Genetic Resources and Crop Evolution

, Volume 60, Issue 1, pp 175–192

Morphological diversity in breadfruit (Artocarpus, Moraceae): insights into domestication, conservation, and cultivar identification

Authors

  • A. Maxwell P. Jones
    • Department of BiologyUniversity of British Columbia
    • Department of ChemistryUniversity of British Columbia
  • Jim Wiseman
    • DigitalMedia Hawaii/Pacific
  • Diane Ragone
    • Breadfruit InstituteNational Tropical Botanical Garden
Research Article

DOI: 10.1007/s10722-012-9824-8

Cite this article as:
Jones, A.M.P., Murch, S.J., Wiseman, J. et al. Genet Resour Crop Evol (2013) 60: 175. doi:10.1007/s10722-012-9824-8

Abstract

Over millennia of breadfruit cultivation, hundreds of named cultivars have been developed that display a high degree of morphological diversity. The current study was undertaken to evaluate morphological diversity within the National Tropical Botanical Garden’s breadfruit germplasm collection, the largest and most diverse breadfruit collection in the world. A set of 57 standardized morphological descriptors including 29 leaf, 22 fruit, four seed, and two male inflorescence characteristics were used to describe and contrast 221 accessions of breadfruit including accessions of Artocarpus camansi Blanco, A. altilis (Parkinson) Fosberg, A. mariannensis Trécul, early generation A. altilis × A. mariannensis hybrids, and domesticated A. altilis × A. mariannensis hybrids. A morphological transition from heavily seeded fruit covered with flexible spines to fewer seeded, smoother skinned fruit of similar size was observed in the domestication of A. altilis from A. camansi. Further selection of true seedless, smooth-skinned cultivars of A. altilis appears to have occurred with human migrations from Melanesia into Polynesia. Cultivars from Micronesia exhibit morphological characteristics indicative of hybridization with the endemic species A. mariannensis. These data were used to generate a multi-access cultivar identification key on the Lucid platform that can be used to identify trees of known cultivars or to predict nearest cultivar relationships for previously undescribed cultivars. Overall, this study provides new insights into the morphological changes that occurred during domestication, helps visualize the diversity that exists across geographical regions, and provides a framework for cultivar identification and germplasm conservation.

Keywords

Agricultural biodiversityArtocarpusBreadfruitConservationDescriptorsDomesticationGermplasm

Introduction

Breadfruit, (Artocarpus, Moraceae) is an important staple crop traditionally cultivated throughout Oceania (Ragone 1997). The long history and importance of the crop is reflected in language, art and cultures. For example, in Pohnpei, Rahk is a term with a connotative meaning of season of abundant food, but is used more specifically in reference to the breadfruit season (Ragone and Raynor 2009; Sakiyama 1998). This ancient relationship between breadfruit and humans has resulted in the development of hundreds of cultivars known by thousands of different names (Ragone 1995). However, the recent trend towards urbanization and the incorporation of non-traditional foods have reduced the reliance on breadfruit in many regions of Oceania. As the reliance on breadfruit wanes, the threat of genetic erosion increases and many of these geographically restricted cultivars face the real threat of disappearing. Breadfruit has recently been included as one of 35 crops identified for their importance for food security and interdependence included in Annex 1 of the International Treaty on Plant Genetic Resources (ITPGR) and it has is classified as a priority crop by the Global Crop Diversity Trust (FAO 2009; http://www.croptrust.org/main/lprioritycrops.php). In order to develop a methodical approach to germplasm conservation of breadfruit it is important to first have detailed information on the amount of diversity that is present, how this diversity is geographically distributed, and to develop methods to identify morphologically unique accessions.

Several authors have recorded morphological descriptions for breadfruit cultivars in a number of different geographic regions (Koroveibau 1967; Navarro et al. 2007; Parham 1966; Ragone 1988, 1995, 2007; Ragone and Wiseman 2007; Sasuke 1980; Sreekumar et al. 2007; Wilder 1928). However, most of these studies did not use standardized methodology for data collection and were primarily focussed on limited geographical areas. The use of standardized measurements is important for the description of all plant species, but is particularly vital for the evaluation of breadfruit due to the high level of variability expressed even among clones of the same cultivar and branches within an individual tree (Ragone 1995). Leaf characteristics, such as the number of lobes, degree of dissection, size, and shape can vary between young shoots and older branches (Ragone 1995). Likewise, as the fruit develops, the skin texture, colour, amount of latex, and other morphological traits often change (Ragone 1995). This morphological variability makes direct comparison of previous studies problematic.

A set of 60 standardized morphological characteristics for the description of breadfruit morphology were recently developed (Ragone and Wiseman 2007). Provenance information and observation data for 183 accessions using these descriptors is available in the USDA National Germplasm Resources Information Network (GRIN) (http://www.ars-grin.gov/cgi-bin/npgs/html/crop.pl?217). The current study uses 57 of these descriptors to describe and contrast the morphology of 221 breadfruit trees growing in a field genebank in the National Tropical Botanical Garden in Hawaii. This repository is the largest and most diverse breadfruit collection in the world and represents trees from 31 Pacific islands spanning Melanesia, Micronesia, and Polynesia, the Philippines, Indonesia, and the Seychelles now growing in a single geographical area. This study represents the most comprehensive morphological comparison of breadfruit conducted to date. It provides insights into the complex heritage and domestication of breadfruit, identifies how the morphological diversity is geographically distributed, and is being used to develop a multi-access cultivar identification key to identify cultivars, clarify vernacular nomenclature, and guide future germplasm conservation initiatives.

Materials and methods

The breadfruit germplasm repository at the National Tropical Botanical Garden

The National Tropical Botanical Garden (NTBG) breadfruit germplasm collection includes 281 individual accessioned trees of breadfruit (Artocarpus altilis (Parkinson)Fosberg and A. altilis × A. mariannensis Trécul hybrids), breadnut (A. camansi Blanco), and dugdug (A. mariannensis) (Fig. 1). The trees were collected between 1978 and 2004 from 31 Pacific islands, the Philippines, Indonesia, and the Seychelles (Ragone 1997) and are conserved in the genebank at Kahanu Garden in Maui (20° 47′ 57.07″N, 156° 02′ 18.42″W). The collection is situated at an elevation of 15 m with a mean maximum temperature of 27.1 °C, mean minimum temperature of 19.7 °C, and receives an average of 2,051 mm of rain each year (Western Regional Climate Center; http://wrcc.dri.edu/). The soil within the collection is classified as Hana Very Stony Silt Clay Loam (http://websoilsurvey.nrcs.usda.gov). This soil is derived from volcanic ash, is typically well draining, slightly/moderately acidic and contains approximately 8 % organic matter in the surface horizon. Root-restrictive obstructions generally occur at a depth of 1.5 m. The soil surface layer is subtended by a base of deep lava. Soil nutrient analysis of this location during the study period can be found in Jones et al. (2011).
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9824-8/MediaObjects/10722_2012_9824_Fig1_HTML.gif
Fig. 1

Representative leaves, whole fruit, and halved fruit from Artocarpus camansi, A. altilis, A. mariannensis, early generation A. altilis × A. mariannensis hybrids, and late generation domesticated A. altilis × A. mariannensis hybrids conserved in the National Tropical Botanical Garden’s breadfruit germplasm repository

Data collection

A total of 57 morphological characteristics were measured using previously standardized descriptors (Ragone and Wiseman 2007). The descriptors include 18 quantitative and 11 qualitative leaf characteristics, 13 quantitative and nine qualitative fruit descriptors, two quantitative and two qualitative seed traits, and two quantitative measurements of the male inflorescences as described in Table 1. Leaf descriptors were measured from fully expanded leaves located three nodes from the distal end of mature branches to provide a standardized stage of development. Fruit descriptors were taken from mature, but not yet ripe fruit (Worrell et al. 1998). Descriptors of male inflorescences where taken when the flowers were mature and had dehisced. Seed descriptors were taken from fully developed seeds found in the mature fruit. Each descriptor was measured 10 times for each breadfruit accession (from 10 individual leaves, fruit, or inflorescences) providing a mean and distribution about the mean for quantitative traits and a distribution for qualitative traits.
Table 1

Morphological descriptors developed for the evaluation of morphological characteristics of breadfruit

Quantitative traits

Measure

Fruit weight

Weight of a whole fruit

Fruit length

Length of the fruit measured from the proximal to the distal end

Fruit width at middle

Diameter of the fruit half way between the proximal and distal end

Fruit width at top

Diameter of the fruit near the distal end

Fruit width at bottom

Diameter of the fruit near the proximal end

Core length

Length of the core measured from where it enters the fruit to where it ends

Core diameter

Diameter of the core at its widest point

Scabbing between sections

Amount of scabbing found between fruit sections 0: none, 1: slight, 2: moderate, 3: heavy

Scabbing around center

Amount of scabbing around center of fruit sections 0: none, 1: slight, 2: moderate, 3: heavy

Flesh color

Color of the fruit pulp; 1: white, 2: creamy, 3: light yellow, 4: yellow

Amount of latex

Amount of latex on cut surface of mature fruit; 0: none, 1: light, 2: heavy

Peduncle length

Length of peduncle

Number of seeds

Number of seeds that show on the surface of a longitudinally halved fruit

Seed length

Length of a mature seed

Seed diameter

Diameter of a mature seed

Male inflorescence length

Length of a mature male inflorescence

Male inflorescence width

Diameter of a mature male inflorescence at its widest point

Leaf length

Length of a mature leaf lamina not including the petiole

Leaf width

Width of a mature leaf lamina at its widest point

Width to length ratio

Leaf width divided by its length

Leaf margin

Description of leaf margin; 1: smooth, 2: slightly wavy, 3: moderately wavy/sinuate, 4: very wavy/undulate

Distance to widest point

Distance from leaf base to the widest point of the leaf

Number of lobes

Number of lobes on an individual leaf

Lobe spacing

Lobe spacing; 1: close, 2: close, overlapping, 3: wide, 4: wide, overlapping

Distance to 1st lobe

Distance from leaf base to the first lobe

Length of lobes

Average distance from midrib to the end of lobes on the leaf

Sinus depth

Distance from the deepest part of the leaf sinus to the midrib

Distance to first sinus

Distance from the base of the leaf to the center of the first sinus

Dissection ratio

Average sinus depth divided by the average length of the lobes

Petiole length

Distance from the basal end of the petiole to the beginning of the leaf lamina

Petiole diameter

Diameter of the petiole

Upper leaf hair amount

Amount of trichomes on the adaxial side of the leaf veins using a 0-5 scale

Upper leaf hair length

Length of trichomes on adaxial vein using a scale of 1–3

Lower leaf hair amount

Amount of trichomes on the abaxial side of the leaf veins using a 0–5 scale

Lower leaf hair length

Length of trichomes on abaxial vein using a scale of 1–3

Qualitative traits

Measure

Fruit shape

Shape of mature fruit; 1: round, 2: broad ovoid, 3: oval, 4: oblong, 5: ellipsoid, 6: heart-shape, 7: irregular

Skin texture

Skin texture at maturity; 1: smooth, 2: smooth with irregular raised sections, 3: sandpapery with persistent stigma dots, 4: flat pebbly, 5: round pebbly, 6: spiky raised centers, 7: pointed flexible spines

Skin color

Skin color at maturity; 1: dark green, 2: light green, 3: yellow green, 4: yellow, 5: brown/green, 6: pink/brick

Scabbing location

Scab location on mature fruit; 0: no scabbing, 1: between sections, 2: around center of sections

Color between sections

Color of scabbing between fruit sections; 1: green, 2: brownish

Color around center

Color of scabbing around center of fruit sections; 1: green, 2: brownish

Peduncle collar

Morphology of fruit collar; 0: none, 1: raised swollen sections, 2: flat with elongated sections

Peduncle insertion

Morphology of peduncle insertion point; 1: open, 2: depressed, 3: tightly clasped

Color of latex

Color of latex on the cut surface of a mature fruit; 1: green, 2: rusty reddish brown, 3: other

Seed color

Color of a mature seed coat; 1: off white, 2: light brown, 3: dark brown

Seed shape

Shape of mature seeds; 1: round, 2: ellipsoid, 3: elongated, 4: oblong, 5: kidney

Shape of leaf base

Shape of the distal end of the leaf lamina; 1: flat, 2: rounded, 3: acute

Shape of leaf apex

Shape of the proximal end of the leaf lamina; 1: rounded, 2: diamond shape

Blade color

Color of leaf lamina; 1: yellow/green, 2: green, 3: dark green

Vein color

Color of leaf veins; 1: green, 2: yellow/green, 3: yellow

Leaf surface

Appearance of leaf lamina; 1: glossy, 2: dull

Leaf flexibility

Flexibility of leaf lamina; 0: none, breaks when folded in hand, 2: flexible, bends without breaking

Velcro

Morphology of trichomes; 0: none or straight trichomes, 1: curled trichomes that act like velcro

Upper hair direction

Orientation of trichomes on adaxial leaf veins; 1: upright, 2: appressed

Upper hair color

Color of trichomes on adaxial leaf veins; 1: white, 2: reddish-white, 3: red

Lower hair direction

Orientation of trichomes on abaxial leaf veins; 1: upright, 2: appressed

Lower hair color

Color of trichomes on abaxial leaf veins; 1: white, 2: reddish-white, 3: red

Statistical analysis

All statistical analyses were conducted using JMP 8.0.2 (SAS Institute, Cary, NC). For each quantitative descriptor the mean and distribution about the mean was calculated for each accession. An analysis of variance (ANOVA) was conducted for each trait to evaluate if there was significant variation within the population for that trait. Accessions were grouped based on their region of origin (Melanesia, Micronesia, Western and Eastern Polynesia) and species (A. altilis, A. camansi, A. mariannensis, and A. altilis × A. mariannensis) to calculate a mean and distribution for each group. Grouping trees into the regions of origin was done to assess the morphological differences that exist among domesticated breadfruit cultivars originating from different geographical regions. As such, domesticated species were included while accessions of the wild progenitor species A. camansi, and A. mariannensis were omitted. Further, early generation A. altilis × A. mariannensis hybrids were also omitted as they represent volunteer trees with intermediate characteristics found growing in locations where seeded A. altilis and wild A. mariannensis grow in close proximity rather than truly domesticated cultivars. There are 281 accessioned trees in the collection including 210 A. altilis, 35 A. altilis × A. mariannensis hybrids, 24 A. camansi and 12 A. mariannensis (Table 2). For each quantitative trait an ANOVA was conducted using each of these grouping to determine if there were significant differences among them. Where significant differences were detected, a student mean separation using Tukey’s adjustment was conducted to determine what these differences were. Linear discriminant analysis was conducted using 29 of the quantitative descriptors with the aforementioned groupings to determine if the species and geographic origin of unknown trees could be discriminated based on their morphology using this method. Fruit stem length, petiole length, seed length, seed diameter, and male inflorescence descriptors were omitted from this analysis due to missing data for some of the accessions.
Table 2

Islands of origin of Artocarpus altilis and related species in the breadfruit germplasm bank at Kahanu Garden, Maui

Region

Island group

Number of Islands

Artocarpus altilis

A. altilis × A.mariannensis

Artocarpus camansi

Artocarpus mariannensis

Melanesia

Fiji

1

10

   

Papua New Guinea

1

  

19

 

Rotuma

1

11

   

Solomon Islands

1

10

   

Vanuatu

1

8

   

Micronesia

Chuuk, FSM

4

1

8

  

Kiribati

1

 

3

  

Mariana Islands

2

2

2

 

9

Palau

2

2

6

1

 

Pohnpei, FSM

1

15

12

1

3

Yap, FSM

1

 

2

  

Polynesia

Cook Islands

3

10

   

Hawaii

1

17

   

Marquesas Islands

1

9

   

Samoa

2

22

   

Society Islands

5

51

2

1

 

Tokelau

2

26

   

Tonga

1

2

   

Non-Pacific Islands

Indonesia

   

1

 

Philippines

   

1

 

Seychelles

 

4

   

Unknown

Unknown

 

10

   

Cultivar identification key

Plants are classified on the basis of the identification of characteristic descriptors that are often assembled as a botanical identification keys. The majority of botanical keys are dichotomous with a process of sequentially narrowing the potential identity of the specimen. In preliminary experiments, a wide variety of different statistical and visual tools were explored with the overall objective of creating a dichotomous botanical key that would accurately identify breadfruit cultivars, but this approach was not successful. Datasets of quantitative and qualitative descriptors were used to create a multi-access identification key on a Lucid 3.3 platform (Centre for Biological Information Technology; http://www.lucidcentral.com) using Lucid3 Builder for data compilation and Lucid3 Player to create a descriptor-based searchable key that will be available on the Breadfruit Institute’s website (Fig. 2; http://www.ntbg.org/breadfruit/).
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9824-8/MediaObjects/10722_2012_9824_Fig2_HTML.gif
Fig. 2

Screen shots from the prototype version of a multi-access breadfruit cultivar identification key on a Lucid 3.3 platform; a The initial start-up screen showing the general layout, b as the user is entering the data set on the filtered sorting mode that eliminates all cultivars that do not match with only 2 possibilities remaining, c after several descriptors have been entered using the ranked sorting mode which shows all cultivars in the database sorted by the number of matching descriptors, and d some images of a cultivar in the database available for the user to compare to their unknown cultivar

Results

Morphological comparison of species

Statistically significant differences were determined between Artocarpus altilis, A. camansi, A. mariannensis, early generation and domesticated A. altilis × A. mariannensis hybrids for the majority of descriptors evaluated (Table 3, Figs. 3, 4). Breadnut, A. camansi, is the putative wild progenitor of A. altilis (Zerega et al. 2004) and typically has large dark green or brown, oval fruit covered in long flexible spines that weigh an average of 1,317 g (SE ± 106.3). The fruit contain more seeds than any of the other species with 34.5 (SE ± 0.92) showing in a longitudinal section, and have a large central core measuring about 13.5 cm (SE ± 0.51) long and 5.3 cm (SE ± 0.15) wide. Breadnut has large leaves with an average length of 56.5 cm (SE ± 1.4) and an average width of 40.5 cm (SE ± 1.14). The leaves typically have about 7.9 (±0.28) lobes per leaf and are covered in trichomes on the upper and lower surface (Table 3).
Table 3

Average value of quantitative fruit, seed, flower and leaf descriptors of breadfruit germplasm grouped by species

Descriptor

Grouped by species

Artocarpus camansi

Artocarpus altilis

A. altilis × A. mariannensis

A.altilis × A. mariannensis*

Artocarpus mariannensis

Fruit weight (g)

1,317a

1,482a

1,457a

459b

444b

Fruit length (cm)

18.2ab

16.8b

19.1a

14.1c

13.3c

Fruit width at middle (cm)

14.1ab

14.0a

13.4b

8.4c

8.4c

Fruit width at top (cm)

8.1ab

8.9a

8.0b

5.6c

4.7c

Fruit width at bottom (cm)

10.0a

8.7a

8.1b

7.1c

6.0c

Core length (cm)

13.5a

10.4c

12.0b

10.1c

9.3c

Core diameter (cm)

5.3a

4.2b

3.6c

2.9d

2.8d

Scabbing between sections

1.3

1.13

0.78

0.32

0

Scabbing around center

1.30b

1.77a

1.54b

1.43b

1.52ab

Flesh colour

2.6bc

2.4c

2.6b

3.1a

3.2a

Amount of latex

0.53b

1.18a

0.82b

0.89b

0.95ab

Peduncle length (cm)

2.0b

6.3ab

6.0ab

4.3ab

9.9a

Number of seeds

34.5a

1.6b

0.4c

1.7bc

3.0bc

Seed length (cm)

2.7ab

2.8b

2.9ab

2.9a

3.0ab

Seed diameter (cm)

2.2b

2.3b

2.5a

2.5a

2.4ab

Male inflorescence length (cm)

23.8a

20.5b

22.5a

16.6c

17.9bc

Male inflorescence width (cm)

3.1a

3.2a

3.1a

2.7b

3.2a

Leaf length (cm)

56.5a

41.0c

44.8b

41.2c

28.8d

Leaf width (cm)

40.5a

33.5c

31.2d

35.6b

18.0e

Width to length ratio

0.72d

0.76c

0.79b

0.81a

0.62e

Leaf margin

2.2b

2.6a

2.4b

2.3b

1.8c

Distance to widest point (cm)

32.2a

24.3c

26.3b

23.8c

17.3d

Number of lobes

7.9a

7.1b

6.3c

6.4c

1.2d

Lobe spacing

2.48b

2.29b

2.99a

2.98a

1.98b

Distance to 1st lobe (cm)

24.1a

13.3c

14.3b

12.7c

11.9c

Length of lobes (cm)

28.7a

20.7c

23.0b

21.1c

9.6d

Sinus depth (cm)

9.5a

4.9c

5.0c

6.5b

3.3d

Distance to first sinus (cm)

10.8a

5.9d

7.0c

8.5b

6.5 cd

Dissection ratio

0.67c

0.74b

0.78a

0.69c

0.43d

Petiole length (cm)

5.1a

4.1d

4.5b

4.4bc

3.8 cd

Petiole diameter (cm)

1.28a

0.97c

1.05b

1.00bc

0.66d

Upper leaf hair amount

3.30a

2.18b

1.52d

1.77c

0.48e

Upper leaf hair length

1.25ab

1.35a

1.07b

1.31a

0.28c

Lower leaf hair amount

2.15ab

1.77b

1.29c

2.49a

1.60bc

Lower leaf hair length

0.98b

1.31b

1.17b

2.08a

1.20b

Means followed by different superscript letters are significantly different than the others in each grouping system based on Student’s means separation with Tukey’s adjustment using a type 1 error rate of 0.05

https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9824-8/MediaObjects/10722_2012_9824_Fig3_HTML.gif
Fig. 3

Distribution of qualitative morphological fruit and seed descriptors among 221 accessions of breadfruit conserved in the National Tropical Botanical Garden’s breadfruit germplasm repository grouped by species (Artocarpus camansi, A. altilis, A. mariannensis, early generation A. altilis × A. mariannensis hybrids, and late generation domesticated A. altilis × A. mariannensis hybrids) and by the region of origin (Melanesia, Western Polynesia, Eastern Polynesia, and Micronesia)

https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9824-8/MediaObjects/10722_2012_9824_Fig4_HTML.gif
Fig. 4

Distribution of qualitative morphological leaf descriptors among 221 accessions of breadfruit conserved in the National Tropical Botanical Garden’s germplasm repository grouped by species (Artocarpus camansi, A. altilis, A. mariannensis, early generation A. altilis × A. mariannensis hybrids, and late generation domesticated A. altilis × A. mariannensis hybrids) and by the region of origin (Melanesia, Western Polynesia, Eastern Polynesia, and Micronesia)

Breadfruit, A. altilis, exhibited a higher degree of variability than breadnut, but on average produces fruit of similar size as breadnut with a mean weight of 1,482 g (SE ± 15.1). However, the fruit of A. altilis are either few seeded, or in many cases, seedless, with an average of 1.6 (SE ± 0.13) seeds showing in a longitudinal section of a fruit. The core is significantly smaller than A. camansi measuring an average of 10.4 cm (SE ± 0.07) long and 4.2 cm (SE ± 0.02) in diameter (Table 3). The shape of the fruit is much more diverse than breadnut and includes cultivars with round, oval, oblong, ellipsoid, heart shape, and irregular-shaped fruit. Most breadfruit cultivars also have a raised or flattened, elongated collar that is generally absent in breadnut (Table 3). The texture of the fruit is variable, but is most often smooth, smooth with raised sections, or covered with small sharp spikes. Skin color is light green or yellow-green more often than the dark green and brown displayed by breadnut. Leaves of A. altilis are significantly smaller than A. camansi with an average length of 41.0 cm (SE ± 023) and an average width of 33.5 cm (SE ± 0.18). The leaves have an average of 7.1 (SE ± 0.05) lobes each and while they have significantly fewer trichomes than those of A. camansi, they are still moderately pubescent (Table 3).

Artocarpus mariannensis is a closely related species that is capable of hybridizing with seeded A. altilis. This species produces a much smaller fruit, with an average weight of 444 g (SE ± 130.2). The fruit are seeded, with about 3.0 (SE ± 1.06) seeds showing in a longitudinally sliced fruit (Table 3). While this number of seeds is statistically similar to what is found in A. altilis it must be recognized that the fruit in which they are produced is much smaller making the fruit proportionally more heavily seeded. The fruit are irregular in shape and the skin texture varies from rounded pebbly to flat pebbly. The outside of the fruit remains dark green even when ripe, but the fruit pulp is more yellow than either A. altilis or A. camansi. The leaves of A. mariannensis are significantly shorter than those of A. altilis or A. camansi with an average of 28.8 cm (SE ± 1.25) long and 18.0 cm (SE ± 1.02) wide (Table 3). The leaves have significantly fewer lobes, with an average of 1.2 (SE ± 0.25) per leaf and are less dissected with a dissection ratio of 0.43 (SE ± 0.02) compared to 0.67 (SE ± 0.02) and 0.74 (SE ± 0.004) for A. camansi and A. altilis respectively. The leaves have significantly fewer trichomes on the adaxial and abaxial surface of the leaf lamina, and the hairs that are present tend to have more of a red pigment than A. altilis or A. camansi (Table 3).

Interspecific hybrids between A. altilis and A. mariannensis were divided into two groups, early generation hybrids that have occurred naturally where the two species grow in close proximity, and highly domesticated cultivars that have been bred for hundreds or thousands of years (Fosberg 1960; Ragone 2007). The early generation hybrids produce fruit that most closely resemble its A. mariannensis parent. The fruit have an average weight of 459 g (SE ± 57), and produce a similar number of seeds with an average of 1.7 (SE ± 0.49) showing in a longitudinal section of fruit. Like its A. mariannensis parent, the fruit are irregularly shaped and have a flattened or rounded pebbly texture (Table 3). However, the fruit of the early hybrids sometimes turn a light green to yellow-green color at maturity. Unlike the fruit characteristics, the leaves of early generation hybrids tend to resemble their A. altilis parent to a higher degree with some traits being intermediate. On average, the leaves are similar in size to A. altilis at 41.2 cm (SE ± 0.55) long and significantly wider at 35.6 cm (SE ± 0.35). They have an average number of 6.4 (SE ± 0.11) lobes per leaf and a dissection ratio of 0.69 (SE ± 0.009), both traits are intermediate between the two parent species but closer to A. altilis (Table 3). The amount of trichomes on the adaxial side of the leaf is intermediate between the two parent species, but is greater than either on the abaxial surface. The trichome color varies from white to red (Table 3).

The second group of interspecific hybrids represent the morphological changes that have occurred over many generations of human selection (Table 3). The fruit of these domesticated hybrids are much larger than A. mariannensis or the early generation hybrids and similar to that of A. altilis and A. camansi with an average weight of 1,457 g (SE ± 32.0). Like cultivars of A. altilis, the majority of these advanced hybrids are seedless, with an average of 0.4 (SE ± 0.27) seeds showing in a longitudinally sliced fruit, and have relatively small core dimensions with an average length of 12.0 cm (SE ± 0.15) and diameter of 3.6 cm (SE ± 0.05). The color of the fruit pulp is intermediate between A. altilis and A. mariannensis. The shape of hybrid fruit is similar to A. altilis in that it varies widely, however, a larger proportion of hybrid cultivars produce irregularly shaped fruit indicative of their A. mariannensis heritage (Table 3). The texture of the hybrid fruit is most often flattened to rounded pebbly like A. mariannensis, but some cultivars have sharp spiked fruit observed in some cultivars of A. altilis. The leaves of domesticated hybrid cultivars are longer, but slightly more narrow than A. altilis with an average length of 44.8 cm (SE ± 0.48) and width of 31.2 cm (SE ± 0.39). The leaves have an average of 6.3 (SE ± 0.10) lobes; intermediate between the parent species but closer to A. altilis. The hybrids have the greatest dissection ratio of all species at 0.78 (SE ± 0.008). The amount of trichomes on the adaxial side of the leaves is intermediate between the parent species, while the amount on the abaxial side is lower than either (Table 3).

Discriminant analysis of the quantitative descriptors indicates that all five of these groups are morphologically distinct at 95 % confidence (Fig. 5a). Using this approach, it would be possible to differentiate between A. camansi and A. mariannensis from one another and from the remaining groups, but the high degree of morphological variation within each group makes it impossible to use this method alone to differentiate A. altilis from either group of hybrids. Overall, 13.9 % of the cultivars were misclassified using this analysis, mostly among the A. altilis and the hybrid cultivars. Discriminant analysis of the same traits with A. altilis and the hybrids combined in a single group resulted in only 1.3 % of the cultivars being misclassified, and the most commonly misclassified were early generation hybrids classified as A. mariannensis. As such, discriminant analysis using these traits is capable of discriminating species, but is not able to adequately distinguish A. altilis from A. altilis × A. mariannensis hybrids. This analysis also indicates that the early generation hybrids are morphologically intermediate between the two parent species. While this is also true for the late generation hybrids, they are morphologically more similar to the A. altilis parent.
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9824-8/MediaObjects/10722_2012_9824_Fig5_HTML.gif
Fig. 5

Linear discriminant analysis of 221 accessions of breadfruit breadfruit conserved in the National Tropical Botanical Garden’s breadfruit germplasm repository grouped by species (Artocarpus camansi, A. altilis, A. mariannensis, early generation A. altilis × A. mariannensis hybrids, and late generation domesticated A. altilis × A. mariannensis hybrids) and by the region of origin (Melanesia, Western Polynesia, Eastern Polynesia, and Micronesia) using 29 quantitative morphological fruit and leaf descriptors. The inner circle represents the 95 % mean of each group at 95 % confidence, and the outer circle represents the area that would include 50 % of the population; individual accessions are not shown to improve overall clarity

Morphological comparison by regions of origin

Significant differences were found among cultivars of breadfruit (both A. altilis and A. altilis × A. mariannensis hybrids combined) grouped by their region of origin for the majority of the descriptors (Table 4, Figs. 3, 4). A. altilis, was first domesticated in Melanesia and this region presumably has the most ancient cultivars (Ragone 1997). The breadfruit cultivars found in Melanesia produce the largest fruit, weighing an average of 1,634 g (SE ± 33.8). The fruit have relatively large cores at 11.9 cm (SE ± 0.16) long and 4.7 cm (SE ± 0.05) in diameter and are typically seeded with an average of 3.2 (SE ± 0.11) seeds showing in a longitudinally sliced fruit. The fruit pulp tends to be more yellow than cultivars from other regions, and the mature fruit accumulate a lot of latex on their surface (Table 4). The skin texture is variable, but the two most common textures are sharp spikes and rough, sandpapery. The leaves are a similar size among locations with the exception of Micronesian cultivars which are larger on average (Table 4). However, Melanesian cultivars have significantly more lobes than the others with an average of 8.2 (SE ± 0.09) per leaf. Melanesian cultivars also have significantly more trichomes on the adaxial surface of the leaf than cultivars from other regions (Table 4). The abaxial surface of the leaf has more trichomes than Eastern Polynesian or Micronesian cultivars, but less than Western Polynesian cultivars.
Table 4

Average value of quantitative fruit, seed, flower and leaf descriptors of breadfruit germplasm grouped by their region of origin

Descriptor

Grouped by region of origin

Melanesia

W. Polynesia

E. Polynesia

Micronesia

Fruit weight (g)

1,634a

1,378c

1,497b

1,405bc

Fruit length (cm)

18.4a

16.9b

17.0b

17.4b

Fruit width at middle (cm)

14.5a

13.8b

13.9b

13.5b

Fruit width at top (cm)

9.1a

9.0a

8.7a

8.3b

Fruit width at bottom (cm)

9.9a

8.6ab

8.6ab

8.3b

Core length (cm)

11.9a

10.9b

10.3c

10.4bc

Core diameter (cm)

4.7a

4.4b

4.0c

3.7d

Scabbing between sections

0.9b

1.0b

1.2a

1.0b

Scabbing around center

1.6b

1.6b

1.8a

1.7b

Flesh colour

2.8a

2.6b

2.2c

2.5b

Amount of latex

1.34a

1.35a

1.08b

0.85c

Peduncle length (cm)

8.5a

4.6c

6.7ab

5.6bc

Number of seeds

3.2a

3.2a

0.5b

0.2b

Seed length (cm)

2.7b

2.9a

2.7b

3.0a

Seed diameter (cm)

2.2c

2.5b

2.2c

2.6a

Male inflorescence length (cm)

21.3a

18.7b

21.6a

21.6a

Male inflorescence width (cm)

3.1b

3.0b

3.3a

3.0b

Leaf length (cm)

41.4bc

39.7c

41.7b

43.2a

Leaf width (cm)

31.4b

30.6b

31.9b

33.7a

Width to length ratio

0.76b

0.77a

0.77ab

0.78a

Leaf margin

2.34b

2.59a

2.63a

2.62a

Distance to widest point (cm)

25.0a

22.4b

24.8a

25.5a

Number of lobes

8.2a

7.7b

6.7c

5.7d

Lobe spacing

2.21c

2.33bc

2.47ab

2.57a

Distance to 1st lobe (cm)

12.5a

12.9a

13.1a

13.1a

Length of lobes (cm)

20.0b

19.0c

21.9a

22.1a

Sinus depth (cm)

5.1b

6.1a

4.5c

4.5c

Distance to first sinus (cm)

5.8b

5.9ab

6.2a

6.2ab

Dissection ratio

0.74b

0.67c

0.79a

0.72b

Petiole length (cm)

4.3a

4.1ab

4.0b

4.4a

Petiole diameter (cm)

0.91b

0.91b

1.02a

1.05a

Upper leaf hair amount

2.19a

2.16ab

1.95c

1.97bc

Upper leaf hair length

1.36a

1.27ab

1.25b

1.27ab

Lower leaf hair amount

1.96b

2.27a

1.46c

1.37c

Lower leaf hair length

1.44a

1.20b

1.09b

1.42a

Means followed by different superscript letters are significantly different than the others in each grouping system based on Student’s means separation with Tukey’s adjustment using a type 1 error rate of 0.05

The breadfruit cultivars originating from Western Polynesia are significantly smaller than those in Melanesia, with an average weight of 1,378 g (SE ± 36.9). The fruit have a proportionally smaller core with an average length of 10.9 cm (SE ± 0.17) and an average width of 4.4 cm (SE ± 0.05) and a similar number of seeds with an average of 3.2 (SE ± 0.10). The fruit pulp is significantly less yellow than the Melanesian cultivars, but the fruit accumulate a similar amount of latex on their surface at maturity (Table 4). The skin textures are most often sandpapery, flattened pebbly, or smooth with irregularly raised sections and they sometimes have sharp spikes, although this is less frequent than in Melanesian cultivars. The leaves of Western Polynesian cultivars have fewer lobes than Melanesian cultivars with an average of 7.7 (SE ± 0.10) lobes per leaf. The adaxial leaf surface has a similar number of trichomes, and the abaxial surface has more than the Melanesian cultivars (Table 4).

Cultivars from Eastern Polynesia produce larger fruit than Western Polynesia, but smaller than those originating from Melanesia with an average weight of 1,497 g (SE ± 22.5). The core is proportionally smaller than Melanesian or Western Polynesian cultivars measuring 10.3 cm (SE ± 0.11) long and 4.0 cm (0.03) wide. Cultivars in this region are predominantly seedless with some producing a few seeds with an average of 0.5 (SE ± 0.07) seeds showing in a fruit sliced longitudinally. The pulp is significantly less yellow than fruit from Melanesia or Western Polynesia, and less latex accumulates on the surface at maturity (Table 4). The most common skin texture of fruit from this region is smooth with irregularly raised sections, followed by sandpapery, and a few bear the sharp spikes common in Melanesia. The leaves have fewer lobes than Melanesian or Western Polynesian cultivars with an average of 6.7 (SE ± 0.06) and have fewer trichomes on both the abaxial and adaxial surfaces (Table 4).

Micronesian cultivars produce fruit similar in size to cultivars from Western and Eastern Polynesia with an average weight of 1,405 g (SE ± 30.9). The core of the fruit is similar in length to that of Eastern and Western Polynesian cultivars at 10.4 cm (SE ± 0.15), but is narrower than either with a diameter of 3.7 cm (SE ± 0.04). Micronesian cultivars are most often seedless, or have very few seeds with an average of 0.2 (SE ± 0.04) showing in a longitudinally sliced fruit. The flesh color is similar to cultivars from Western Polynesia, less yellow than those from Melanesia but more yellow than Eastern Polynesian cultivars (Table 4). At maturity, Micronesian cultivars accumulate significantly less latex on the surface of the fruit than any other region. While fruit shape varies among cultivars from all regions, there is a greater tendency for Micronesian cultivars to produce irregularly shaped fruit (Table 4). They also most often have flattened to rounded pebbly fruit textures, with spiked, sandpapery, and smooth being less common than in other regions. The leaves of Micronesian cultivars are larger than those from the other regions with an average length of 43.2 cm (SE ± 0.45) and width of 33.7 cm (SE ± 0.37). However, their leaves have the lowest number of lobes with an average of 5.7 (SE ± 0.08). The adaxial surface of the leaves have a similar amount of trichomes as the Polynesian cultivars. The abaxial surface has relatively few trichomes similar to Eastern Polynesian cultivars (Table 4).

Discriminant analysis of the quantitative descriptors indicates that cultivars from Melanesia, Western Polynesia, Eastern Polynesia and Micronesia are distinct at a confidence level of 95 % (Fig. 5c). However, due to the high degree of morphological variability among cultivars collected from each of these regions 34.2 % of the cultivars were misclassified using these characteristics alone. As such, these descriptors can provide some clues to the origin of an unknown cultivar, but cannot provide a definitive answer.

Using descriptors to distinguish cultivars

Initial efforts were made to create a dichotomous key for the identification of unknown breadfruit cultivars but were unsuccessful due to the seasonality of the species, variability of some descriptors within a single tree or between trees of the same cultivar, and overlap between different cultivars, making a discrete separation impossible. Therefore, a multi-access key was developed using Lucid 3.3 as a platform. A database containing the 57 morphological descriptors for 221 accessions of breadfruit in the NTBG collection was developed using the Lucid Builder system. For qualitative traits, the most frequent state for each cultivar was entered as the common score, and less frequent states were entered as rare scores. In cases where two states were equally frequent they were both entered as common scores. For quantitative traits the maximum and minimum of the 10 values recorded were entered into the database as the extreme maximum and minimum scores.

The Lucid system allows the user to begin with the descriptor characteristic that is most available and to enter the data in any order they choose (Fig. 2). As the user enters more data, the system eliminates the cultivars in the database that do not match and ranks the cultivars based on their similarity to user-entered data. At any time, the user will be able to select a potential match, view larger images and be directed to a detailed description of the cultivar. As new cultivars are identified they can be added to the database so that it can continually grow and improve over time.

Discussion

The National Tropical Botanic Garden (NTBG) maintains the largest and most diverse collection of breadfruit in the world with accessions collected throughout the crop’s traditional range (Ragone 2007). Within this collection exist accessions representative of the entire domestication process including the two wild progenitor species Artocarpus camansi and A. mariannensis, seeded diploid and seedless triploid A. altilis, early generation inter-specific hybrids between A. altilis and A. mariannensis, and highly domesticated advanced A. altilis × A. mariannensis hybrids (Ragone 1997, 2001, 2007; Fig. 6). Breadfruit is most often asexually propagated, and its spread and domestication is intimately linked to human migration patterns (Zerega et al. 2004, 2006). As humans migrated out of Melanesia and colonized the Pacific Islands, breadfruit and other crops collectively referred to as “canoe plants” were carried with them. Each time a new island group was colonized, the settlers would select their favourite cultivars of each crop to bring with them. Over many generations, repeated vegetative selection such as this can lead to dramatic morphological changes in the crop such that they no longer resemble their wild progenitors. These changes, often referred To as the crops “domestication syndrome (Hammer 1984), generally include traits that make the crop more productive or more suitable for cultivation. In the case of breadfruit, this process of repeated bottlenecks appears to have resulted in heavy selection pressure evident by the high level of morphological diversity observed in breadfruit from different regions. A detailed morphological comparison of the NTBG’s collection of breadfruit germplasm described herein provides new insights into the domestication of this species and how it is related to human migration, provides a framework for cultivar classification, and identifies regions and accessions with distinct morphological traits that will help guide future germplasm conservation initiatives.
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9824-8/MediaObjects/10722_2012_9824_Fig6_HTML.gif
Fig. 6

A flow chart depicting the domestication of breadfruit starting with a the wild ancestor Artocarpus camansi transitioning into b Seeded diploid A. altilis and its final development into c seedless triploid A. altilis, d the closely related wild species A. mariannensis which can hybridize to produce e early generation interspecific A. altilis × A. mariannensis hybrids that represent the basis for the development of f domesticated A. altilis × A. mariannensis hybrids

Previous morphological comparisons and more recent molecular evidence suggests that breadfruit, Artocarpus altilis (Parkinson) Fosberg, is a cultigen originally derived from the wild species Artocarpus camansi Blanco (Ragone 1997; Zerega et al. 2004, 2006). Artocarpus camansi is native to Papau New Guinea, and possibly the Moluccas (Ragone 1997). The initial domestication of A. altilis from A. camansi likely occurred in New Guinea or the surrounding islands before continuing as it was moved eastward (Zerega et al. 2004, 2006). The domestication syndrome (Hammer, 1984) exhibited during the transition from A. camansi to A. altilis is interesting in that the average fruit size has not increased relative to the wild progenitor species (Table 3). Many species, including the closely related jackfruit, A. heterophyllus, have significantly larger fruit than their wild counterparts (Khan et al. 2010). However, while domesticated breadfruit is similar in size and weight to breadnut, the fruit has significantly fewer seeds and in many cases is seedless (Table 3). This is the major change that enabled the transition from a seed crop into the important starchy staple that it is today. Other significant differences in fruit morphology that occurred during domestication include a reduction in the length and width of the core, and the surface of the fruit transitioned from being covered by long flexible spines towards a smoother texture. Some of these traits such as the reduction in core width and shape increase the edible starchy portion of the fruit (Jones et al. 2011). The transition from long flexible spines to a smoother skin texture may have been selected for because breadfruit is often peeled before cooking or being preserved by pit fermentation (Cox 1980; Ragone 1997, 2001), or could be in part a consequence of the morphological changes that contribute to reduced seed production (Hasan and Razak 1991). The reduction in leaf size exhibited by A. altilis relative to A. camansi is also different than what have been observed in the domestication of A. heterophyllus, where cultivated varieties have larger leaves than wild trees (Khan et al. 2010).

After the initial domestication of A. altilis, the early cultivars accompanied pioneers as they colonized the Pacific islands (Zerega et al. 2004, 2006). These early domesticates from New Guinea travelled east into the rest of Melanesia and then into Polynesia (Zerega et al. 2004). During the course of this migration, the process of domestication continued and several distinct trends in breadfruit morphology occurred. The most well-documented of these changes is the further reduction in seed number as this crop moved east (Ragone 1997, 2001). The seeded cultivars found in Melanesia give way to fewer-seeded diploid and true seedless triploid cultivars in western Polynesia, and almost exclusively seedless triploid cultivars in Eastern Polynesia indicating that there was a strong selection pressure for seedlessness as peoples migrated east (Ragone 2001; Zerega et al. 2004, 2006). The overall morphology of cultivars from Melanesia and Western Polynesia is more similar to each other than cultivars from Eastern Polynesia. This may indicate that there was a significant event in breadfruit selection that occurred somewhere in Polynesia that affected its overall morphology, perhaps the development of triploid cultivars (Ragone 2001). Unlike other crops such as A. heterophyllus where increased fruit size was selected for (Khan et al. 2010), the size of breadfruit declined as it moved out of Melanesia and into Polynesia. However, this reduction in fruit size is accompanied by fewer seeds and a continued reduction in the length and width of the core which increases the proportion of starchy pulp in each fruit. The reduction in yellow pigmentation in the flesh of Polynesian cultivars may indicate a preference for white starch which is common in some other staple crops such as rice (Sweeney et al. 2007). The transition from spiny skin towards a smoother surface that occurred during domestication appears to have been further selected for as the crop moved east into Polynesia. While there are some trends in leaf characteristics, such as a reduction in the number of lobes and a reduction in the amount of trichomes present, leaf morphology appears to change less than fruit characteristics suggesting there was less selective pressure for leaf traits.

The history of breadfruit domestication is further complicated when it was introduced into Micronesia where the closely related endemic species A. mariannensis Trécul originates (Fosberg 1960; Ragone 1997). These two species are sexually compatible and based on morphological characteristics and AFLP markers, many Micronesian cultivars are interspecific hybrids between the two (Fosberg 1960; Zerega et al. 2005). Early generation hybrids provide insights into the early stages of the domestication process. The fruit from these early hybrids resemble those of A. mariannensis to a large extent. However, the leaves are morphologically more similar to A. altilis, with many leaf traits being intermediate between the two. These early hybrids likely provided early settlers with agronomic or nutritional benefits as they were perpetuated and developed into a highly domesticated crop. Hybrids are known to be more tolerant of saline soils and do well on low lying atolls (Ragone 1997), often have altered production seasons (Jones et al. 2010) contain significantly higher levels of iron and some other minerals (Jones et al. 2011), and contain higher concentrations of nutritionally important carotenoids (Jones 2010). These traits may have provided the necessary advantage to early breeders to develop these hybrids into the large-fruited, seedless cultivars that exist today. The domesticated hybrid cultivars are morphologically more similar to A. altilis than the early hybrids (Fig. 5a). This may be a consequence of subsequent backcrossing with A. altilis, the selection for similar characteristics, or a combination of the two. Regardless, some of the domesticated hybrid cultivars produce seedless fruit that weigh over 3.5 kg each, making them some of the largest-fruited cultivars in the NTBG collection (Jones et al. 2011). This hybridization and subsequent domestication has resulted in an overall morphologically distinct set of cultivars within Micronesia (Fig. 5c).

While the differences in morphological characteristics among regions reveal a pattern in the ongoing domestication pattern of breadfruit, it is important to recognize the high level of diversity that has been preserved within each geographical region. More than 2000 vernacular cultivar names have been recorded across Oceania (Ragone 1995), and even within single island groups a large number of cultivars are maintained. For example, 130 cultivar names have been recorded from Pohnpei (Fownes and Raynor 1993; Ragone and Raynor 2009), over 40 from Samoa (Ragone et al. 2004), and 132 from Vanuatu (Walter 1989). Describing and contrasting the overall morphology of such a large number of cultivars presents significant challenges. Multivariate statistics such as discriminant analysis provide powerful tools to assess the overall morphological similarity of germplasm collections that can remain undetected using univariate methods (Carter 1987; Erskine et al. 1989; Flores et al. 1997; Veronesi and Falcinelli 1988). These methods have been used to identify regions with morphologically distinct cultivars to help guide conservation efforts in several other crops (Carter 1987; Erskine et al. 1989; Flores et al. 1997; Veronesi and Falcinelli 1988). Discriminant analysis of breadfruit descriptors indicates that cultivars originating from Melanesia, Western Polynesia, Eastern Polynesia and Micronesia are morphologically distinct from one another (Fig. 5c). However, due to the morphological diversity exhibited among cultivars within each region, discriminant analysis using these traits misclassified 23.3 % of the cultivars. While this approach provides some insight into the morphological relationship among these regions and indicates that germplasm conservation efforts will need to include them all, it is impossible to reliably determine the region of origin of unknown cultivars using this model.

Cultivars are traditionally named and identified based on their morphological characteristics such as skin texture or leaf shape (Ragone 1991; Ragone et al. 2004). The traditional system of nomenclature varies among indigenous peoples and there is no unified method of identification or classification. Additionally, while thousands of cultivar names have been documented, the actual number of morphologically or genetically distinct cultivars is unknown as it is possible to have multiple names for what is actually a single cultivar or two morphologically distinct cultivars that are known by a single name (Ragone 1995, 2007). These issues were the impetus for the development of a universal cultivar identification key. A multi-access key was selected over the more traditional dichotomous key because it allows more flexibility for the user as it allows them to use the characteristics that are available to them, and is highly interactive. The software has the ability to function by eliminating all trees that do not match the selection criteria, or by ranking the cultivars based on the percent of selected traits that match thus identifying other similar cultivars. As the user narrows down potential matches, a series of detailed photos and other relevant information can be accessed to help in identification. A prototype of this identification key is currently being field tested and will soon be publically available on the Breadfruit Institute’s website (http://www.ntbg.org/breadfruit/). This identification key will aid in the identification of cultivars, help identify cultivars with novel traits for germplasm conservation, and help clear the ambiguity of vernacular cultivar names.

Conservation of breadfruit diversity has been identified as a high priority. It is included in Annex 1 of the International Treaty on Plant Genetic Resources for Food and Agriculture and has been identified as a priority by the Global Crop Diversity Trust (FAO 2009; http://www.croptrust.org/main/lprioritycrops.php). The current study provides the most detailed and comprehensive morphological comparison of breadfruit germplasm conducted to date, providing new insights into the morphological changes that occurred during the domestication of breadfruit and a basis to evaluate the variation that exists among species and geographical regions. Multivariate statistics and the development of a universal cultivar identification key described herein can be used to help identify known and previously undescribed cultivars and regions that exhibit unique morphological traits to guide conservation effort.

Acknowledgments

The authors would like to acknowledge the Trustees and Fellows of the National Tropical Botanical Garden who supported this project. Funding was also provided by the United States Department of Agriculture Agricultural Research Service through CA No. 58-5320-5-765. We are grateful for the guidance and assistance provided by Dr. Francis Zee and Carol Mayo Riley at the USDA Pacific Basin Agricultural Research Center in Hilo, and Kamaui Aiona at Kahanu Garden.

Copyright information

© Springer Science+Business Media Dordrecht 2012