genetic engineering for disease resistance resulted in public protest, due in part to the fact that the plant is considered an ancestor of the Hawaiian people (Farran 2014; Gugganig 2017). Although sweet potato does not generally hold as important a position as taro in Hawaiian origin stories, it is considered a body form (kinolau) of the god Lono, is used in medicines and ceremony, and is considered a more important staple than taro in some regions (Handy et al. 1991; Kagawa-Viviani et al. 2018). Scientists should engage with community members to vet their proposed research, ideally a priori, to ensure that communities are not just aware of, but are involved in, consent to and, to the extent possible, drive research that involves cultural resources (Chung-Do et al. 2019; Keaulana et al. 2019).

Conclusion

The sweet potato varieties present in Hawai‘i represent a diverse array of cultivars. Yet many of the named Hawaiian varieties of sweet potato remain genetically distinct from sweet potato varieties grown in the United States and elsewhere around the world. Our analysis highlighted a need to address naming inconsistencies or record-keeping errors among varieties with Hawaiian names, as identically-named varieties appeared in multiple genetic groupings. Our chloroplast analysis showed that cultivars maintained at the Amy Greenwell Ethnobotanical Garden are genetically distinct, even within the broader pool of Hawaiian sweet potato varieties. These unique genotypes should be maintained and leveraged as certified “heirloom” which may provide added appeal in the markets and incentive to farmers to grow them. At the same time, more effort is needed to elucidate the basis for differences between genetic clusters. Lastly, leaf morphology of Hawaiian sweet potatoes varies considerably even among varieties carrying the same name, highlighting a gap in our knowledge of naming conventions. Our work demonstrates how integrating genetic, morphological, and ethnographic data can strengthen our understanding of Hawai‘i and Hawaiian sweet potato diversity and contribute to the preservation, historical significance, and vitality of this important crop.