Volume 226, Issue 5 p. 1240-1255
Tansley review
Free Access

Origin and domestication of Cucurbitaceae crops: insights from phylogenies, genomics and archaeology

Guillaume Chomicki

Corresponding Author

Guillaume Chomicki

Department of Plant Sciences, University of Oxford, South Parks Road, Oxford, OX1 3RB UK

The Queen's College, University of Oxford, High St, Oxford, OX1 4AW UK

Authors for correspondence:

Guillaume Chomicki

Email: [email protected]

Hanno Schaefer

Tel: +49 8161 71 5884

Email: [email protected]

Search for more papers by this author
Hanno Schaefer

Corresponding Author

Hanno Schaefer

Plant Biodiversity Research, Technical University of Munich, Emil-Ramann Str. 2, Freising, 85354 Germany

Authors for correspondence:

Guillaume Chomicki

Email: [email protected]

Hanno Schaefer

Tel: +49 8161 71 5884

Email: [email protected]

Search for more papers by this author
Susanne S. Renner

Susanne S. Renner

Systematic Botany and Mycology, University of Munich (LMU), Menzinger Str. 67, Munich, 80638 Germany

Search for more papers by this author
First published: 22 June 2019
Citations: 113


Some of the World's most valuable crops, including watermelon, honey melon, cucumber, squash, zucchini and pumpkin, belong to the family Cucurbitaceae. We review insights on their domestication from new phylogenies, archaeology and genomic studies. Ancestral state estimation on the most complete Cucurbitaceae phylogeny to date suggests that an annual life cycle may have contributed to domestication. Domestication started c. 11 000 years ago in the New World and Asia, and apparently more recently in Africa. Some cucurbit crops were domesticated only once, others multiple times (e.g. melon from different Asian and African populations). Most wild cucurbit fruits are bitter and nonpalatable to humans, and nonbitterness of the pulp apparently was a trait favoured early during domestication, with genomic data showing how bitterness loss was achieved convergently. The genetic pathways underlying lycopene accumulation, red or orange pulp colour, and fruit size and shape are only just beginning to be understood. The study of cucurbit domestication in recent years has benefitted from the increasing integration of archaeological and genomic data with insights from herbarium collections, the most efficient way to understand species’ natural geographic ranges and climate adaptations.

  Summary 1240
I. Introduction 1240
II. The diversity of cultivated Cucurbitaceae: traits of major and minor crops 1241
III. Time and place of domestication 1245
IV. The genomics of Cucurbitaceae domestication 1250
V. Conclusions 1252
  Acknowledgements 1252
  References 1252

I. Introduction

The gradual transition from hunting and gathering to plant cultivation and animal husbandry began between the end of the Pleistocene and the beginning of the Holocene, some 12 000–10 000 years ago (Fuller et al., 2014; Arranz-Otaegui et al., 2018). The resulting sustainable nutrition of large sedentary populations represents one of the most significant transitions in the c. 300 000 year-long history of Homo sapiens (Larson et al., 2014; Richter et al., 2017). The reasons behind the implied behavioural changes in various human populations remain much debated, but are likely to have involved a combination of climate change, accompanying changes in the vegetation, prey animal densities, human demography, and human social systems (Belfer-Cohen & Goring-Morris, 2011; Larson et al., 2014). Eleven geographic regions have been identified as centres of plant and animal domestication in the New World, Africa, the Middle and Near East, Asia and New Guinea (Piperno, 2011; Fuller et al., 2014; Larson et al., 2014, and references therein).

Domestication involves humans acting as dispersers and modifiers of a crop's biotic and abiotic environment (Larson et al., 2014). It is a gradual process and often is not restricted to a single place or human population (reviewed in Meyer et al., 2012). As a result of this process, crop plants and domesticated animals share suites of modified traits referred to as the ‘domestication syndrome’, differentiating them from their wild ancestors (Darwin, 1868; De Candolle, 1884; Hammer, 1984; Pickersgill, 2007; Meyer et al., 2012; Stetter et al., 2017). In plants, the domestication syndrome often involves larger and more sugary fruits, reduction of physical and chemical defences in the parts used by humans, a change towards a more compact architecture, larger seeds, reduction in seed dormancy, larger inflorescences and non-shattering seeds (in grasses and legumes).

This review synthesises recent insights with older studies to produce a big picture of the domestication of Cucurbitaceae crops, highlighting different domestication trajectories as informed by phylogenetics, archaeology and genomics. The gourd family (Cucurbitaceae) has c. 1000 species, including numerous crops, such as cucumber (Cucumis sativus), bitter gourd (Momordica charantia), watermelon (Citrullus lanatus), preserving melon (Citrullus amarus), honey melon (Cucumis melo), squash and zucchini (Cucurbita pepo), and bottle gourd (Lagenaria siceraria) (Schaefer & Renner, 2011a,b; Renner & Schaefer, 2016; Fig. 1). Archaeobotanists have long documented the roles of cucurbits in ancient cultures based on remains of fruits, seeds and even leaves (e.g. Schweinfurth, 1883, 1884; Piperno, 2000, 2011; Wasylikowa & van der Veen, 2004), and old texts and illustrations document the spread of cucurbit crops from India or the Americas to the Mediterranean and northern Europe (Paris et al., 2012a,b,c, 2013; Paris, 2015). However, it is DNA sequencing of old and new material (usually leaf tissues), informed by meta-data from herbarium specimens, that has yielded the most surprising insights. Such work has by now identified closest relatives, ancestral areas and divergence times of pumpkin, zucchini, squashes, bottle gourd, bitter gourd, chayote, honey melon, watermelon, cucumber, bryonies, and sponge gourds (Sanjur et al., 2002; Erickson et al., 2005; Clarke et al., 2006; Renner et al., 2007; Volz & Renner, 2009; Schaefer & Renner, 2010b; Sebastian et al., 2010, 2012; Telford et al., 2011a,b; Filipowicz et al., 2014; Chomicki & Renner, 2015; Endl et al., 2018).

Details are in the caption following the image
Cucurbitaceae crops and their wild progenitor to illustrate the domestication syndrome. (a, b) Honey melon (Cucumis melo). (a) Wild melon progenitor, Asian agrestis (Cucumis melo subsp. melo f. agrestis). (b) Domesticated melon (Asian lineage). (c, d) Watermelon (Citrullus lanatus). (c) Wild watermelon progenitor, Kordofan melon (Citrullus lanatus subsp. cordophanus). (d) Domesticated watermelon. (e, f) Cucumber (Cucumis sativus). (e) Wild cucumber progenitor (C. sativus f. hardwickii). (f) Domesticated cucumber. Photographs credited to: (a) Balkar Singh; (c) Harry Paris; (b, d, f) Creative commons; (e) Hanno Schaefer.

II. The diversity of cultivated Cucurbitaceae: traits of major and minor crops

Out of the c. 1000 species of Cucurbitaceae, 10 are of worldwide economic importance, cultivated globally, and are here considered ‘major crops’ (Table 1); another 23 are of more local commercial importance, are often cultivated in their native range, and might be called ‘minor crops’ (Table 2). The distribution of these 33 cultivated species over a phylogeny that samples 554 of the family's species and 95 of its 96 genera (Fig. 2) shows that the species used by humans are fairly clustered. All major crops are cultivated for their fruit, a particular type of berry called ‘pepo’ that is consumed ripe (e.g. pumpkin) or unripe (e.g. zucchini), raw (watermelon), cooked (squash), or pickled (gherkins). Cucurbita pepo is the only major crop also cultivated for its seeds and the oil pressed from them. Minor crops are cultivated for a wide range of purposes including raw fruits (e.g. Cucumis metuliferus), nutritious seeds (e.g. Citrullus mucosospermus) or to make oil (e.g. Telfairia occidentalis, T. pedata, Hodgsonia macrocarpa), cooked fruit (e.g. Trichosanthes cucumerina), sugary fruit used as a sweetener (Siraitia grosvenorii) or for their physical (Luffa aegyptiaca used as sponge) or medicinal properties (Citrullus colocynthis; Barghamdi et al., 2016).

Table 1. Traits and geographic origins of the major domesticated Cucurbitaceae crops.
Species Common names Area of origin Area of domestication Type of evidence for area of domestication Timeline of domestication (years bp, minimum age) Wild progenitor found? Sister species References
New World
Cucurbita argyrosperma Silverseeded gourd Mexico Mexico W, P - Yes (C. argyrosperma subsp. sororia) C. moschata Sanjur et al. (2002); Kates et al. (2017)
Cucurbita maxima Pumpkins, Squashs South America Argentina W, P 4000 Yes (C. maxima subsp. andreana) C. ecuadorensis Sanjur et al. (2002); Kates et al. (2017)
Cucurbita moschata Butternut squash Unknown Unknown - - No C. argyrosperma Sanjur et al. (2002); Kates et al. (2017)
Cucurbita pepo subsp. pepo Summer squashs, zucchini, pumpkins Mexico Mexico A, G, P, W 8000–10 000 No C. okeechobeensis + C. lundelliana Smith (1997); Sanjur et al. (2002); Kates et al. (2017)
Cucurbita pepo subsp. texana Acorn, crookneck and pattypan squashes Texas and nearby states Texas G, W - Yes C. okeechobeensis + C. lundelliana Sanjur et al. (2002); Kates et al. (2017)
Citrullus lanatus


*Red and black seed melon (China)

Sudan (?)

Nile Valley

*China (fleshy seed cultivar)

A, C, P, W?




Yes, potentially C. mucosospermus Mote (1977); Wasylikowa & van der Veen (2004); Chomicki & Renner (2015); Renner et al. (2017, 2019)
Asia and Melanesia
Benincasa hispida Wax gourd Southeast Asia, Australasia Southeast Asia, Indonesia? A 2450 or before Yes B. fistulosa Matthews (2003), Marr et al. (2007), Kocyan et al. (2007), Schaefer et al. (2009)
Cucumis melo subsp. melo Numerous names for cultivars of Melon India and Australia Asia A, G, P 2200 Yes C. trigonus + C. picrocarpus Walters (1989); Sebastian et al. (2010); Endl et al. (2018)
Cucumis sativus *Cucumber (Eurasian), **pickling cucumber (East Asian), ***Xishuangbanna cucumber India *Asia, **East Asia, ***Xishuangbanna region A, C, P, W ***2500 Yes (C. sativus var. hardwickii) C. hystrix Renner et al. (2007); Sebastian et al. (2010); Qi et al. (2013)
Lagenaria siceraria Bottle gourd Africa Independently domesticated in Eurasia* and South America** A, P

*11 000

**10 000

Yes L. breviflora Kistler et al. (2014)
Momordica charantia Bitter gourd Africa Africa or India, unclear C, G ? Yes M. angolensis Schaefer & Renner (2010a)
  • For the type of evidence for the domestication area; A, archaeological remains; C, other cultural evidence (e.g. linguistic or iconographic); G, genetic diversity of primitive landraces; P, phylogenetic evidence; W, inferred from wild progenitor distribution. ; (?) uncertainty remaining; ? unknown; *, ** and *** are used when crop species have been independently domesticated.
Table 2. Minor Cucurbitaceae crops.
Species Tribe Common names Area of origin Area of cultivation Evidence for domestication Habit Sexual system Timeline of cultivation (years bp, minimum age) Use References
Benincasa fistulosa Benincaseae Tinda NW India (?) India, Pakistan, E Africa Yes Annual Monoecious   Fruit Schaefer et al. (2009); Schaefer & Renner (2011b)
Citrullus amarus Benincaseae Preserving melon South Africa, Namib desert Mediterranean Documented as crop in the Mediterranean region for > 1900 years Perennial Monoecious >1900 Fruit rind boiled with sugar for jams Bailey (1930); Chomicki & Renner (2015)
Citrullus colocynthis Benincaseae Colocynth Northern Africa Northern Africa, Middle East to India No Perennial Monoecious 4000 Various medicinal uses from Ancient Egypt on Renner et al. (2017)
Citrullus mucosospermus Benincaseae Egusi melon West Africa West Africa No Annual Monoecious   Oil and protein rich seeds used for cooking Achigan-Dako et al. (2015); Chomicki & Renner (2015); Renner et al. (2017)
Coccinia grandis Benincaseae Scarlet gourd, Kowai East Africa India and Southeast Asia No Perennial Dioecious   Young shoots cooked and fruits Holstein (2015)
Cucumis anguria Benincaseae Cackrey, Maroon cucumber, West Indian gherkin Africa, naturalised in the New World Tropical regions Yes, nonbitter and smooth fruit Perennial Monoecious   Fruits pickled Kirkbride (1993)
Cucumis melo subsp. meloides Benincaseae Tibish and Fadasi Melon Africa, perhaps Nile valley Sudan Yes, nonbitter and larger fruit Probably annual Monoecious   Young fruits eaten cooked, seeds used in soups Endl et al. (2018)
Cucumis metuliferus Benincaseae Kiwano Sub-Saharan Africa Worldwide Yes, larger fruit and sweeter cultivars Annual Monoecious   Fruit eaten raw Kirkbride (1993)
Cyclanthera pedata Sicyoeae Stuffing cucumber, caigua, achocha Neotropics South America Yes, larger and smooth fruits Perennial Monoecious   Fruit cooked Jeffrey (1980)
Hodgsonia macrocarpa Sicyoeae Lard fruit Himalaya China No Perennial Monoecious   Seeds used to make oil or eaten roasted Chien (1963)
Luffa acutangula Sicyoeae Sponge gourd, angled luffa Arabian Peninsula, India Tropical regions Yes, larger fruit Annual Monoecious   Unripe fruit cooked, mature dried fruit used as a sponge Filipowicz et al. (2014)
Luffa aegyptiaca Sicyoeae Sponge gourd Southeast Asia Tropical regions Yes, larger fruit Annual Monoecious   Unripe fruit cooked, Mature dried fruit used as a sponge Filipowicz et al. (2014)
Melothria mannii (syn. Cucumeropsis mannii) Benincaseae Egusi gourd Central America West Africa and Neotropics No Perennial Monoecious   Seeds Schaefer & Renner (2010b)
Melothria scabra Benincaseae Mouse melon, Mexican miniature watermelon,pepquinos Mexico and Central America Mexico and Central America No Perennial Monoecious   Fruit eaten raw Schaefer & Renner (2010b)
Momordia dioica Momordiceae Bristly balsa, pear, spiny gourd South Asia South Asia, especially India No Annual Dioecious   Leaves as vegetable fruit boiled Schaefer & Renner (2010a)
Momordica balsamina Momordiceae Balsam apple Tropical Africa Mediterranean? No Annual Monoecious     Schaefer & Renner (2010a)
Momordica cochinchinensis Momordiceae Gac Southeast Asia to North Australia Southeast Asia Yes, larger fruit Perennial Dioecious   Fruit eaten raw Schaefer & Renner (2010a)
Sicana odorifera Cucurbiteae Cassabanana South America Latin America and Southern United States No Perennial Monoecious   Fruit eaten raw Schaefer & Renner (2011b)
Sicyos edulis (Sechium edule) Sicyoeae Chayote, christophine, chouchou Mexico Tropical regions Yes, larger fruit, loss of bitterness and spines on fruits, increased sugar content Perennial Sicyos (Sechium) chinantlense >500 Fruit eaten as vegetable Newstrom (1991); Sebastian et al. (2012)
Siraitia grosvenorii Siraitieae Luo Han Guo Southern China and Northern Thailand China No Perennial Dioecious   Fruit used as a low-calorie sweetener and in traditional Chinese medicine Schaefer & Renner (2011b)
Telfairia occidentalis Joliffieae  Fluted gourd, Fluted pumpkin Tropical West Africa Tropical West Africa Yes, larger fruit Perennial Dioecious   Seeds used to make oil or roasted Schaefer & Renner (2011b)
Telfairia pedata Joliffieae  Oysternut, Zanzibar oil vine Tropical East Africa Tropical East Africa No Perennial Dioecious   Seeds used to make oil or roasted Schaefer & Renner (2011b)
Trichosanthes cucumerina Sicyoeae  Snake gourd, Serpent gourd Southeast Asia? China Yes, larger fruit Annual Monoecious 1000–800 Fruit cooked Walters (1989)
  • This include species that are cultivated but not domesticated and species domesticated but only of local importance.
Details are in the caption following the image
Phylogenetic distribution of cultivated Cucurbitaceae and ancestral state estimations on a phylogeny sampling 554 of the family's c. 1000 species of annual vs perennial taxa. Name of major crop species printed in red (and marked by arrows), names of minor crop species in blue.

The domestication syndrome of Cucurbitaceae crops includes nonbitter fruits, increased fruit size, sometimes with higher sugar or carotenoid content, decreased physical defences (e.g. wild chayote fruits are spiny), and more compact and less branched growth with increased apical dominance. Selection of plants with desirable properties was probably begun by individual farmers and did not involve mass selection. All crop species went through population bottlenecks during domestication (Gross & Olsen, 2010). However, the number of rice plants sown by a household would be significantly higher than that of cucumber or squash plants, and the population size of founder plants in cucurbit domestication may therefore have been smaller than in Poaceae crops (Qi et al., 2013). Traits, such as nonbitter fruits, clearly were consciously selected for and were not a byproduct of selection for other features (see section IV.1 'Loss of bitterness').

Most Cucurbitaceae are annual or perennial herbs, and c. 50% of the species are monoecious, 50% dioecious (Kocyan et al., 2007; Schaefer & Renner, 2011b). Most major crop species (Table 1) are annuals with a monoecious sexual system in which every individual sets fruit. To test whether these traits were present already in the wild relatives of these crops or are overrepresented among domesticated forms, we performed a family-wide ancestral state estimation of life form evolution, scoring species as annual or perennial based on personal observations, labels of herbarium specimens and relevant literature. The results reveal that all but one of the major crop species are nested in lineages with an annual habit, while minor crops are either annual or perennial (Fig. 2). An annual life cycle may have facilitated domestication. Cucurbitaceae all have indeterminate growth, meaning that annual domesticated species are actually perennials that die at the end of the productive cycle due to the exhaustion of the vegetative organs imposed by the selection for high productivity or due to unfavourable climate conditions outside their centre of origin. In other words, perennial Cucurbitaceae crops are cultivated as annuals, but this is a management practice, not the result of genetic changes. Perennial wild species often build up large underground tubers from which annual shoots emerge (Schaefer & Renner, 2011b) that flower in their first year (H. Schaefer, pers. obs.).

Most Cucurbitaceae have unisexual flowers, and based on outgroup comparison, dioecy appears to be the ancestral state in the family (Zhang et al., 2006). There have been numerous evolutionary changes between monoecy and dioecy (Volz & Renner, 2008; Schaefer & Renner, 2010b), and the sexual systems of many wild species are neither reliably known nor can be extrapolated from herbarium material, especially in climbers in which male and female flowers appear distant from each other and at different times. Some cucurbit climbers, such as species of Gurania and Psiguria, only bear male flowers until they reach sunny conditions (in the canopy or because of a disturbance) and then form female flowers (Condon & Gilbert, 1988). In many dioecious species, prolonged careful observations of numerous individuals also reveal ‘leaky dioecy’, the occasional development of flowers of both sexes in dioecious individuals (Schaefer & Renner, 2010b).

III. Time and place of domestication

1. The New World as a centre of cucurbit domestication

Several important Cucurbitaceae crop species were domesticated in the New World in pre-Columbian times (Larson et al., 2014). The most prominent are the globally important squashes and pumpkins (Cucurbita spp.; Table 1; Fig. 3). The genus Cucurbita had a wide pre-Holocene distribution and was adapted to the disturbed habitats maintained by large-bodied mammals that also dispersed the bitter fruits, potentially because these mammals had few bitter taste receptor genes (Kistler et al., 2015). Population fragmentation inferred from plastid genomes suggests that the Holocene megafaunal extinction led to a decline in wild Cucurbita, which were then ‘rescued’ by anthropogenic dispersal during domestication (Kistler et al., 2015). Archaeological evidence points to northern South America and Central America as places of the earliest pumpkin domestication, c. 10 000 years ago (Smith, 1997).

Details are in the caption following the image
Phylogeny and domestication of Cucurbita. The maximum likelihood (ML) tree was inferred from 44 nuclear loci taken from Kates et al. (2017). Blue-coloured taxa indicate likely progenitor taxa, red-coloured taxa indicate domesticated species. ML bootstrap support values are shown for the backbone and the main clades (all values are shown in Kates et al., 2017).

Of the four major crop species in the genus Cucurbita, C. pepo is native to North America (Northeast Mexico, and the Southeast and Central United States; Bailey, 1943; Paris et al., 2012c), and archaeological remains suggest an initial domestication some 8000–10 000 yr ago in Mexico (Smith, 1997). Independent domestication occurred in Eastern North America (Sanjur et al., 2002). These two domestication events led to the forms now classified as C. pepo subsp. pepo and C. pepo subsp. ovifera (Paris et al., 2012c; Table 1; Fig. 3). No wild populations of subsp. pepo have been found, but wild populations of subsp. ovifera occur in Texas and nearby states, and a 3rd taxon (C. pepo subsp. fraterna) occurs in Northeast Mexico (Nee, 1990). However, C. pepo subsp. ovifera is not monophyletic, instead appearing in at least two places in a nuclear phylogeny, with a mix of domesticated and nondomesticated forms, suggesting either introgression, multiple domestication centres and/or broad domestication bottlenecks (incomplete lineage sorting in the crop) (Kates et al., 2017; our Fig. 3). American cultivars of C. pepo subsp. pepo reached Italy in the 16th century, where in the mid-19th century they gave rise to the cultivar group known as zucchini. From Italy, zucchini returned to the Americas, and today, they are one of the most widely cultivated crops of the family (Lust & Paris, 2016).

The second domesticated species of Cucurbita is C. maxima, with the subspecies maxima (Hubbard squash, buttercup squash, giant pumpkin) and andreana. The former was cultivated as early as 4000 bp along the coast of Peru where it may have been domesticated, but apparently never left the Neotropics during pre-Columbian times (Sauer, 1993; Piperno & Stothert, 2003). The other subspecies, andreana, appears to have been domesticated in what is today Argentina (Nee, 1990; Sanjur et al., 2002; Kates et al., 2017).

The third domesticated species of Cucurbita, C. moschata – the butternut squash – may have originated in Northern South America (Sanjur et al., 2002), consistent with high landrace diversity, and commonness of landraces with primitive traits such as small dark seeds, lignified, warty rind and bitter pulp (Nee, 1990). However, nuclear sequence data cast doubt on a South American origin of this species (Kates et al., 2017).

The fourth domesticated species of Cucurbita, C. argyrosperma is native to Mexico (Sanjur et al., 2002; Kates et al., 2017) and probably was domesticated there, although the timeline of its domestication remains unclear because of a lack of archaeological remains.

Another Cucurbitaceae crop from a different genus domesticated in the New World is Sicyos edulis, the chayote. Linguistic data, together with cultivar genetic diversity, pinpoint Mexico as its centre of domestication (Newstrom, 1991), consistent with a biogeographic analysis showing Mexico as the ancestral area of the clade containing chayote (Sebastian et al., 2012). Hernandez (1550) documented the use of the chayote by the Aztecs. It became a worldwide crop in tropical, subtropical and warm temperate regions in the 19th century when it was introduced in Africa, Southern Europe, Asia and the West Indies, with the precise dates of introduction known in most cases (Newstrom, 1991). A potential wild progenitor population of chayote has been found in Mexico, with small, spiny and bitter fruits, but some less bitter and containing sugar (Newstrom, 1991).

Melothria mannii is a minor crop (Table 2) cultivated for its nutritious seeds as a so-called ‘egusi-crop’ (for making stews) in both West Africa and Central and South America. Biogeographic reconstruction indicates that it is native to America (Schaefer & Renner, 2010a; treating the species under the erroneous name Melothria sphaerocarpa). Whether it has been domesticated independently in America and Africa is unclear. It could have been brought to Africa as a domesticate during the slave trade, but natural transatlantic dispersal of either wild or cultivated forms cannot be excluded (Schaefer & Renner, 2010a).

2. Africa as a centre of cucurbit domestication

The genus Citrullus is the most economically important Cucurbitaceae lineage in Africa where it has seven species (Chomicki & Renner, 2015) of which several are cultivated and some are used as bush food (C. amarus, C. naudinianus). The taxonomy of the genus Citrullus until recently contained misapplications of names, including that of the watermelon, C. lanatus, itself (Renner et al., 2014; Chomicki & Renner, 2015). The erroneous nomenclature led to a number of confusions, including the reference to ‘wild’, ‘semi-wild’ and ‘domesticated’ populations of watermelon that are in fact distinct species ‘separated’ by undoubted wild species. Citrullus lanatus is the most economically important species in the genus. It was long thought to originate from South Africa because of a taxonomic mistake involving the oldest collection of a wild South African species described by a student of Linnaeus, which in the 1930s was wrongly synonymised with the cultivated watermelon (Bailey, 1930). Phylogenetic evidence, including DNA from the original specimen collected near Cape Town in 1773 by Linnaeus's student Carl Peter Thunberg, showed that no South African material is closely related to the domesticated watermelon. The earliest archaeological evidence of cultivated watermelon consists of seeds dated to c. 5000 bp (8000? bp) from Wan Muhuggiag in Libya (Wasylikowa & van der Veen, 2004). An oblong, large fruit served on a tray and with the characteristic striped fruit skin of watermelon is shown on a wall painting in a Pharaonic tomb from Meir, Northwest of Asyut (reproduced in Paris, 2015), suggesting that the fruit was eaten raw, which would indicate the availability of sweet, nonbitter watermelons in Egypt 4000 years ago (Chomicki & Renner, 2015).

Four hypotheses have been proposed for the origin of the watermelon: First, that it descends from the northern African colocynth (C. colocynthis; Singh, 1978; Sain et al., 2002; McCreight et al., 2013). Second, that it derives from the South African citron melon, C. amarus (Robinson & Decker-Walters, 1997; Maynard & Maynard, 2000; Rubatsky, 2001). Third, that it stems from the West African C. mucosospermus (Guo et al., 2013; Chomicki & Renner, 2015), and fourth, that it both originated from and was domesticated in Northeastern Africa (Paris, 2015). A phylogenetic analysis of Citrullus (Chomicki & Renner, 2015), together with genetic, archaeological and historical data (reviewed by Paris, 2015) rejects the first two hypotheses, leaving West Africa and Northeast Africa. Phylogenomic analyses of nuclear gene sequences suggested that the white-fleshed Sudanese Kordophan melon is the closest relative of the domesticated watermelon (Renner et al., 2019; Fig. 4). The lack of bitterness and high sugar content of Kordophan melons (Ter-Avanesyn, 1966), however, leaves open the possibility that these melons represent a landrace. Phylogenomic data also imply that the West African egusi melon, C. mucosospermus, bred for its nutritious seeds (Achigan-Dako et al., 2015), was domesticated from a different gene pool than that of the watermelon.

Details are in the caption following the image
The phylogeny and domestication of Citrullus. The phylogeny results from a maximum likelihood (ML) analysis of 143 nuclear loci from Renner et al. (2019) with ML bootstrap values shown at branches. ML analysis of a matrix with 100 plastid loci yielded the same topology, with high bootstrap support at all nodes. Blue-coloured taxa indicate likely progenitor, red-coloured taxa indicate domesticated species.

Another crop in the genus Citrullus is the preserving melon, C. amarus, which originates from Southern Africa (Chomicki & Renner, 2015; Renner et al., 2017 provide a distribution map). This species, referred to as ‘citron melon’ in its wild South African form, is an important crop in the Mediterranean region where the ripe fruit, including its rind, is boiled with sugar to make jams (Laghetti & Hammer, 2007). It is also used in specialty baking, for example, in fruit cake mixes and Christmas baking in Mexico and the United States (Bush, 1978). The species was introduced into the Mediterranean region from at least the Roman era, as evidenced by mention in a Latin cookbook (De Re Coquinaria, ad 77–516) as citrium (Paris, 2015), and there is archaeological and iconographic evidence that it was a widely used crop all around the Mediterranean and Europe from at least the 14th century (Paris et al., 2013; Paris, 2015). The species was also introduced to Australia by Afghan cameleers in the mid 1800s to early 1900s, who used its fruits as a feed source for their camels; it is now a weed of summer fallows in Australia (Shaik et al., 2017).

The colocynth, C. colocynthis, has also been cultivated since Ancient Egyptian times as a medicinal plant and a source of oil but has not been domesticated. Colocynth seeds are found from 5000 bp (8000 bp) onwards in archaeological sites in Libya, Egypt and the Near East (Wasylikowa & van der Veen, 2004), and the species may have reached India and Pakistan by human transport.

3. Asia and Melanesia as centres of cucurbit domestication

The most important Cucurbitaceae crop from Asia is the cucumber, Cucumis sativus. Cucumbers have been domesticated from wild Indian C. sativus var. hardwickii (Qi et al., 2013), which has small bitter-tasting or sour fruits that are ellipsoid or subglobose and 5–8 cm long (Fig. 1e); it is known from India and Northwest and West Thailand, growing wild and sold in local markets. Three main cultivar groups of cucumber were subsequently selected for distinct purposes, namely Eurasian cucumbers (our slicing cucumbers eaten raw and immature), East Asian cucumbers (pickling cucumbers) and Xishuangbanna cucumbers (Fig. 5, inset). Based on demographic modelling, the East Asian C. sativus cultivars diverged from the Indian cultivars c. 2500 years ago, that is close to a Chinese text reporting the introduction of cucumber to China during the Han dynasty (Qi et al., 2013). The Xishuangbanna cultivar (C. sativus var.  xishuangbannanensis) was selected for high β-carotene content in the mature pulp; its fruits are eaten boiled or raw during different stages of maturity (Yang et al., 1991; Renner, 2017).

Details are in the caption following the image
The phylogeny and domestication of Cucumis. The maximum likelihood (ML) tree was inferred from six plastid and one nuclear locus from Endl et al. (2018), with ML support above branches and Bayesian poster probabilities below branches. Blue-coloured taxa indicate likely progenitor taxa, red-coloured taxa indicate domesticated taxa. The inset shows the three domestication directions of the cucumber, C. sativus, based on Qi et al. (2013). Photograph credited to: bottom right inset: Michel Pitrat.

Honey melon (Cucumis melo) was domesticated at least once in Asia and once in Africa (Endl et al., 2018; Fig. 5). The oldest melon remains from Asia date to 3000 bc in China (Watson, 1969) and to 2300–1600 bc in the Indus valley (Vishnu-Mittre, 1974). The oldest African melon seeds date to a site from 3700 to 3500 bc in Lower Egypt (van Zeist & de Roller, 1993; El Hadidi et al., 1996). The honey melon cultivars grown today can be traced back to two wild lineages that are estimated to have diverged 1–3 Ma. The first one is restricted to Asia (C. melo subsp. melo), while the second, C. melo subsp. meloides Endl & H. Schaef., is widespread throughout Africa. The Asian lineage gave rise to all modern cultivars including the commercially important ‘Galia’, ‘Cantaloupe’, and ‘Yellow Honeydew’ melons, while the African lineage gave rise to the ‘Tibish’ and ‘Fadasi’ melons, landraces still grown in the Sudanese region but gradually being replaced by imported cultivars from Asia. A third wild C. melo lineage is restricted to Australia and New Guinea but apparently was never domesticated (Endl et al., 2018). The C. melo clade is sister to a long overlooked Indian perennial, C. trigonus and the Australian C. picrocarpus (Endl et al., 2018).

Lagenaria siceraria, the bottle gourd, the dry empty fruits of which are used as watertight containers, is native to Africa (Table 1). The species appears to have been cultivated as early as 11 000 years ago in Asia and independently in the Americas, where it was widespread 8000 bp (Erickson et al., 2005; Kistler et al., 2014). For Polynesia, genetic data suggest arrivals from both Asia and South America, after the species had become a well established crop in those regions (Clarke et al., 2006). An older scenario suggested that the bottle gourd was brought to the New World by Paleoindians as they colonised the continent (Erickson et al., 2005), but this would imply that a tropical plant could have been planted and harvested under Arctic climate conditions. A more recent phylogenetic analysis of a denser sample of archaeological wild and cultivated American and Asian bottle gourds indicated that the species was brought from Africa to Eurasia by humans, but reached America by natural transoceanic dispersal from Africa before the arrival of humans there (Kistler et al., 2014). Strangely, archaeological evidence for early bottle gourd cultivation is restricted to the Pacific coast of South America, which does not match the species’ presumed transoceanic arrival from Africa. Bottle gourd used to be an important vegetable in Europe, later replaced by zucchini (Lust & Paris, 2016).

The bitter gourd, Momordica charantia, another major cucurbit crop (Table 1), is native to Africa (Schaefer & Renner, 2010b) and Madagascar, but today is most extensively used in Asian cuisines. Its use in Kerala (Southern India) is documented from the 17th century onwards (Rheede, 1688; Marr et al., 2004). It remains unclear where this important crop was domesticated, and more sampling of landraces and wild forms from Madagascar, mainland Africa and Southern India is necessary to confirm its region of domestication.

The wax gourd, Benincasa hispida, is documented from an archaeological site in Thailand dated from 9980 to 9530 bp (Pyramarn, 1989), and in New Guinea, rind and seeds have been found at the Kuk swamp archaeological site (Golson et al., 2017) dated to 2450 bp (Matthews, 2003). Seed size does not resolve the status of these archaeological finds as cultivated or wild because at least the size of the New Guinean seeds is within the range of domesticates (Marr et al., 2007). The related minor crop Tinda or Indian round gourd, Benincasa fistulosa (syn. Praecitrullus fistulosus), is cultivated in India and Pakistan and at a very small scale in East Africa, where it has been introduced. Its origin and area of domestication are unclear but wild forms exist in Northwest India (Renner & Pandey, 2013).

The area of domestication of the sponge gourd, Luffa aegyptiaca, which occurs throughout Southeast Asia, is unknown and the entire genus with eight species worldwide is surprisingly poorly understood (Marr et al., 2005; Filipowicz et al., 2014). Small-fruited wild forms occur in Australia and Indonesia (Marr et al., 2005). Wild forms of the angled loofah, L. acutangula, are known from the Arab Peninsula and India (Filipowicz et al., 2014). Interestingly, none of the three Neotropical Luffa species has been domesticated.

IV The genomics of Cucurbitaceae domestication

So far, 11 reference genomes of cucurbits have been produced, namely C. sativus var. sativus cv 9930 and cv Gy14, one wild cucumber (C. sativus var. hardwickii PI 183967), one cultivated melon (C. melo cv DHL92), two cultivated watermelons (C. lanatus subsp. vulgaris cv 97103 and cv Charleston Gray), four cultivated Cucurbita species (C. maxima cv Rimu, C. moschata cv Rifu, C. pepo cv MU-CU-16, and C. argyrosperma), and one cultivated bottle gourd (Lagenaria siceraria cv USVL1VR-Ls) (Huang et al., 2009b; Garcia-Mas et al., 2012; Guo et al., 2013; Qi et al., 2013; Sun et al., 2017; Wu et al., 2017; Wang et al., 2018; Montero-Pau et al., 2018; Ruggieri et al., 2018; Barrera-Redondo et al., 2019), and the number is rapidly increasing. Zheng et al. (2019) summarised the genomic resources available by 2018. Comparisons of nondomesticated wild forms with domesticated accessions imply that domestication of cucurbit fruits always first required loss of fruit bitterness, a trait conferred by toxic cucurbitacins (Zhou et al., 2016). Selection also affected the seed coat, carotenoid content and sugar content. Below we summarise how this may have occurred. We restrict our review of comparative and functional genomics of cucurbit crop traits to traits directly relevant to domestication. Molecular mechanisms underlying flower sex-determination, fruit development, and spine development in cucumber are reviewed by Che & Zhang (2019).

1. Loss of bitterness

Some wild cucurbit species have sweet fruit pulp, and some of these are minor crops (Cucumis anguria, Cucumis metuliferus, Melothria scabra); others are bush food collected from the wild and eaten by kids, such as the fruit pulp of wild Kedrostis foetidissima, Momordica species, Solena heterophylla, and some Trichosanthes (e.g. Murthy et al., 2013). Most wild cucurbits, however, have bitter fruits, with bitterness conferred by a group of terpenoid compounds called cucurbitacins. Cucurbitacins are very effective in direct plant defence against herbivores and are present in leaves, roots and fruits in many nondomesticated cucurbits. With the exception of the bitter gourd (Momordica charantia), for which some cultivars are very bitter when it is harvested, Cucurbitaceae fruit crops have lost bitterness during domestication. Nonbitter remains of the squash Cucurbita moschata that date to 9200 bp indicated that automatic selection (sensu Harlan et al., 1973) for nonbitter fruits occurred very early (Piperno & Dillehay, 2008).

The Bi gene confers bitterness to the whole plant (Huang et al., 2009 and references therein), and a genetic and biochemical study showed that Bi encodes a cucurbitadienol synthase that catalyses the first committed step in cucurbitacin C biosynthesis in cucumber (Shang et al., 2014). Two bHLH transcription factors regulate cucurbitacin C biosynthesis by upregulating Bi in the leaves (Bl) and fruits (Bt) directly via binding in the E-box elements of the Bi promoter (Shang et al., 2014). In cucumber, no selective sweep was associated with the Bi gene (Qi et al., 2013), suggesting that loss of cucurbitacin biosynthesis in the whole plant would have been deleterious, even in cultivated crops (notably in terms of defence against generalist herbivores). Examination of different cucumber lines with varying bitterness has revealed that domestication of nonbitter cucumber occurred via the downregulation of Bt, either by mutation in its cis-acting elements or by affecting the binding site (Qi et al., 2013; Shang et al., 2014; Zhou et al., 2016). A genomic analysis of homologues comparing cucumber, melon and watermelon (producing cucurbitacin C, B and E, respectively) revealed a convergent domestication sweep at the Bt loci and the loss of bitterness is, in all cases, due to convergent mutations at this locus (Shang et al., 2014; Zhou et al., 2016). Selected mutations in an upstream regulatory gene controlling the expression of a given pathway in specific tissue may be common during domestication, as this can help avoid pleiotropic effects associated with modification of the pathway itself (Lenser & Theißen, 2013).

2. Selection for sweetness

For several Cucurbitaceae crop species, especially watermelon (Citrullus lanatus) and honey melon (Cucumis melo), selection for sweet fruits has occurred during domestication. Genome analysis and transcriptomics in the watermelon revealed that α-galactosidase, insoluble acid invertase, neutral invertase, sucrose phosphate synthase, UDP-glucose 4-epimerase, soluble acid invertase and UDP-galactose/glucose pyrophosphorylase are the main enzymes that regulate sugar unloading and metabolism during ripening (Guo et al., 2013). In total, 62 sugar metabolism genes and 76 sugar transporter genes have been annotated, with 13 and 14, respectively, differentially expressed in fruit flesh (Guo et al., 2013). In honey melon, 63 genes involved in sugar metabolism have been annotated (Garcia-Mas et al., 2012). Of the sugar transporters found in domesticated (sweet and red-fleshed) watermelon, but not in citron melons (Citrullus amarus (PI296341-FR)), two SWEET-like sugar transporters (Sugars Will Eventually be Exported Transporter (SWEET)) were highly expressed in developing red-flesh of watermelon (but not rind), and their expression level was correlated to the amount of fruit sugar, suggesting that they increase fruit flesh sweetness by facilitating the active transmembrane transport of sugars (Guo et al., 2015).

Work on watermelon has further identified a tonoplast sugar transporter (ClTST2) that pumps sucrose, glucose and fructose inside the fruit's vacuoles (Ren et al., 2018). Cross-comparisons across hundreds of Citrullus accessions spanning several species revealed that increased expression of ClTST2 was a major event in the domestication of the watermelon. ClTST2 is regulated by a sugar-induced transcription factor called SUSIWM1, and a single SNP in the tonoplast sugar transporter (ClTST2) has been fixed in the sweet, domesticated watermelon, and underlies a sugar accumulation quantitative trait locus (QTL) in watermelon (Ren et al., 2018). Therefore, it appears that increased ClTST2 expression by enhanced binding efficiency of SUSIWM1 to the ClTST2 promoter was a key event increasing sugar content during the domestication of the watermelon.

Altogether these data for honey melon and watermelon suggest that increased sugar content resulted from selection on dozens of genes altering sugar metabolism or transport.

3. Selection for larger fruits

In all 10 major Cucurbitaceae crops (Table 1), humans selected for larger fruits. Fruit size is essentially controlled by three processes in the ovary: cell differentiation (e.g. the definition of the number of carpels), cell division and cell expansion. Carpel number only varies from three to seven, with even very large specimens of Cucurbita maxima only having five to seven carpels (Grumet & Colle, 2017). The increase in fruit size therefore mostly involves shifts in the regulation of cell division and in cell expansion. The transcriptional master regulators of the fruit size gene networks are still unclear, but so far phytohormone-related genes (auxin, cytokinins and gibberelins), microtubule-related genes and cyclin-related genes appear to be the key players controlling fruit size.

In watermelon, Citrullus lanatus, two MADS box genes similar to the tomato ripening and expansion gene TAGL1 have been characterised and are highly expressed during fruit expansion and ripening, suggesting that they regulate both processes (Guo et al., 2013). In cucumber, C. sativus, analysis of QTL identified a single selective sweep with 19 genes, one of which encodes a cyclin involved in cell proliferation (Qi et al., 2013). Another study identified 12 QTLs for fruit size in this species that were linked to different aspects of size, in particular length and width (Weng et al., 2015), and that have been selected in different ways in the three domestication directions of Cucumis sativus (Fig. 5 inset). Transcriptome analysis of cucumber revealed thousands of differentially expressed genes, including many linked to microtubules and cyclins (cell division) (Jiang et al., 2015). A single recessive gene, spontaneous short fruit 1 (sf1), appears to regulate cucumber length through auxin and cytokinin signalling (Wang et al., 2017).

Clearly the regulation of fruit size is a complex process, and domestication has affected it in different ways (for instance, long and thin Eurasian salad cucumbers vs short and thick East Asian pickling cucumbers; Fig. 5), implying changes in the spatio-temporal fine-tuning of genes controlling cell division and expansion. The world's largest fruits are produced by Cucurbita maxima cv Dill's Atlantic Giant, with record fruits weighing > 1000 kg. Despite the huge variation in fruit size and weight in Cucurbitaceae crops and their close relatives, the regulatory control underpinning these changes is not well understood. Functional characterisation of candidate genes regulating cell division and expansion in fruits is an exciting future research path. This approach would pave the way for interesting comparisons of fleshy fruit development across angiosperms of which the mechanisms are currently almost exclusively known in tomato (Karlova et al., 2014).

4. Seed traits

Relatively few Cucurbitaceae crops are cultivated for their seeds. Most important economically are pumpkin seeds (Cucurbita pepo). The Styrian pumpkin, C. pepo subsp. pepo var. styriaca, is a variety that lost the seed testa following a single natural mutation in the 19th century (Teppner, 2004). Another cucurbit cultivated for its seeds is the West African egusi melon, C. mucosospermus, which has high nucleotide divergence from the domesticated watermelon (C. lanatus) in loci controlling fatty acid metabolism in seeds (Guo et al., 2013), consistent with the egusi melon's role as a seed crop. Mendelian genetics identified that a single recessive locus controlled the egusi seed type in Citrullus (Gusmini et al., 2004). In China, a form of watermelon with fleshy reddish seeds is grown for its seeds, which are toasted and eaten as a snack, and apparently this use dates to ad 1250 (Mote, 1977).

5. Selection for high carotenoid content

Several cucurbit crops accumulate β-carotene or lycopene, antioxidants that are highly beneficial to human health. Orange or red fruit pulp due to naturally high carotenoid concentrations are also found in wild species, for instance in the genera Momordica, Solena and Trichosanthes. Studies of crops have focused on watermelon and honey melon, in which the carotenoid accumulation results from the regulation of several metabolic genes (Garcia-Mas et al., 2012; Guo et al., 2013). In domesticated red-flesh watermelons, two key genes of the carotenoid biosynthetic pathway – phytoene synthase (PSY) and phytoene desaturase (PDS) – are upregulated during fruit development (Guo et al., 2011, 2015).

Lycopene accumulation in watermelon has been linked to 19 transcription factors, with the accumulation mirrored by low expression of lycopene cyclase, the enzyme that converts lycopene into β-carotene (Grassi et al., 2013). Another lycopene-rich cucurbit is the Xishuangbanna cucumber (Cucumis sativus var. xishuangbannanensis, Renner, 2017), a form of cucumber domesticated c. 2500 years ago by the Hani ethnic group in China, Laos and Vietnam, with high sugar and β-carotene content, similar to that of a melon. Its orange flesh is due to a substitution of alanine to asparagine at position 257 of the BCH1 gene, encoding a β-carotene hydroxylase (Qi et al., 2013). This mutation leads to a nonfunctional β-carotene hydroxylase that cannot convert β-carotene into zeaxanthin, leading to an accumulation of β-carotene (Qi et al., 2013).

Building on these results, Zhang et al. (2017) reported that high expression of a chromoplast-localised phosphate transporter (ClPHT4;2) is needed for lycopene accumulation. While ClPHT4;2 is present in both domesticated red-fleshed watermelon and white-fleshed close relatives, it is expressed at a very low level in white-fleshed watermelons. High expression of ClPHT4;2 relates to the presence of key cis-acting elements in its promoters that enables binding of sugar, abscisic acid and ethylene inducible transcription factors ClbZIP1 and ClbZIP2. These cis-acting elements are lacking in white-fleshed Citrullus (Zhang et al., 2017). That ClbZIP1 and ClbZIP2 are sugar-inducible led Zhang et al. (2017) to suggest that selection for sweetness and red colour occurred concurrently in the watermelon. However, the fruits of Citrullus lanatus subsp. cordophanus have white flesh combined with a high sugar content (Ter-Avanesyn, 1966).

V. Conclusions

We have highlighted the current state of research on Cucurbitaceae domestication, discussing both major and minor crops. Major crop species arose from annual and monoecious lineages within four of the 15 phylogenetically supported tribes of the family (Table 1), and were domesticated in North, Central and South America, Northeastern Africa, East, Southern, and Southeastern Asia, and on the island of New Guinea, in some cases overlapping with well known centres of plant and animal domestication. In many cases, the first trait selected on by man likely was nonbitterness of pulp, and comparative genomic analyses have revealed that, in cucumber, honey melon and watermelon, loss of fruit bitterness results from convergent domestication sweeps involving mutations at the Bt locus (Shang et al., 2014; Zhou et al., 2016). Both honey melon and watermelon were domesticated in Northeast Africa, perhaps in the Upper Nile region of what is now Sudan, but honey melon was also domesticated independently in Asia. Only a tiny subset of all flowering plants has been brought into cultivation with certain traits, such as an herbaceous habit or the ability to grow in disturbed anthropogenic environments, apparently key reasons. For Cucurbitaceae, monoecy combined with an annual life cycle appears to have been a factor (Fig. 1). Our review also highlights how the study of crop domestication needs to combine genomic efforts with traditional systematics and herbarium-verified specimens; this approach is the most efficient way to infer species’ geographic distributions and likely climatic adaptations.

Deciphering the domestication history of the major Cucurbitaceae crops will require further integration of population genomics with collection-based research and archaeological samples. The huge array of useful traits and the large genetic diversity found in the 23 minor cucurbit crops, as well as in wild progenitor populations of major crops, opens up a number of exciting avenues for de novo domestication, using CRISPR–Cas9 genome editing (Zsögön et al., 2018). This approach would be especially useful for plants like the watermelon, in which disease-resistance genes have been lost during domestication (Guo et al., 2013).


We thank three reviewers for their comments, and Harry Paris, Shaogui Guo and Bob Jarrett for discussion, and Oscar Alejandro Pérez-Escobar for generating the phylogeny shown in Fig. 4. Financial support came from the DFG (grants RE 603/27-1 and SCHA 1875/4-1) and the Elfriede and Franz Jakob Foundation for research at the Botanical Garden Munich, Germany. GC is supported by a Glasstone Research Fellowship and a Junior Research Fellowship at Queen's College, University of Oxford, UK.