HISTÓRIA TAXONÔMICA E FILOGENIA DA PUPUNHA

8/9/2019 HISTRIA TAXONMICA E FILOGENIA DA PUPUNHA

1/154

CHAPTER 1

INTRODUCTION

The pupunha (Bactris gasipaes Kunth) is the only

domesticated neotropical palm species and is found only as

a cultivated plant (Fig. 01, a, b). It was selected and

artificially distributed in the past by the Amerindians,

well before the first Europeans arrived in the Americas

(Clement, 1988). Today, the artificial selection and

distribution of pupunha has been conducted mainly by

research centers and commercial planters. As a result, its

current distribution ranges from Northern Honduras, in

Central America, to southeastern Bolivia and the Atlantic

coast of Brazil, in South America (Fig. 02). It grows more

easily in the lowlands and humid areas with short dry

seasons, although it is possible to find it in the Andean

foothills of Bolivia, Ecuador, Colombia and Venezuela

(Mora-Urp, 1984).

In the past, several specimens of cultivated pupunha

were described as distinct species or varieties. A few

wild palms, also called pupunha by local people, were

described as species and considered to be closely related

to the domesticated plants. With a few exceptions, all

described species and varieties of the cultivated and wild

pupunha share two main characters. They all have large

1


2/154

stems, the biggest in the genus Bactris, and flattened

fibers adnate to the endocarp. The cultivated plants are

easily recognized by theirs large fruits, usually more than

3 cm long, and ovoid shape. The wild species are

vegetatively very similar to the cultivated plants, but

usually have much smaller globose fruits.

The main products obtained from the cultivated pupunha

are edible fruits and the palm heart, which rivals in

quality that extracted from Euterpe palms (Fig. 03, a, b).

The fruits are widely sold and are eaten after cooking.

They also can be used for oil extraction, flour

preparation, and animal feed (Clement, 1995). More

recently palm heart extraction is becoming the most

important economic use of this palm. In Brazil pupunha is

now reaching the status of an industrial crop due to the

establishment of large plantations for palm heart

extraction (Fig. 04).

Although important economically, there are

several disagreements regarding the taxonomy of pupunha.

It is not totally clear what are the correct names to be

applied to the cultivated and wild species, how many of the

species can be considered as valid, synonyms, or simple

varieties. It is also unknown what taxon or taxa are

related to the domestication process that gave rise to the

2


3/154

cultivated pupunha and one important question is still

unanswered: in what place was pupunha domesticated?

Due to the confusion related to its taxonomy, origin,

and ancestors involved in its domestication, there is an

unavoidable necessity to present a detailed and careful

account of the taxonomic history and origin of pupunha.

For this reason, this study is presented in two distinct

sections. The first section is dedicated to a review and

discussion of the several aspects related to the taxonomic

history and origin of pupunha. The second section

presents: (a) a review and discussion of the recent

taxonomic and cladistic studies of Bactris and the

placement of pupunha and related species, (b) a discussion

of the classification system for cultivated and wild

pupunha developed by agronomists, and (c) a cladisitic

analysis of pupunha and related species that was performed

during the development of the present study.

The results of the cladistic analysis are used to

propose a new phylogenetic arrangement for the cultivated

and wild species, and to indicate the most likely place

where pupunha was domesticated.

3


4/154

SECTION I

HISTORY AND ORIGIN OF PUPUNHA (Bactris gasipaes Kunth).

CHAPTER 2

TAXONOMIC HISTORY OF PUPUNHA (Bactris gasipaes Kunth) AND

ALLIED SPECIES.

Introduction

The binomial Bactris gasipaes was first used by Kunth

(1816) to describe cultivated plants found in Colombia by

Humboldt and Bonpland. However, the taxonomic history of

pupunha began eighteen years before Kunths publication,

when Ruiz and Pavn (1798) described Martinezia ciliata

from Peru. It is not clear, however, if Ruiz and Pavn

described their species from cultivated or wild plants.

Martinezia Ruiz & Pavn was rejected as a genus name for

pupunha species and M. ciliata was transferred to Bactris

by Martius (1826).

4


5/154

Martius (1824) created the genus Guilielma and

described G. speciosa from cultivated plants he had found

in the eastern Brazilian Amazon. The genus Guilielma was

widely accepted at that time and many palm species and

varieties were described or transferred under this name in

the following years. This situation persisted until the

beginning of the 1960's, when Guilielma was replaced by

Bactris as a valid genus in several publications (MacBride,

1960; Wessels Boer, 1965, 1988; Uhl and Dransfield, 1987).

However, some non-taxonomist researchers working with the

cultivated pupunha started challenging this new proposition

suggesting that Guilielma could not be discarded as a valid

genus at all (Mora-Urp, 1984; Clement, 1988). The

discussion about the validity of Guilielma was definitely

over when Sanders (1991) published the first cladistic

study of Bactris in which he included Guilielma. Despite

being criticized, Sanders results regarding the monophyly

of Bactris are accepted by current palm taxonomists

(Henderson, 1995; 1996; Henderson et al. 1995; d'Granville

1997). The reason for this acceptance is that Bactris is

not monophyletic if Guilielma is segregated. In fact,

Bactris would have to be divided into eight or more poorly

defined genera in order to keep Guilielma apart. Sanders

gave the name Guilielma to the clade that included the

pupunha species, both cultivated and wild.

5


6/154

In the following pages I will review and discuss all

species and varieties of pupunha described in the past

(Guilielma clade Sanders). In this review I will also

include Bactris setulosa, Bactris riparia, and all taxa

currently accepted as synonyms of these species by Wessels

Boer (1988), Henderson (1995) and Henderson et. al.,

(1995). The reason for this inclusion is explained in the

topics I discuss each of these species.

The taxa of the cultivated and wild pupunha

The species and varieties here included have been

considered by several authors (Clement, 1988; Henderson,

1995; Henderson et. al., 1995; Mora-Urp, 1944, 1994, Mora-

Urp and Clement, 1981; Wessels Boer, 1965, 1988) as the

ones that better represent the several forms of the

cultivated and wild pupunha. The historical aspects and

the taxonomic disagreements over the correct use of the

names are discussed in most cases.

Martinezia ciliata Ruiz & Pavn, Systema veg. 275.

1798. Bactris ciliata (Ruiz and Pavn) Martius, Hist.

Nat. Palm. 2: 95, t. 71 fig. 3, 1826. Guilielma ciliata

(Ruiz and Pavn) Wendland in Kerchove, Palm. 246.

1878. Type. Peru. Hunuco: Pozuzo, s.d., Pavn s.n.

(Type at MA, not seen).

6


7/154

This was the first pupunha to be described. Ruiz and

Pavn, however, used the genus name Martinezia and included

under it several species that are now known to belong to

other palm genera, such as M. ensiformes (=Prestoea

ensiformis), M. linearis (=Chamaedorea linearis)and, M.

interrupta (=Geonoma interrupta). As a result, Martinezia

Ruiz and Pavn was rejected, and is now recognized only as

a synonym of Prestoea (Uhl and Dransfield, 1987). Martius

(1826) studied the type of M. ciliata and transferred it to

Bactris. It is unknown, however, the reasons why Martius

did not transferred M. ciliata to the genus Guilielma,

which he himself had created. He probably did not noticed

the endocarps of Pavn's specimen were similar to his G.

speciosa, although of a smaller size. The transfer of M.

ciliata to Guiliema was made later by Wendland (in

Kerchove, 1878).

B. ciliata is still poorly known and, besides the

type, apparently no additional collections from the type

locality have been made. Bernal (1989) believe it is the

same species as B. gasipaes Kunth and proposed the

conservation of the later name, arguing it has been used

for a long time in the botanical and agricultural

literature. However, it is difficult to conclude if the

endocarps of the type specimen (illustrated in Martius,

1826) are from cultivated or wild plants. Ruiz and Pavn

7


8/154

did not note in the description if the fruits were edible,

a common characteristic of all domesticated pupunha. Ruiz

(in Dahlgren, 1940) commented only about the uses of the

palm heart, reported to be tender and tasty, raw or

cooked, but reported to be coarse. Palm hearts of

cultivated pupunha are known by their high quality, while

many wild pupunha are scarcely reported as having edible

palm heart. Ruiz noted the uses of the palm trunk, but

this does not help decide if the palm is domesticated or

wild. The final answer for this puzzle will be achieved

only when new collections from the type locality is made.

Bactris gasipaes Kunth, in Humboldt, Bonpland &

Kunth, Nov. gen. sp. 1: 302. 1816. Guilielma

gasipaes (Kunth) L. H. Bailey, Gentes Herb. 2: 187,

1930. Type specimen. Colombia, Tolima: Ibagu, n.d.,

Bonpland s.n. (Type at P, not seen; F neg. 38701).

Kunth described this species from material collected

in Colombia by Bonpland and Humbold during their voyage to

tropical America. The epithet is derived from the common

name "gachipaes" given to the species in that region of

Colombia. For many years it was accepted to attribute to

Humbold, Bonpland and Kunth (H.B.K.) the description of

this species, however, it is now known that Kunth was the

one who really wrote the description, therefore the correct

authority for the name must be given to him alone.

8


9/154

Guilielma Martius, Palm Fam. 21, 1824.

Martius never gave or discussed the reasons why he

decided to create Guilielma as a new genus apart from

Bactris, already known for almost 50 years. He knew that

Kunth had published B. gasipaes from Colombia and it was a

cultivated species, similar to the first Guilielma species

described by himself as G. speciosa (see discussion about

this species).

The adoption of Guilielma as a valid genus in the past

century was, perhaps, due to Martius influence in palm

taxonomy at that time. Karsten (1857) was the only one to

challenge Martius proposition. Wendland (in Kerchove,

1878) transferred several species from Bactris to

Guilielma, but never discussed the reasons why he decided

to do so. Furthermore, his publication is essentially a

list of previously published palm species. Burret (1934),

on the other hand, in his taxonomic treatment of Bactris

decided to keep Guilielma as a segregated genus and gave

data to support it. Sanders (1991) showed, however, that

Burret had found only two reliable characters to maintain

his proposition: a massive stem and an endocarp with adnate

9


10/154

bands. I consider that massive stems can not be used as a

character to segregate these genera since non-Guilielma

species of Bactris, such as B. setulosa, B. riparia, B.

guineensis and all the Antillean species, can have a

massive stems and were never described under the genus

Guilielma. In following this logic, Astrocaryum, a closely

related genus to Bactris, would have to be segregated

because it also has species with massive or slender stems.

The endocarp with an adnate band is characteristic almost

unique to all species originally described as Guilielma,

but it can be also found in some specimens of B. setulosa

and B. riparia. It would be very difficult to justify

segregating Guilielma on the basis of only one good

character.

It is very common to find in papers, especially those

by agronomists (Mora-Urp, 1984; Clement, 1988; Silva,

1994) suggestions that the controversy about the validity

of Guilielma is not resolved, and that the "modern

tendency" is to accept this genus as a synonym of Bactris.

In my opinion Sanders (1991), has exhausted this

discussion, and I cannot see reason for any further

suggestion of Guilielma being a segregated genera. Mora-

Urp (1984) presented a incorrect discussion of the pupunha

taxonomy. I think Mora-Urp's publication deserves some

comments to prevent unaware researchers, especially those

unfamiliar with taxonomy, for being confused by wrong

10


11/154

concepts regarding Bactris. First: not all species of

Bactris have a staminoidal ring. This character is found

in only a small group of black-fruited species (the

Pyrenoglyphys group). There are Bactris species not

closely related to the Guilielma clade without a

staminoidal ring and variously colored fruits, such as red,

yellowish and even black fruits. Second: crispate leaves

are throughout Bactris, and are not related to a specific

group of species.

Guilielma speciosa Martius, Hist. Nat. Palm. 2: 82.

1824. Bactris speciosa (Martius) Karsten, Linnaea

28: 402. 1857. Type. Brazil. South of Par, n.d.,

Martius s.n.(Type at M, not seen).

This was the first species described under Guilielma.

Martius considered B. gasipaes Kunth as synonym of this

species, in a clear transgression of the nomeclatural

rules. Spruce (1871) commented that speciosa was not the

epithet for this species, and Guilielma was

indistinguishable from Bactris. Martius' G. speciosa is a

nomen illegitimun according to article 63.1 of the ICBN

(Bernal, 1989). If B. gasipaes Kunth was placed as a

synonym of G. speciosa, in order to follow the Code,

Martius should have made the combination G. gasipaes, a

correction proposed by Bailey (1930) more than 100 years

later.

11


12/154

Karsten (1857) transferred G. speciosa to Bactris when

he described his variety chichagui. Karsten understood

that all species without androecium structures in the

corolla of the pistillate flowers should have been placed

under Bactris. Therefore, if B. speciosa (Mart.) Karsten

is taken alone, it is a superfluous name (as implied by

Wessels Boer, 1965; 1988), because G. speciosa was

described with B. gasipaes as a synonym, and two names can

not be used to represent a single taxon.

G. speciosa represents the eastern Amazonian

population of the cultivated pupunha and is closely related

to the small-fruited B. insignis (G. insignis) from

Bolivia, according to Mora-Urp (cited in Clement, 1988).

Bactris speciosa (Martius) Karsten var. chichagui

Karsten, Linnaea 28: 402. 1857. Guilielma gasipaes

(Kunth) L. H. Bailey var. chichagui (Karsten)

Dahlgren, Field Mus. Nat. Hist. Bot. 14: 185, 1936.

Type. Colombia. Magdalena and Cauca. (Type

destroyed?).

As commented above, Martiu's G. speciosa is a nomen

illegitimun. Should Karsten's var. chichagui be considered

also a nomen illegitimun? No according to the article 68.2

of the ICBN.

12


13/154

Guilielma speciosa Martius var. flava Barb. Rodr.

Enum. Palm. Nov. 23, 1875. Guilielma gasipaes

(Kunth) L. H. Bailey var. flava (Barb. Rodr.) L. H.

Bailey, Gentes Herb. 2: 187, 1930. Lectotype

designated by Henderson (1995): Barb. Rodr., Sert.

Palm. Brasil. 2: t. 52b. 1902.

Barbosa Rodrigues was the first to propose a

classification of the cultivated varieties of pupunha.

However his system was never adopted because it is based on

fruit color, which is not a good character to segregate

cultivated traits of pupunha (Mora-Urp, 1984). As implied

by the name, this variety is characterized by its yellowish

(gold color) fruits.

Guilielma speciosa Martius var. coccinea Barb.

Rodr., Enum. Palm. Nov. 23, 1875. Guilielma gasipaes

(Kunth) L. H. Bailey var. coccinea (Barb. Rodr.) L. H.


designated by Glassman (1972): Barb. Rodr., Sert. Palm.

Brasil. 2: t. 52c. 1902.

This variety has red fruits.

Guilielma speciosa Martius var. mitis Drude. in

Martius, Fl. Brasil.: Cyclanthaceae et Palmae I, Fasc.

13


14/154

85, vol. 3(2): 363. 1881. Type. Brazil. Rio de

Janeiro, cultivated, 26 Dec 1887, A. Glaziou 17342.

(Type: P, not seen)

This variety was proposed by Drude using as the main

character to segregate it the lack of spines in the stems.

It seems Drude was trying to complete the range of

varieties initially proposed by Barbosa Rodrigues, who

never described any spineless variety of pupunha. It is

worth noting that Glaziou collected the type specimen in

Rio de Janeiro from cultivated plants, probably from

material originally imported from the Amazon.

Guilielma speciosa Martius var. ochracea Barb. Rodr.

Vellosia 1, ed. 1: 40, 1888. Guilielma gasipaes

(Kunth) L. H. Bailey var. ochracea (Barb. Rodr.) L. H.


designated by Glassman (1972): Barb. Rodr., Sert. Palm.

Brasil. 2: t. 52d. 1902.

This variety was described as having a dark-yellowish

color.

Guilielma insignis Martius, Palm. Orbign., in

d'Orbigny, Voy. dans l'Amer. Mr. 7(3): 71-73. 1847.

Bactris insignis (Martius) Bailon, Hist. Pl. 13: 305.

14


15/154

1895. Type. Bolivia. Beni. Moxos, s.d., d'Orbigny 18

(Holotype at P, not seen)

Guilielma insignis was found by Alcides d'Orbigny in

several localities of Beni and Santa Cruz, in Bolivia.

Martius (1847) gave a detailed description for this

species, but the accompanying illustrations he presented

are not precise. The leaves in plates X and XXIX are drawn

as being in the same plane, a character not found in

Guilielma. The fruit, although showing the yellowish color

common to many of the cultivated pupunha, has an obovoid

shape, conflicting with his description "magnitude et

figura ovi gallinacei minoris."

Saldias (1991) collected and studied several specimens

of pupunha, either cultivated, semi-cultivated and growing

wild from Santa Cruz, Beni, Cochabamba, and La Paz, in

Bolivia, places that were partially visited by d'Orbigny.

He was able to identify B. gasipaes and B. dahlgreniana,

but not B. insignis (=G. insignis). From his data it is

possible to understand why he could not find G. insignis:

it has a fruit similar in size and shape to the smaller

cultivated pupunha that belongs to the microcarpa or

mesocarpa landraces proposed by Mora-Urp and Clement

(1989). Therefore, B. insignis is similar to B. gasipaes,

but its fruits are at the low range of size.

15


16/154

Guilielma utilis Oersted, Vidensk. Meddel. Kjoeb.

1858: 46, 1859. Bactris utilis (Oersted) Benth. &

Hook. ex Halm. in Goodman & Salvin Bio. Cent. Amer.

Bot. 3: 413. 1885. Type. Costa Rica, n.d., Oersted

s.n. (Type at C, not seen).

This was the only Central American taxon described as

a closely related species to B. gasipaes. However, since

it is known that pupunha is believe to have been introduced

to Costa Rica as early as 2,300 BC (Mora-Urp, 1994),

Oersteds species is in fact a cultivated pupunha

(mesocarpa landrace sensu Mora-Urp & Clement, 1989)

introduced to Central America. Therefore, this species is

a synonym of B. gasipaes Kunth.

The epithet utilis has been applied in different ways

by agronomists. Mora-Urp (1994) uses it to describe one

of the four western varieties of the pupunha mesocarpa

landrace. Clement (1995), on the other hand, lumped all

known pupunha ladrances into two subspecies of B. gasipaes

(?) naming one of them utilis. The other subspecies is

called speciosa and includes an undescribed and unpublished

variety of Karstens G. chontaduro Triana var. chichagui

known locally in Colombia as "chinamato."

Bactris caribaeaa Karsten, Linneae 28: 403. 1857.

Guilielma caribaeaa (Karsten) Wendland in Kerchove,

16


17/154

Palm. 246, 1878. Type. "Crescit ad pedem montis

glacialis St. Marthae in planitie sicca et calida

Maracaibensis et St. Marthensi." No date, Karsten

(?)(Type at LE?, not seen).

Dugand (1940) suggested Karstens description of the

type locality does not make geographic sense. He also

suggested, perhaps based on his field experience in that

region, that B. caribaea should be placed as a synonym of

G. macana (=B. macana). Clement (1988) decided to follow

Burret (1933-1934) and Glassman (1972) by recognizing both

species, detracting Dugands proposition of synonymy. It

is necessary, however, to make clear that Clement (1988)

apparently did not understood Karstens (1857) reasons for

describing a "true" Guilielma species under the name Bactris

as was the case for this taxon (see discussion of Guilielma

of Karsten).

Guilielma chontaduro Triana, Nuev. jen. i spec. Neo-

Gran. 15, 1854. Type. Colombia. "Entre Pedra de Moler

i Cartago" (Colombia). Type. Not designated. Guilielma

gasipaes (Kunth) L. H. Bailey var. chontaduro (Triana)

Dugand, Caldasia 1: 63, 1940.

Dugand (1976) gave additional data regarding the type

locality for this species, found in the Cauca River valley

and Caldas. Triana's fruit description says: "fructus

17


18/154

aurantiaci, flavi, velcorallini, magnitudine Guilielma

speciosa, pericarpo carnoso eduli crasso." Therefore, it

is possible to conclude that this is a small-fruited and

cultivated pupunha. Clement (1988) placed this species as

a synonymy of B. gasipaes. Mora-Urp (1994), on the other

hand and without further discussion, considered B.

chontaduro (=G. chontaduro?) as one of the species that

gave rise to the cultivated pupunha. Perhaps the common

name of Trianas species "chontaduro de los naturales" or

the high altitude at which it was found (1,200m) led

researchers to think about this possibility. It is worth

to note that B. chontaduro proposed by Mora-Urp is a name

found nowhere in the literature, and can not be used,

unless validly published.

Guilielma macana Martius, in d'Orbigny, Voy. dans

l'Amer. Mr. 7(3): 74. 1847. Bactris macana (Martius)

Pittier, Man. Plant. Usual. Venz., 276. 1926. Type.

Venezuela. Zulia, Maracaibo, n.d., Ple s.n. (Type at

P?, M?, not seen).

This species was described from the Maracaibo region

in Venezuela, which is close to the Colombia border.

Karsten (1857) described his B. caribaeaa from plants

collected in an unknown place near the same type locality

of G. macana (either at the Colombian or Venezuelan side of

the border). Although Karsten's species description is far

18


19/154

more complete than that of Martius, the data presented by

the later is within the range of Karsten's description,

making it clear that they had described similar species.

Pittier (1926) transferred G. macana to Bactris and

commented that it was a poorly known species. Perhaps the

lack of an illustration for Martiuss (and also Karsten's)

species contributed to this situation.

Guilielma matogrossensis Barb. Rodr., Palmae

Matogross. 33. 1898. Lectotype designated by

Glassman (1972): Barb. Rodr., Sert. Palm. Brasil.

2: t. 46b. 1903.

This species was found by Barbosa Rodrigues (1898)

without flowers and fruits, in the "virgin forests of the

morro do Capito-mr, at the margin of Rio da Casca,

tributary of Rio Manso, itself a tributary of Rio Cuiab,

in Serra da Chapada, Mato Grosso," located 70-90 kilometers

north of the city of Cuiab. "It is found deep in the

forest, usually solitary and far from each other, reaching

the forest canopy." From his description it is clear that

Barbosa Rodrigues definitely saw a species matching the

main vegetative characters of the Guilielma sensu Martius:

massive stem (10-12 m long and 12 cm in diameter), long

internodes (30 cm long at the base) fully covered by black

spines, plumose leaves, with sheath, petiole, and rachis

covered by a whitish tomentose layer. The fruits,

19


20/154

according to the information Barbosa Rodrigues obtained

from his field assistant, were red like "pitanga," globose,

and no more than 2 cm, a size much smaller when compared to

the cultivated B. gasipaes and even the small-fruited B.

insignis from Bolivia. Based on this information, Barbosa

Rodrigues suggested G. matogrossensis could be the ancestor

of the cultivated pupunha, arguing that the seeds may have

migrated through the Madeira and Tapajs Rivers to the

Amazonas River, and there evolved into the cultivated

pupunha.

When Barbosa Rodrigues renamed his G. matogrossensis

as B. coccinea (see discussion under B. riparia), it was

generally accepted that G. matogrossensis was really a

Bactris sensu strictu (Clement, 1988; Henderson, 1995).

However, my opinion is that G. matogrossensis is de facto a

wild pupunha from the western Amazon, similar to the later

published G. microcarpa Huber (=B. dahlgreniana), also

found in Rondonia, Brazil, by Clement et al. (1989) and

Santa Cruz, Bolivia, by Saldias (1991). A collection from

the type locality of G. matogrossensis is needed to clarify

its status as a true member of the wild pupunha species.

Guilielma microcarpa Huber, Bol. Mus. Goeldi 4: 476,

1904. Type. Brazil. Amazonas: Boca do Acre. Type

specimen not designated (Type at MG?, not seen).

20


21/154

Bactris dahlgreniana Glassman, A Revision of B. E.

Dahlgren's Index of American Palms, 150. 1972.

This species was found by Huber in several localities

near the town of Boca do Acre, Amazonas, Brazil, during his

trip to the upper Purus River in March-May of 1904. Huber

had already found, some years before, similar plants

(although sterile) in the vrzeas of the Chipurana River in

pampa Sacramento, between the Ucayali and Huallaga Rivers,

in Peru. He suggested that the geographic distribution of

this species included a large area, from the Huallaga and

upper Juru River to the Madeira River basin. More

recently Clement et al. (1989) extended this distribution

some 250 km eastward to the Madeira River, to the Ouro

Preto do Oeste region, in Rondnia, Brazil.

Huber (1904), discussed the differences between G.

matogrossensis and G. microcarpa, suggesting that they

could be synonyms. He was the first to propose a hybrid

origin for the cultivated pupunha, with G. microcarpa and

G. insignis as the parents.

Glassman (1972) transferred all Guilielma species to

Bactris. Not being able to use the binomial B. microcarpa

(already used by Spruce, 1871), he renamed G. microcarpa as

B. dhalgreniana.

21


22/154

The Guilielma species published by Karsten

Karsten (1857) proposed a new system of classification

for some genera in Bactridinae. He emended the description

of Guilielma Martius charact. emend. calyx et corolla

gampophyllus and interpreted it in a completely different

point of view, proposing that species belonging to this

genus should have sterile androecium structures in their

pistillate flowers, and the ones lacking such structures

should be members of Bactris. Karstens proposal would

result in the placement of all previously described

Guilielma species as members of Bactris and he did

published two "true" Guilielma species under the generic

name of Bactris: B. speciosa var. chichagui (a variety of

the G. speciosa Martius) and B. caribaea. Therefore, all

Guilielma species described by Karsten were never intended

to be part of Guilielma Martius, as it has been suggested

by several authors (Mora-Urp and Clement, 1981; Mora-Urp,

1984; Clement, 1988).

Karsten described under the name Guilielma three small

statured species he found in northern Colombia and

Venezuela: G. piritu, G. tenera and G. granatensis. These

species shared basically two common features: pistillate

flowers with sterile androecium structures, either as a jug

or staminoidal ring or as distinct staminodes, and black or

22


23/154

violaceous fruits, characters never found in the Guilielma

Martius.

Guilielma piritu Karsten, Linnaea 28: 397, 1857.

Bactris piritu (Karsten) Wendland in Kerchove, Palm.

234, 1878. Type area. "in regionibus calidis, planis,

argillosis, Venezuelae et Novo-Granada (Colombia)." No

type specimen designated.

The type locality given by Karsten is vague and a

precise geographic placement of the area referred it is

very difficult. Bactris piritu was considered by Wessels

Boer (1988) to be a synonym of B. guineensis, a proposal

accepted by Henderson et. al., (1995). B. guineensis it is

known today to be distributed in the llanos of northern

Venezuela and Colombia, and along the Pacific coast of

Central America. According to Karsten (cited by Wessels

Boer, 1988), this taxon is distinct from Bactris minor

Jacq. (=B. guineensis) by its pistillate flowers and stem

and petiole indument, being almost identical for the

remaining characters. Braun (1983, 1995) and Braun and

Chitty (1987) consider B. piritu and B. guineensis to be

distinct species, and their plates show B. piritu to be a

much smaller palm.

Guilielma tenera Karsten, Linnaea 28: 399, 1857.

Bactris tenera (Karsten) Wendland in Kerchove, Palm.

23


24/154

234, 1878. Type area. Colombia. Rio Magdalena. No type

specimen designated.

When Karsten published G. tenera (1857) he discussed

its relationship with B. brongniartii and B. maraja.

Wendland (in Kerchove, 1878) transferred G. tenera to

Bactris sensustrictu without discussing the reasons for

it, as well as to what group of species it was related.

Burret (1933-1934) included all previously described

species of Bactris with androecium structures in the calyx

and corolla of pistillate flowers in the genus Pyrenoglyphis

Karsten, making the combination P. tenera. Wessels Boer

(1965; 1988) and Galeano and Bernal (1987), however,

maintained B. tenera (=G. tenera) as synonym of B. maraja,

a conflicting proposal since B. maraja was originally

described by Martius without flowers and the fruit

description does not mention the presence of a staminoidal

ring. Recently, Henderson (1995) maintained that B.

brongniartii is a good species, distinct from B. maraja on

the basis of the pinnae shape (linear x sigmoidal) and the

presence of a staminoidal ring, more easily found in the

fruit perianth. As a result, B. tenera was placed as a

synonym of B. brongniartii.

Guilielma granatensis Karsten, Linnaea 28: 400, 1857.

Bactris granatensis (Karsten) Wendland in Kerchove,

Palm. 234, 1878. Type specimen. "Locis umbrosis,

24


25/154

humidis calidisque Venezuelae et Novo Granatae

[Colombia] habitans." Type specimen destroyed at

Vienna.

The description of the type locality for this species

is also vague and cannot be located precisely in Colombia

or Venezuela. It is considered by Henderson et. al. (1995)

a synonym of B. pilosa Karsten. B. pilosa is a member of a

large group of Bactris that have purple-black fruit lacking

androecium structures in the pistillate corolla. Since B.

granatensis Karsten was originally described as having

these structures, the present placement of this species

under the name B. pilosa does not seems to be an acceptable

decision.

The species of Sanders Antillean clade

In 1991 Sanders proposed a group of Bactris species

found in the Greater Antilles as members of a clade he

named Antillean. Salzman and Judd (1995) revised the

members of this clade and suggested it was formed by three

species, B. cubensis, B. jamaicana and and B. plumeriana,

and one ambiguous name. In both studies the Antillean

clade is considered a sister clade to the species of

pupunha (Guilielma clade) and together they form the Non-

Ocreate clade. The phylogenetic relationship of these two

25


26/154

clades is supported by several synapomorphies according to

Sanders (1991), but geographically they appear to have

evolved independently from each other.

Bactris cubensis Burret, Kungl. Svenska Akademiens

Handlingar 6(7): 25. 1929. Type. Cuba. Oriente:

Sierra de Nipe, in charrascalles ad Ro Piloto,

29 Jul 1914, Ekman 2286 (Holotype at S, not seen).

This species is restricted to the eastern part of

Cuba, where it usually grows in forested areas. According

to Salzman and Judd (1995), its fruits approach the form of

the small-fruited B. macana.

Bactris jamaicana L. H. Bailey, Gentes Herb. 4: 177.

1938. Lectotype designated by Salzman and Judd, 1995:

Jamaica. St. Catherine Parish: Gilbralter, on

Gibralter Road in pasture of Frank Roper State, 10 Nov

1935, Bailey 216 (Lectotype at BH, not seen).

Bactris jamaicana is restricted to Jamaica where it

grows in sub-montane forest and wet savannas. Salzman and

Judd (1995) considered it the most distinctive of the

Antillean Bactris species. The fruits of B. jamaicana that

I have examined at NY are very similar in shape to those of

B. riparia, and their endocarps have an unusually large

26


27/154

fertile pore, a character useful to identify the specimens

of this species.

Bactris plumeriana Martius, in d'Orbigny, Voy. dans

l'Amer. Mr. 7(3): 64. 1847. Type. Martinique or

Guadaloupe. Plumier, Tab. 43-45. Nova Plantarum

Americanarum Genera, Paris, Museum, Bibliotheque

Centrale.

Although Martius had expressed doubts if this species

was originally collected in two of the Lesser Antilles

islands, B. plumeriana seems now endemic only to the

Hispaniola. This species grows in forest or disturbed

areas.

Bactris chaetophylla Martius, in d'Orbigny, Voy. dans

l'Amer. Mr. 7(3): 71. 1847. Type. Dominican

Republic. Herb. Vetenatti nunc Webbii.

Martius described this species but placed question

marks on the name. It is probably a synonym of B.

plumeriana, according Salzman and Judd (1995).

Bactris setulosa and related taxa

27


28/154

Galeano and Bernal (1987) discussed the presence of

adnate and flattened fibers in an unidentified species

(Bactris sp. 2) they collected in Antioquia, Colombia

(Bernal, Galeano & Bolivar, 693). The presence of

flattened fibers adherent to the endocarp is one of the

apomorphies that defines the Guilielma clade proposed by

Sanders (1991). I had the opportunity to examine Bernals

collection at NY and it is, without any doubt, B. setulosa.

This species and closely related taxa are characterized by

their large size, similar to the cultivated pupunha, and

have an Andean distribution, where they can be found up to

1.700 m of elevation, in Ecuador (H. Balslev 4288).

Galeano and Bernal (1987) suggested their unknown species

(=B. setulosa) resembled slightly B. gasipaes and B.

riparia. Henderson et. al., (1995) suggested this species

as related to B. gasipaes.

Bactris setulosa Karsten, Linnaea 28: 408. 1857. Type.

Venezuela. Carabobo: Cumbre de Valencia, Puerto

Cabello, s.d., Karsten s.n. (Type at LE, not seen).

This species is very common at high elevations in the

Andes, but can also be found at low elevations in Colombia,

Ecuador, Venezuela, and Trinidad and Tobago. It has a

series of characters that make easy its identification:

pinnae very wide (sometimes up to 9cm), leaf spines

clustered, spines on the stem arranged in regular rings,

28


29/154

and fruits obovoid with the perianth usually entire.

According to A. Henderson (pers. com.) the broad pinnae of

B. setulosa is an uncommon feature in Bactris and, when

combined with the clustered spines on the leaves, are the

best characters to identify herbarium specimens of this

species. Two collections from Ecuador (Skov et al. 64.824;

Balslev & Steere 3121) and one from Surinam (Wessels Boer

506) have very distinct staminate flowers, with 9-12

stamens. In Wessels Boers samples seems to have occurred

a labeling mistake because the specimen collected in

Suriname by Wessels Boer (# 506) was supposedly taken from

forest camp, Coppenname Riverside, land floods with high

water, tidal influence, loamy soil, an environment

completely different from that cited for the two remaining

species collected in Ecuador, and unlikely to be typical

for a species belonging to the B. setulosa group.

Bactris cuvaro Karsten, Linnaea 28: 406. 1857. Type.

Colombia. Meta: Villavicencio. (Type at LE, not seen).

This species is poorly known, and was considered

synonym of B. setulosa by Henderson et. al. (1995).

Bactris cuesa Grisebach & H. Wendland in Grisebach,

Fl. British West Ind. Type. Trinidad. Maracas, s.d.,

Crueger s.n. (Holotype at K, not seen).

29


30/154

B. cuesa is considered a synonym of B. setulosa by

Wessels Boer (1988) and Henderson et. al. (1995). Bailey

(1947) presented illustrations of B. cuesa and B.

sworderiana, which were originally described without

illustrations. From Baileys description it is possible to

conclude they are closely related species. As a matter of

fact, the fruit illustrations are basically the same, with

a slight and insignificant difference in the shape.

Bactris falcata Johnston, Proc. Amer. Acad. Arts 40:

683. 1905. Type. Venezuela. Margarita Island: Ro

Assuncin, 16 Jul 1903, J. Johnston 220 (Holotype at

GH).

The Margarita Islands are adjacent to continental

Venezuela and it would not be unusual to find B. setulosa

growing there. The type specimen of B. falcata consists of

the apex of a young leaf and a small section of the stem,

similar to several specimens of B. setulosa at NY. In

Johnson's description it is clear that the spines on the

stem are arranged in regular rings and the fruits are

broader than long, typical features of B. setulosa.

Bactris sworderiana Beccari, Repe. Spec. Regni Veg.

16: 437. Type. Tobago. Caledonia, s.d., W. E.

Broadway 4014 (Holotype at FI, not seen).

30


31/154

Wessels Boer (1988) followed Burret (1933-1934) and

considered B. sworderiana and B. setulosa distinct species

on the basis of fruit shape (obovoid x depressed globose)

and the perianth size. All of the samples of plants

belonging to this group that I have examined at NY,

however, have obovoid fruits with or without a flattened

(depressed) apex. This small variation may have led

researchers to give the depressed-globose shape for the

fruits that present the flattened apex. Perianth size in

palm fruits are known to show some degree of variation when

they are not fully developed.

Beccari's description also has many of the typical

characters of B. setulosa. He suggested that B. sworderiana

has some differences form B. falcata, including broader

pinnae. This is irrelevant since in Johnson's description

of B. falcata a measure for the "segment intermediis" was

not given. Therefore Beccari's comparison was made with

the lower pinnae in Johnson's description, which are only 3

cm wide x 4 cm of the intermediate pinnae described by

Beccari. Henderson et. al., (1995), also considered this

species a synonym of B. setulosa.

Bactris kalbreyeri Burret, Repe. Spec. Regni Veg.

34: 231. 1934. Type. Colombia. Antioquia: Guadalito,

670-1000 m, 7 Feb 1880, W. Kalbreyer 1394 (Holotype

31


32/154

destroyed at B, Holoneotype at COL, designated by

Bernal et al., 1989, not seen).

This species is considered synonym of B. setulosa by

Henderson et. al., (1995). Burret (1933-1934) had

originally placed his own species as incertae sedis,

because he did not have fruits or flowers to make a

complete description. To make the situation worse, the

type collection of this species was destroyed in Berlin

during the Second World War and Burrets description does

not show any illustration for the species. In the original

description the number of pinnae on each side of the leaf

is only 23, a value far below the 40-65 accepted by

Henderson et. al., (1995) for B. setulosa. Bernal et al.

(1989) designated a Holoneotype for B. kalbreyeri and

admitted they failed to trace the original place of

collection in Guadalito or Guadualito. With so much

missing information, it would be better to consider this as

incertaesedidspecies, as Burret did.

Bactris circularis L. H. Bailey, Gentes Herb. 7: 388.

1947. Type. Trinidad. St. Andrews: Melajo, 26 Feb

1946, L. H. Bailey 158 (Holotype at BH, not seen).

This species was decribed and illustrated by Bailey in

his Palmae Indigenae Trinitensis et Tobagensis (1947).

Wessels Boer (1988) comments that Bailey's B. circularis is

32


33/154

a synonym of B. setulosa because the spine rings on the

stem can not be used to identify specific differences.

Bactris bergantina Steyermak, Fieldiana Bot. 28 (1):

71. 1951. Type. Venezuela. Anzategui: along Ro Leon,

northeast of Bergantin, 500m, 20 Feb 1945, J.

Steyermark 61039. (Holotype at F, not seen).

Steyermark (1951) comments that this species is

separated from the closely related B. cueso and B.

sworderiana by the flattened or nearly flattened spines on

the rachis of the petiole and the scales on the lower

surface of the pinnae. These are non-reliable characters

and should not be used to segregate species in Bactris. In

almost all other characters Steyermark's species is similar

to B. setulosa, including the clustered spines on the leaf

and the broad pinnae (4 cm wide). B. bergantina is

considered a synonym of B. setulosa by Wessels Boer (1988)

and Henderson et. al. (1995).

Bactris riparia

Like the species of B. setulosa, some herbarium

specimens of B. riparia have an endocarp with fiber

flattened and adnate to the endocarp, similar to those

found in pupunha. B. riparia and the species considered to

33


34/154

be synonym of it are of a large size, although usually

smaller than pupunha. The fruits are depressed-globose and

reddish. The crown of leaves is a feature that makes them

very similar to the wild and cultivated pupunha, as noticed

by Martius in the original description. Its geographic

distribution overlaps that of the cultivated and wild

pupunha in the lowlands of the Amazon basin. Henderson

(in prep.) propose it as close to B. gasipaes.

Bactris riparia Martius, Hist. Nat. Palm. 2: 97. 1826

Type. Brazil. Amazonas: Rio Japur, s.d., C. Martius

s.n. (Holotype, M, not seen).

Martius described this species without fruits and

noticed similarities with pupunha in the vegetative habit

"habitu fere frondium Guilielma speciosa." As implied by

its name, this is a species frequently found close to the

margin of lakes, Rivers, and small streams, having a

widespread distribution in the lowlands of the Amazon. In

Colombia (Dugand, 1940) it is known as pupua-brava.

Bactris inundata Martius, in d'Orbigny, Voy. dans

l'Amer. Mr. 7(3): 58. Type. Bolivia. Beni: Moxos,

s.d., A d'Orbigny 24 (Holotype at P, not seen).

Alcides d'Orbigny found this species in Beni, Bolivia,

and in the adjacent Mato Grosso, Brasil. The fruits were

34


35/154

described and an illustration provided (Pl. 27b, 6, 7 and

8). Herbarium specimens of plants collected in central

Amazon (the type locality of B. riparia) are similar to

those found in Bolivia and Mato Grosso, especially the

fruits, the habit, and the staminate flowers. B. inundata

is considered a synonym of B. riparia by Henderson (1995).

Bactris littoralis Barb. Rodr., Enum. Palm. Nov. 36.

1874. Type area. Brazil. Amazonas: cercanias de

Parintins. Lectotype (designated by Glassman, 1972):

Barb. Rodr., Sert. Palm. Brasil. 2: t. 32a and 33.

Barbosa Rodrigues species is similar to Martius B.

riparia not only in the vegetative and reproductive traits,

but also in the habit.

Bactris coccinea Barb. Rodr., Contr. Jard. Bot. Rio

de Janeiro 4: 110. 1907. Lectotype designated by

Glassman (1972): Barb. Rodr., Contr. Jard. Bot. Rio

de Janeiro 4: t. 24c.

When Barbosa Rodrigues described Guilielma

matogrossensis (1898) he did not see its fruits, but only

young and closed buds. He arranged to have the mature

fruits sent to him in Rio de Janeiro as soon as they became

available. However, seems it took almost eight years for

him to get the fruits. After he examined them, he notice

35


36/154

they were not of Guilielma type and decided to transfer his

G. matogrossensis from Guilielma to Bactris. Since he

could not use the binomial Bactris matogrossensis (he has

used already it for another species) he applied the name B.

coccinea. From Barbosa Rodrigues drawings it is possible

to conclude the fruits he examined in 1907 were similar of

B. riparia. This is corroborated by the fact that this

taxon can be found in Mato Grosso and in the adjacent

Bolivian territory, as reported by dOrbigny in the notes

that follow the description of B. inundata (=B. riparia).

Henderson (1995) consider B. coccinea and B. inundata to be

synonyms of B. riparia.

From Barbosa Rodrigues description this species is a

mix between the vegetative features of a Guilielma taxon

(G. matogrossensis) and the fruits of a Bactris (strictu

sensu) species.

36


37/154

CHAPTER 3

THE ANCESTORS AND PLACE OF DOMESTICATION OF PUPUNHA (Bactris

gasipaes Kunth)

There are disagreements about the origin of the

cultivated pupunha, the parent species, number of times it

was domesticated, and the place were it occurred. Some

authors suggest pupunha is of monophyletic origin (Barbosa

Rodrigues, 1898; Burret, 1933-1934; Clement, 1988;

Henderson, 1995), while a few others states it has a

polyphyletic origin (Huber, 1904; Mora-Urp, 1993). Mora-

Urp (1983) support the hypothesis of multiple

domestication events while Clement (1988) believe in a

single event followed by artificial dispersion. There are

37


38/154

also a dispute about the area were the domestication

process took place. For single or multiple events the most

frequently cited region includes a large area that follows

the Andes cordilleras, from northern to the Southeastern

South America.

Wallace (1853) had already observed "this palm appears

to be indigenous to the countries near the Andes, on the

Amazon and Rio Negro it is never found wild." Spruce

(1871), who traveled from the Amazon River mouth to the

Andes, comments when he asked local people in central

Amazon about the origin of pupunha, they pointed westward

and said "From the Cordilleras."

Barbosa Rodrigues (1898), who also worked in the

Amazon valley and had found only cultivated pupunha,

believed the wild pupunha could not be found in Brazil but

in some area near the Andes of Bolivia and Peru. He was

surprised when he found his G. matogrossensis near the city

of Cuiab, in an area relatively close to the type locality

of the small fruited and cultivated B. insignis, from the

adjacent Bolivia. Barbosa Rodrigues species was apparently

a primitive type of the cultivated pupunha, and he

hypothesized the seeds of G. matogrossensis could have

migrated throughout the Madeira and Tapajs River, reaching

the Amazonas River, where they acclimated, acquiring along

the time the features of the cultivated plants he had found

38


39/154

there. Barbosa Rodrigues G. matogrossensis was,

therefore, the first westward wild pupunha described.

I do not discard Barbosa Rodrigues hypothesis about

seed migration in the Tapajs River. The fact only

mesocarpa and macrocarpa landraces are found along the

middle course of the Solimes River (Mora-Urp and Clement,

1989), has put a question mark over the origin of the

microcarpa landrace found in Par, originally described by

Martius as G. speciosa. Mora-Urp (1984) suggested this

microcarpa landrace as being closely related to the

cultivated B. insignis Martius, from Bolvia, and believe

their seeds had been carried through the Mamor-Madeira

River (Bolvia) to the Amazon in Brazil.

Huber (1904) was the first to suggest the pupunha had

a hybrid origin, with G. microcarpa and G. insignis being

the parent species. This would explain the variations

already identified on the cultivated plants, such as fruit

size, color, mesocarp composition sometimes oily or

starchy, and the reduction or the spineless states found in

some plants. The most striking feature he discussed,

however, was the seed abortion (seedless fruits), perhaps

suggesting hybrids could not produce fertile descendent.

Huber did not accept the natives as being capable of

selecting plants because, as he states, they practiced a

primitive agriculture. He thought the appearance of

39


40/154

"accidental" hybrids in the dispersal zones of the parents

species, located someplace in the upper Purus, or Beni or

Mamor Rivers basin, gave the opportunity for the natives

to choose the improved ones for cultivation, and later

dispersal to the north and eastern South America.

Clement et al. (1989) studied the phenotypic variation

of two populations of B. dahlgrenian from Acre and

Rondonia, in the southwestern Amazon of Brazil, and

compared them with one population of the mesocarpa landrace

Pastaza from Ecuador and another possible B. dahlgreniana

population, from Peru. Their results suggest B.

dahlgreniana could be the ancestor of the cultivated

pupunha because this species is vegetatively similar to the

cultivated pupunha and differences are present only in the

vegetative parts. Although they have studied only the

fruits, a tendency towards the increment of infructescence

weight and a decrease of fruit number was identified. This

trend would be expected in the case pupunha was

domesticated from B. dahlgreniana by human selection.

Although Clement et al. (1989) do not state conclusively

that B. dahlgreniana is the ancestor of pupunha, their

results match Spruce's (1871) expectations about how the

wild pupunha would look like: "with fruits so much smaller

and drier than what it has become by long cultivation as to

be not easily recognizable."

40


41/154

Mora-Urp (1984) suggested different kinds of

cultivated pupunha had distinct origin. Some are simple

wild plants poorly improved, while others had hybrid

origin. He conclude saying pupunha does not have a common

center of origin since it is native to an extensive region

and the domestication process took place several times

along this area, hence, pupunha is a synthetic species.

Mora-Urp (1993, 1994) gives a list of the nine species he

believe are related to the multiple origin of pupunha.

Four of these species were validly described in the past as

B. ciliata, B. insignis, B. macana, and B. caribaea. Four

are undescribed species and one, B. chontaduro, is a name

not found in the literature. According to this author

pupunha evolved independently from each of these nine

species around their type locality. He comments the

ancestor for the macrocarpa landrace Vaups, which has the

biggest fruit among the cultivated pupunha, is still to be

described since all the wild pupunha found to date have

only small fruits.

To follow Mora-Urps logic one has to admit, e. g.,

that B. macana and B. caribaea must have been used by

natives in the Northern South America as a parent species

of some cultivated trait of pupunha. Considering they have

such small fruits (the smallest among the wild pupunha),

the expected cultivated variety derived from them must had

to had small fruits, and would be classified as a

41


42/154

microcarpa landrace type. However, Mora-Urp (1993) map of

distribution of the cultivated varieties of pupunha in

South America does not shows any cultivated landrace in

northern South America that overlaps the type locality of

the wild B. caribaea and B. macana, and this is a clear

indication that domestication of pupunha was not developed

in that area.

Several authors (Galeano and Bernal, 1987; Henderson

et. al., 1995; Henderson, 1995; Henderson, com. per.) have

suggested B. setulosa and B. riparia as new members of

the pupunha group on the basis of characters I already

discussed. These two species cannot be considered to have

a direct relationship to the cultivated pupunha, as are the

cases of the wild species, such as B. macana and B.

dahlgreniana, but may have been hybridizing with them,

wherever they grow allopatrically, as suggested by Clements

(1995). This author discussed phylogenetics relationship

in pupunha using Harlan's & Wet's (1971) gene pool concept.

He considers all cultivated pupunha as the primary gene

pool (GP-1), while the wild species and spontaneous

populations of pupunha are classified as the secondary gene

pool (GP-2). The GP-2 species are known to hybridize

freely with the cultivated B. gasipaes when found growing

close to it (Mora-Urp, 1984, 1994) All the remaining

species of Bactris are suggested as to represent the

tertiary gene pool (GP-3) and may hybridize naturally with

42


43/154

pupunha. I think only a partial number of Bactris species

could be considered as member of the GP-3 proposed by

Clement. This group would include only the orange,

yellowish or reddish fruited species that share common

features of their staminate and pistillate flowers with

pupunha species, such as non-spiny calyx and corolla, and

B. setulosa and B. riparia are members of this group.

SECTION II

THE PHYLOGENY OF PUPUNHA (Bactris gasipaes Kunth) AND

ALLIED SPECIES

CHAPTER 4

TAXONOMIC AND CLADISTICS STUDIES OF Bactris Jacq.

AND THE PLACEMENT OF PUPUNHA (Bactris gasipaes Kunth) AND

RELATED SPECIES.

43


44/154

The more recent taxonomic study of Bactris and related

genera was published by Burret (1933-1934). In this study

Guilielma Martius was kept segregated and the species of

Guilielma of Karsten, as well as all other Bactris with

androecium structures in the pistillate corolla, were

transferred to the genus Pyrenoglyphys Karsten. The

remaining species were dumped in the genus Bactris sensu

strictu, that was divided into two sub-genus and several

sections. He recognized seven species in Guilielma,

including G. utilis and G. insignis, species obviously

described from samples of cultivated populations of B.

gasipaes. Burret's treatment was a mix of ideas proposed

previously by several taxonomists and the result are

considered a compendium (Sanders, 1991), a catalog of names

(de Nevers et. al., 1996) or a synopsis (Salzman and Judd,

1995), in which 257 species are listed (38 of them

described as new by Burret), and no key or illustrations

were provided.

Using Burret's treatment as a starting point, Sanders

(1991) produced the first cladistic study of Bactris in

which the segregated genera Guilielma and Pyrenoglyphis

were lumped together to form a unique and large genus under

the name Bactris. He studied 139 herbarium specimens that

represented 61 species of Bactris sensustrictu, Guilielma

and Pyrenoglyphis. He scored 106 morphological characters,

although a large number of multistate characters were

44


45/154

broken down into separate binary characters. Sanders'

result shows that Bactris is composed of four "roughly

equivalent major clades" that do not correspond to any of

Burret's segregated genera. Regarding the Guilielma

species, Sanders used samples that represented only a

fraction of the species accepted by Burret (B. gasipaes, B.

dahlgreniana and B. macana), and they formed a small clade

he named Guilielma. This clade alone is equivalent to the

genus Guilielma Martius and in Sanders analysis it was

placed together with the Antillean clade to form one of the

major clades of Bactris, the Non-Ocreate clade.

From Sanders results (Fig. 5), the clade Guilielma

diverges from Bactris only as a member of the Non-Ocreate

clade and can not be segregated as the genus Guilielma

Martius alone without all the other equivalent clades being

equally considered. In this case, according to Sanders,

Bactris would be fragmented into eight or more poorly

defined genera.

Sanders was criticized because there is a difficulty

in interpreting the characters he used in his analysis (de

Nevers et. al., 1996). Hanh (1993) suggest the

shortcomings of Sanders study were due to the fact he has

"relied on original species descriptions for character

state assessment, redundancy of true characters,

multistates characters broken down into separate binary

45


46/154

characters, and considerable homoplasy in most characters."

From my point of view all these problems are related to the

fact he did not have the access to a good monograph of

Bactris and used a very small number of specimens while

searching for the most reliable characters, giving a

generalized impression he was unable to cover successfully

all the morphological variation present in Bactris. In

some cases he did not have the opportunity to examine

samples of common and widespread species, as was the case

of B. riparia and this may explain the fact the placement

of some species and even clades (Balonophora) in Sanders

analysis are presently disputed (de Nevers and Henderson,

1996). Besides these problems, Sanders results are of

good resolution regarding the monophyly of Bactris and this

has finished the dispute about the validity of Guilielma

as a good genus.

Salzman and Judd (1995) reviewed the species of the

Antillean clade proposed by Sanders and studied its

phylogenetics relationship with the sister clade Guilielma.

In their cladistic analysis they used morphological and

anatomical characters, but the data used to assess the

characters states of the Guilielma clade were partially

obtained from several bibliographical sources, while most

of the material used for the study of the Antillean species

were collected by them in the field. They also did not

specify the species members of the clade Guilielma they

46


47/154

used. The results (Fig. 06), confirmed Sanders hypothesis

about the relationship of the Guilielma and the Antillean

clade as being sister clades.

Henderson (1995) treated only the Bactris found in the

Amazon and did not recogniz Guilielma as a valid genus.

According to him there are only two pupunha species growing

in the Amazon basin. The domesticated B. gasipaes in which

he include as synonym several varieties, G. speciosa and G.

insignis, and the wild B. macana, with G. microcarpa under

it. The key he provided places B. setulosa in the same

group of pupunha on the basis of stem size.

CHAPTER 5

THE AGRONOMISTS SYSTEM OF CLASSIFICATION FOR THE

CULTIVATED AND WILD SPECIES OF PUPUNHA.

Mora-Urp and Clement (1989) have traveled extensively

in the Amazon basin, Andes and Central America, analyzing

different pupunha populations and collecting fruits for

germplasm banks. They have proposed (Clement, 1988; Mora-

Urp, 1994; Mora-Urp and Clement, 1989) a non-taxonomic,

but practical and efficient system to classify the

cultivated species of pupunha, including the wild plants

they believe are related to its domestication process.

47


48/154

Their system is mainly based on the fruit size. They have

divided all populations of cultivated and wild pupunha in

the occidental group, when found West of the Andes, and

oriental, when found at the East. So far they have

classified the cultivated and wild pupunha in three major

landraces (they do not use the term variety in agronomic

sensu), named microcarpa, mesocarpa and macrocarpa. They

have been studying several populations in the field and

each of them is characterized and classified into one of

the major landraces. There are 17 studied populations,

named as cultivated variety by Mora-Urp (1994) (Table I)

or landraces by Clement (1995).

Cultivated and wild populations of pupunha are known

to hybridize freely when in contact (Mora-Urp, 1984) and

it may be the explanation for the high diversity of the

species in character such as fruit color, size and content,

spines presence and plant size. Being now widely

cultivated in farmlands, and with an intense process of

seed exchange taking place, the occurrence of germplasm

contamination, landraces hybridization and genetic erosion

are the most likely results. A possible impact of this

process at long term would be the ruin of Mora-Urp and

Clement's system of classification.

48


49/154

CHAPTER 6

A CLADISTIC ANALYSIS OF PUPUNHA (Bactris gasipaes Kunth)

AND RELATED SPECIES: A MORPHOLOGICAL AND ANATOMICAL APPROACH

Introduction

This analysis was performed using morphological and

anatomical data and the new information is expected to

contribute to a better understanding of two important

aspects related to the taxonomy and history of pupunha: (a)

the phylogenetic relationship within the domesticated and

wild group of pupunha, and their relationship with the

49


50/154

species suggested to be closely related to them, and (b)

the possible place were the domestication event of the

cultivated species was developed. The hypothesis is that

cladistic analysis can be used to elucidate taxonomic

problems at species level and at the same time the results

of the phylogenetic reconstruction would allow a more

precise definition of the place were domestication events

occurred.

Material and Methods

A total of seven Bactris species were used in the

present analysis (Table II). The two Desmoncus and two

Astrocaryum species were chosen as the source of

information for building the outgroup data set. I decided

to follow Henderson (1995) and Henderson et. al. (1995)

taxonomic arrangement for the species of pupunha. The

cultivated species will be represented by B. gasipaes and

the wild species by B. macana. This means that all

herbarium specimens belonging to these species found at NY

will be sorted into two distinct piles and the further

anatomical and morphological surveys of each group will

show if there is any necessity to subdivide them before

completing the data matrix and running the cladistic

analysis. The same procedure was applied to B. setulosa

and B. riparia specimens. For the Antillean species I

followed the recent treatment of Salzman and Judd (1995).

50


51/154

The cladistic analysis was performed with an IBM-

compatible computer, using Wagner parsimony method to

analyze the characters. The programs Nona 1.1 (Goloboff,

1993) and Hennig86 1.5 (Farris, 1988) were used to generate

the trees. In Nona the commands Hold100, Hold/20 and

Mult*20 were used to search for the trees. The trees were

saved in Nona(sv* command), retrieved and generated in

Hennig86. Dada 1.0 (Nixon, 1995) and Clados 1.1 (Nixon,

1992) were used to examine and manipulate the data matrix

and the trees.

The outgroup comparison was used to asses the polarity

of the characters (Stevens, 1980; Watrous & Wheeler, 1981;

Madison et al., 1984). The outgroup used to assess the

polarity of the Bactris species analyzed here was

Desmoncus. In the cases when Desmoncus was too variable or

the character is considered to be an autapomorphy (stem

habit, modified leaves, e.g.), characters states were

obtained from Astrocaryum. Sanders (1991) and Salzman and

Judd (1995) discussed in more details the aspects of choice

of outgroup when cladistic analysis of Bactris were

performed.

The characters used in the analysis are mostly

qualitative and were searched among the outgroup and the

in-group taxa. Quantitative characters were used only when

51


52/154

a clear gap or discontinuity were detected. Multistate

character were considered to be unordered. Thirty four

morphological and anatomical characters were selected in

the analysis of the nine terminal taxa resulting from the

anatomical and morphological surveys (Table III). The

plesiomorphic or ancestral state is coded as 0 and the

apomorphic condition as 1, 2, or 3, in the cases of the

multi-states characters. The data matrix is presented in

the Table IV.

Only herbarium specimens were used during the present

study. One hundred and fifty six collections from the NY

herbarium were examined in a morphological and anatomical

survey of characters (Table V). Each survey involved the

use of a specific number of herbarium specimens, as well as

specific methods. The morphological survey used only the

usual and conventional methods and a short description of

it is given. On the other hand, the anatomical survey

required a longer time and the use of several techniques to

prepare multiple samples that were used in determining the

more reliable anatomical features that were incorporated as

characters state in the computerized analysis. A detailed

account of the anatomical survey is given for two reasons.

The first is to allow future replication of the methods

here employed. The second is derived from the difficulties

I had in collecting all the necessary methods and

protocols, which are scattered over several different

52


53/154

publications and not always were developed to be used

specifically with palm samples, requiring several

modifications.

Morphological Survey

Morphological observations were performed throughout a

careful examination of 156 herbarium specimens (table V).

A stereomicroscope Bausch & Lomb, with magnification of up

to 30x was used to study the comparative data for

vegetative, floral (staminate and pistillate) and fruiting

structures. All flowers and fruits were rehydrated prior

to their examination.

Anatomical survey.

Leaf clearing

The procedure recommended by Martens and Uhl (1994)

was mostly followed and minor changes were made (see

Appendix 01). The samples used are listed in Table V. All

procedures were done in small glass vials. Some samples

required a longer time in 5% NaOH (in a oven at 58C), and

in some cases up to one week. When the solution became

very dark it was changed for a new one, in most cases after

the first 24 hours. The samples were washed in deionized

water (three times, 15-30 min. each), and placed in one-

53


54/154

third strength commercial bleach, for a period of up to 30

min. After being washed again in deionized water, they

were stained in aqueous Safranin (0.5%) for a minimum

period of 30 min. or up to overnight. The samples were

washed and dehydrated for final mount using kleermount.

Flower Clearing

A total of 3-4 flowers (staminate and/or pistillate)

were cleared for each specimen sampled. The specimens

and flowers types cleared are listed in the Table V, and

were cleared according to the protocol described in the

Appendix 02.

The flowers were dissected only after the staining

procedure, immediately before the slide preparation. This

was necessary in order to avoid the loss of floral parts

during the process, that require many changes of solutions.

Some flowers with thick petals, such as B. gasipaes, B.

macana and B. setulosa, required a longer time in 5% NaOH

and resulted very difficult to handle due to the excessive

maceration. The tannin rich staminate flowers of B.

setulosa 2 (W. Boer 506, H. Balslev and Steere 3121 and F.

Skov et. al. 64.824) required to two changes of the NaOH

solution.

54


55/154

The long washing time required after the bleach

treatment(at least 30 min., step 7 of the protocol) was

necessary to ensure good water infiltration and complete

bleach removal. If the bleach is not completely removed

from the flower tissues, the staining procedure with

Safranin will not work properly.

Leaf Anatomy

Epidermal peels maceration, leaf clearings and cross

section were the techniques used in this part of my study.

The species and specimens sampled are listed in Table V.

Sampling

For the accuracy and uniformity of data collection,

only the middle portion of leaflets located along the

middle part of the leaves were sampled. Cross-section

samples included the midrib and one leaflet margin, and

measured 0.5-1 cm wide. Sample for epidermal peels were 1

cm and required one untrimmed leaflet margin. For

clearings, a 3 cm long section of leaflet with midrib and

one of the margins was used. In all cases one or two

additional spare sample was taken, rehydrated and kept in

55


56/154

Glycerin-Alcohol ready to be processed in case of material

loss.

The sampling procedure was done using a scissors and

most of the samples for cross section and epidermal peels

were trimmed at this point. However, if the material was

too brittle, cutting was difficult and usually resulted in

longitudinal fissures on the lamina and/or midrib

partition. To avoid this problem a larger piece of the

leaflet (4-5 cm long) was taken, and the samples trimmed

only after the rehydration procedure. When necessary, the

cross section samples taken from B. setulosa (which has the

widest leaflet) were subdivided in order to included the

midvein.

Rehydration and fixing

The rehydration solution used was that recommended by

Martens and Uhl (1984). Samples were rehydrated overnight

or up to 24 hours in a vacuum chamber (12 p.s.i.) and

finally washed in three changes of deionized water (15

minutes each). They were fixed in FAA (formalin, acetic

acid, 50% ethanol) for 48 hours.

Epidermal peels

56


57/154

The methodology of O'brien and MaCully (1981) was

chosen, following the recommendations of Pinheiro (1997).

The samples used are listed in Table V. Small glass vials

were used during most of the procedure (see Appendix 03).

The samples were placed in Jeffrey's solution, in half-

filled glass vials. The epidermis were easily separated

after 4-5 days and no additional cleaning with paintbrush

was necessary. Only one change of Jeffrey's solution was

made, usually after 24-48 hours. The vials were carefully

shaken once a day during the treatment period, since we

found it helpful for the epidermis and mesophyll

separation. To avoid the loss of samples during the

washing procedure a small glass pipette was used to

exchange the solutions. If the second Jeffrey's solution

was too dark, being impossible to see the samples, water

was added to dilute and clear it, making easy the use of

the pipette and diminishing the possibility of sucking up

the epidermis. The samples were moved to petri-dishes

after the staining step with 0.5% aqueous Safranin, and

processed there until the final slide mounting. Although

recommended, the use of paintbrush to move the sample from

the petri-dish to the slide is a very difficult task,

usually resulting in partial damage to the sample or even

total loss. To avoid it the better solution we found was

to fill up (to the top) a petri-dish with Hemo-De, dip

partially the slide and push the sample over it with the

use of a paintbrush. Some drops of Hemo-De on the slide

57


58/154

surface is helpful to position the epidermis in the desired

position on the slide surface.

Leaf cross section

The methodology described in Martens and Uhl (1984)

was modified and used to prepare the cross sections. The

samples used are listed in Table V. Samples were left 6-15

days in a 2:1 mixture of glycerin-alcohol 70% (1:6) and

Hidrofluoric Acid (HF). Due to its dangerous nature, HF

must be handled with extreme care under the hood and small

plastic vials must be used during this part of the

treatment. The length of time the samples can be treated in

HF without being damaged is quite variable. All species of

Bactris used in this study were left in the HF solution 6-9

days, and some, such as B. gasipaes and B. setulosa, 15

days without problems. Desmoncus (used as outgroup in the

cladistics analysis) was damaged after 6 days in the

solution. At the end of HF treatment the samples were

transferred to plastic vials (with small holes on the

sides) and washed 24 hours in running water to eliminate

all traces of HF. As it is known, even small amounts of HF

are enough to destroy the samples after they are mounted in

the slides.

The samples were transferred to small glass vials and

embedded in Paraplast Plus according to a protocol using a

58


59/154

Hemo-De series (Appendix 04). For each of the changes in

the Hemo-De series (up to the 100% Hemo-De step) the vials

were kept in a vacuum chamber (12 p.s.i.) for 1 hour to aid

the penetration of the solutions. Since Hemo-De is less

volatile than Toluene or Butyl Alcohol, two additional

changes of 100% melted paraplast, at 24 hour intervals,

were made in the oven. The resulting blocks were trimmed

(exposing the leaf edge) and soaked for additional

softening in glycerin-alcohol 70% (1:6) for 2-5 weeks.

This last step is of great importance for a successful leaf

sectioning. All samples sectioned without this further

softening were of very bad quality and not suitable for

mounting in slides. In the last series of cross section

preparation a 2:5 mixture of glycerin:alcohol 70% was used

as the further softening agent, also with good results.

The blocks were sectioned with an A. O. Spencer 820

rotary microtome, and the knives sharpened after use.

Sections of 15-20 were obtained and mounted in glass

slides. Problems of sections falling off the slides

reported by Pinheiro (1997) were not observed during this

study. Only Haupt adhesive was used, and a thin layer was

spread out with the finger on the slide surface. Abundant

Formalin 4% was placed over the adhesive and the section

transferred to the slide with the help of a scalpo. The

slides were transferred to a warming plate (43C) and after

the sections were stretched, the excess of Formalin was

59


60/154

eliminated using bulbilous paper. The slides were left up

to three hours in the warming plate and then transferred to

trays for an overnight period. To avoid the formation of

air bubbles in the sections it is important that the

warming plate does not overheat.

The sections were stained using a modified Safranin-

Astra Blue protocol (Appendix 05), developed by P. Endress

(Switzerland?). Although Martens and Uhl (1984) and

Pinheiro (1997) recommended the use of 0.5% Safranin-Fast

Green, we found it not satisfactory and expensive (it uses

absolute Alcohol to stop the fast green stain action).

Unlike Fast Green, Astra Blue does not stain the adhesive,

and therefore there it is no necessary to scrape the slides

with a razor blade, risking to damage the sections. Astra

Blue also gives a better time control of the stain

intensity, which is not the case when using Fast Green,

which requires a maximum of 20 seconds, sometimes less.

Several slides were overstained by Fast Green, and this

stain usually takes the place of Safranin, resulting in a

totally green section. The sections were mounted with

Kleermount and the slides were left on a warming plate for

2-3 days. A Nikon light microscope was used to observe and

take all the pictures presented in this study.

Results

60


61/154

The cladisitic analysis of the resulting nine terminal

taxa of Bactris generated a single most parsimonious tree

with 69 steps, a consistency index (C.I.) of 0.65 and a

retention index (R.I.) of 0.69 (Fig. 07).

Although I have started the characters survey giving

the name B. macana for all herbarium specimens of the wild

pupunha, the morphological and anatomical studies show two

distinct and isolated populations for this species, treated

separately in the analysis. The northern South America

population bears the name Macana1 in the data matrix and

tree, and the southwestern Amazonian population is named

Macana2. The reason for the splitting was the presence of

a high number of polymorphic characters which occured when

all specimens were grouped together as an isolated

terminal taxon. The splitting of B. setulosa was also

necessary and it was done because two very distinct

staminate flower types were found among the herbarium

specimens examined. These specimens also had a strong

variation in the arrangement of the fruit endocarp fibers.

The resulting two terminal taxa originating from the

splitting of B. setulosa were not associated with any

specific populations or geographic areas.

The monophyly of the Guilielma and the Antillean

clades of Sanders was confirmed, but they are not placed

together to form the Non-Ocreate clade as observed in

61


62/154

Sanders (1991), and Salzman and Judd (1995). In my study

the Antillean clade is the sister clade of all other taxa

analyzed, while the Guilielma Clade forms a more inclusive

clade grouped with B. riparia.

A new clade comprising B. setulosa plus B. riparia-

Guilielma is the sister clade of the Antillean clade. This

new clade is supported by two synapomorphies; the anther

with folded filaments and the lack of adaxial non vascular

fiber bundles, or if they are present, they do not show the

columnar or pentagonal shape typical of the taxa belonging

to the Antillean clade (Fig. 08, a, b).

The Guilielma clade is the best supported clade and

presents four synapomorphies, one of them, unique to this

clade, is the staminate petals with fibers heavily branched

apically (Fig. 09, a, b). The in-group arrangement in this

clade shows that B. gasipaes is the sister species to the

southwestern Amazonian population of B. macana (Macana 2).

These two species share two synapomorphies; the endocarp

irregularly oblong in lateral view and the fertile pore

displaced and located near the top of the endocarp (Fig.

10, a). B. gasipaes can be distinguished from the

remaining taxa of the Guilielma clade only by their fruit

with a characteristic ovoid shape (character 23) (Fig. 10,

b). The northern South America populations of B. macana

(Macana1) are defined by one reversal character state

62


63/154

(character 120) and the homoplasy represented by the

distinct palisade mesophyll, also found in the clade

comprising the Jamaican and Hispaniola taxa (Fig. 11, a,

b).

The clade B. riparia-Guilielma is supported by two

unique synapomorphies; pinnae with a pendulous apex and the

presence of buttresses touching the hypodermis in the small

veins (Fig. 12, a, b). The B. riparia clade is

differentiated by two apomorphies; the square-shaped

adaxial hypodermal cells and the ellipsoidal shape of its

primary veins.

The Setulosa clade is supported by two unique

synapomorphies; the spines on the stem forming distinct

rings close to the nodes and the trullate or obtrullate

shape of their primary veins (Fig. 13, a, b). Setulosa1 is

segregated by the presence of one apomorphy; the fruits

with an obovoid shape, a condition that can be partially

explained by the fact Setulosa2 was scored as missing (?)

for this character. All four homoplasia and the reversal

character 12 (120) that segregate Setulosa 2 from its

sister taxon are related to its specific staminate flowers.

The in-gro

Documents

HISTÓRIA TAXONÔMICA E FILOGENIA DA PUPUNHA