975
American Journal of Botany 85(7): 975–985. 1998.
ORIGINS OF DOMESTICATION AND POLYPLOIDY IN OCA
(OXALIS TUBEROSA: OXALIDACEAE): NRDNA ITS
DATA1
EVE EMSHWILLER2 AND JEFF J. DOYLE
L. H. Bailey Hortorium, Cornell University, Ithaca, New York 14853
As part of a study aimed at elucidating the origins of the octoploid tuber crop ‘‘oca,’’ Oxalis tuberosa, DNA sequences
of the internal trancribed spacer of nuclear ribosomal DNA (nrDNA ITS) were determined for oca and several wild Oxalis
species, mostly from Bolivia. Phylogenetic analysis of these data supports a group of these species as being close relatives
of oca, in agreement with morphology and cytology, but at odds with traditional infrageneric taxonomy. Variation in ITS
sequences within this group is quite low (0–7 substitutions in the entire ITS region), contrasting with the highly divergent
(unalignable in some cases) sequences within the genus overall. Some groups of morphologically differentiated species were
found to have identical sequences, notably a group that includes oca, wild populations of Oxalis that bear small tubers, and
several other clearly distinct species. The presence of a second, minor sequence type in at least some oca accessions suggests
a possible contribution from a second genome donor, also from within this same species group. ITS data lack sufficient
variation to elucidate the origins of oca precisely, but have identified a pool of candidate species and so can be used as a
tool to screen yet unsampled species for possible progenitors.
Key words: crop evolution; domestication; nrDNA ITS; oca; Oxalidaceae; Oxalis tuberosa; polyploidy.
Oxalis tuberosa (Molina), commonly known as ‘‘oca,’’
is one of over two dozen crops first domesticated in the
Andes (National Research Council, 1989). It is cultivated
at high altitude (2800–4100 m) in the central Andes, pri-
marily in small plots by traditional agriculturists, along
with other Andean tuber crops of unrelated families: Tro-
paeolum tuberosum R. & P. (Tropaeolaceae), Ullucus tu-
berosus Lozano (Basellaceae), and several tuber-bearing
Solanum species (Solanaceae). It is grown in greatest
abundance in the highlands of Ecuador, Peru, and Bolivia,
although it is found as far north as Venezuela and as far
south as Chiloe Island in Chile (National Research Coun-
cil, 1989). In recent decades it has become a commercial
crop in New Zealand (National Research Council, 1989)
and is also a common minor crop in the Transverse Neo-
volcanic Axis of Mexico, but the diversity of cultivars
there is very low, and several sources of evidence indi-
cate that it arrived in Mexico after the Spanish conquest
(King and Bastien, 1990). The crop has always been con-
sidered to be undoubtedly of Andean origin (Leo´n, 1967;
Bru¨cher, 1969, 1989). Although published reports have
consistently denied the existence of wild oca, or any re-
lated species bearing tubers (Leo´n, 1967; Hermann,
1 Manuscript received 13 January 1997; revision accepted 21 Novem-
ber 1997.
The authors thank the personnel of Programa de Investigacio´n de la
Papa (PROINPA), especially Maria Luisa Ugarte, Franz Terrazas, Andre
Devaux, and Sonia Echeverri, for collaboration in the collection of spec-
imens, providing germplasm accessions of cultivated oca, and essential
help with application for permission to export material of Oxalis from
Bolivia. The following also collaborated in transportation and collection
of specimens: Joseph Miller; Bente Eriksen and Ulf Molau; and Rudy
Vargas of SEMTA (Servicios Mu´ltiples de Tecnologı´a Apropriada). EE
thanks Jane L. Doyle for technical training. We thank David Spooner
and an anonymous reviewer for helpful comments on the manuscript.
Support is acknowleged from NSF grants DEB-9420215 (JD) and DEB-
9623227 (EE & JD), the Mellon Foundation Graduate Systematics
Training Program, the Harold E. Moore Fund, a Cornell Sage Graduate
School travel grant, and a Foreign Language and Area Studies Fellow-
ship from the U.S. Department of Education.
2 Author for correspondence.
1992), some wild populations of Oxalis bearing small
tubers have been found recently in Bolivia (see below).
The wild progenitor of domesticated oca is unknown,
as is the origin(s) of polyploidy (most reports [see below]
consider oca to be octoploid with 64 chromosomes). It is
unknown whether parental genomes were contributed in
single or multiple events and whether one or more eco-
types of a single progenitor species or more than one
well-differentiated species were involved. The occurrence
of meiotic abnormalities and frequent high levels of pol-
len sterility (Gibbs, Marshall, and Brunton, 1978) has
been cited as evidence of autopolyploidy, but cannot be
considered conclusive because no controlled crosses
demonstrating whether inheritance is polysomic or disom-
ic (e.g., Jackson and Casey, 1982) have been reported.
Taxonomic background—Oxalis is a large genus of
over 800 species, variable in habit and ecology, with
greatest diversity of species in South America and south-
ern Africa. The most recent monographic treatment of the
entire genus is that of Knuth (1930). However, it is gen-
erally acknowledged that the 38 sections defined by
Knuth are quite artificial, being based primarily on a few
vegetative characters, and his keys are considered nearly
unusable (Macbride, 1943; Salter, 1944; Bru¨cher, 1969;
Denton, 1973; A. Lourteig, Museum national d’Histoire
naturelle, Paris, personal communication), so this mono-
graph has not been very useful as a guide to which spe-
cies may be most closely related to oca. Knuth’s treat-
ment (e.g., 1930, and addenda 1931, 1935, 1936, 1940)
is the most comprehensive work on the genus, however.
Earlier monographs did not cover many South American
species, and their infrageneric classifications were clearly
artificial (e.g., Jacquin, 1794; DeCandolle, 1824; Zuccar-
ini, 1825, 1831). Subsequent published work has consid-
ered particular sections or geographical regions, but not
the whole genus (e.g., Salter, 1944; Eiten, 1963; Denton,
1973; Lourteig, 1975, 1979, 1981a, b, 1983, 1994). No
published revision has considered a group including O.

976 [Vol. 85AMERICAN JOURNAL OF BOTANY
TABLE 1. Accessions of Oxalis (including Xanthoxalis) sampled for ITS sequence. Collections of E. Emshwiller et al. except as noted below.
Collection
no.
Vouchera
no. Country, department, province, locality Species ID Sectionb 2n 5
Sequence
type
MHG861c EE133 Bolivia, Oruro, Avaroa, Challapata O. tuberosa Molina
(cultivated)
Ortgieseae Knuth 64i A
MHG884c EE136 Bolivia, Potosi, Tomas Frias, Chuitara O. tuberosa Molina
(cultivated)
Ortgieseae Knuth 64i A
MHG913c EE140 Bolivia, Cochabamba, Carrasco, Totora O. tuberosa Molina
(cultivated)
Ortgieseae Knuth 64i A (&E?)
EE260 EE260 Bolivia, Cochabamba, Ayopaya, near In-
dependencia
aff. O. tuberosa Molina
(wild)
Ortgieseae Knuth — A
EE284 EE284 Bolivia, Cochabamba, Chapare, Cande-
laria
aff. O. tuberosa Molina
(wild)
Ortgieseae Knuth — A
EE168 EE168 Bolivia, La Paz, Nor Yungas, Unduavi O. unduavensis (Rusby)
Knuth
Ortgieseae Knuth — A
EE291 EE291 Bolivia, La Paz, Nor Yungas, below
Chuspipata
O. unduavensis (Rusby)
Knuth
Ortgieseae Knuth — A
EE351 EE351 Bolivia, La Paz, Nor Yungas, Unduavi O. unduavensis (Rusby)
Knuth
Ortgieseae Knuth — A
EE308 EE308 Bolivia, La Paz, Nor Yungas, between
Chuspipata and Yolosa
O. sp. ‘‘O’’g — A
EE294 EE294 Bolivia, La Paz, Nor Yungas, between
Chuspipata and Yolosa
aff. O. distincta Knuth
(hybrid?)
Clematodes Knuth — A (&B?)
EE289 EB289 Bolivia, La Paz, Nor Yungas, between
Chuspipata and Yolosa
aff. O. distincta Knuth Clematodes Knuth — B
EE321 EE321 Bolivia, La Paz, Nor Yungas, between
Chuspipata and Yolosa
aff. O. distincta Knuth Clematodes Knuth — B
EE447 EE447 Bolivia, La Paz, Nor Yungas, between
Chuspipata and Yolosa
aff. O. distincta Knuth Clematodes Knuth — B
EE187 EE187 Bolivia, La Paz, Larecaja, Sorata O. spiralis R & P Ortgieseae Knuth 16j,k,l or 48i,m B
EE64 EE64 Bolivia, Cochabamba, Chapare, Locotal O. longissima (Kuntze)
Schumann
Ortgieseae Knuth — B
EE247 EE247 Bolivia, Cochabamba, Chapare, above
Locotal
O. longissima (Kuntze)
Schumann
Ortgieseae Knuth — B
EE418 EE418 Bolivia, La Paz, Murillo, Mecapaca O. peduncularis HBK Carnosae (Reiche)
Knuth
16i C
EE184 EE184 Bolivia, La Paz, Larecaja, Sorata aff. O. spiralis R & P
(not in flower)
Ortgieseae Knuth 16j,k,l or 48i,m C
EE249 EE249 Bolivia, Cochabamba, Chapare, above
Locotal
O. spiralis R & P Ortgieseae Knuth 16j,k,l or 48i,m C
EE250 EE250 Bolivia, Cochabamba, Chapare, above
Locotal
O. spiralis R & P Ortgieseae Knuth 16j,k,l or 48i,m C
EE190 EE190 Bolivia, La Paz, Larecaja, Sorata O. melilotoides Zucca-
rini
Clematodes Knuth 16i C
EE359 EE359 Bolivia, La Paz, Sud Yungas, between
Unduavi and Chulumani
O. mollissima (Rusby)
Knuth
Clematodes Knuth 16i C
EEx415 Eex415 Bolivia, La Paz, Murillo, Rı´o Zongo O. mollissima (Rusby)
Knuth
Clematodes Knuth 16i C
VULC1d EE124 unknown (El Salvador, Costa Rica, Pan-
amae)
O. vulcanicola Donn.
Smith
Ortgieseae Knuth — D
HERR1d EE459 unknown (‘‘Peru’’f) O. herrerae Knuth Herrerea Knuth 16i E
PED1d EE466 unknown (‘‘Ecuador, Peru’’f) O. peduncularis HBK Carnosae (Reiche)
Knuth
16i F
EE246 EE246 Bolivia, Cochabamba, Chapare, below
Colomi
X. flagellata Rusbyh Clematodes Knuth — G
EE331 EE331 Bolivia, La Paz, Sud Yungas, Ikiko X. flagellata Rusbyh Clematodes Knuth — G
EE345 EE345 Bolivia, La Paz, Murillo, Rinconada O. nubigena Walpers Capillares (Reiche)
Knuth
48–50n —o
EE315 EE315 Bolivia, La Paz, Nor Yungas, between
Chuspipata and Yolosa
O. andina Britton Clematodes Knuth — H
EE437 EE437 Bolivia, La Paz, Murillo, Rı´o Zongo O. yungasensis Rusby Corniculatae DC — H
EE350 EE350 Bolivia, La Paz, Sud Yungas, between
Unduavi and Chulumani
O. yungasensis Rusby Corniculatae DC — H
EE292 EE292 Bolivia, La Paz, Nor Yungas, between
Chuspipata and Yolosa
O. dolichopoda Diels Myriophyllum
Knuth
— I
EE219 EE219 Bolivia, La Paz, Nor Yungas, between
Cotapata and Chuspipata
O. sp. ‘‘R’’g — I
EE295 EE295 Bolivia, La Paz, Nor Yungas, below
Chuspipata
O. sp. ‘‘Z’’ (new spe-
cies?)
— J

July 1998] 977EMSHWILLER AND DOYLE—ORIGINS OF CULTIVATED OXALIS TUBEROSA
TABLE 1. Continued.
Collection
no.
Vouchera
no. Country, department, province, locality Species ID Sectionb 2n 5
Sequence
type
EE310 EE310 Bolivia, La Paz, Nor Yungas, between
Chuspipata and Yolosa
O. boliviana Britton Ortgieseae Knuth — K
ORT1d EE458 unknown (‘‘Andes of Peru’’f) O. ortgiesii Regel Ortgieseae Knuth 14p,q,r L
EE387 EE387 Bolivia, La Paz, Murillo, near La Paz O. pachyrrhiza Weddell Acetosellae DC — M
REG1d EE456 unknown (‘‘Peru, Brazil, Bolivia, Para-
guay, Argentina’’f)
O. regnellii Miquel Articulatae Knuth 28s N
EE235 EE235 Bolivia, Cochabamba, Tiraque, Toralapa O. latifolia HBK Ionoxalis (Small)
Knuth
14u, 28u, 42u O
EE445 EE445 Bolivia, Cochabamba, Chapare, between
Colomi and Locotal
O. martiana Zuccarini Ionoxalis (Small)
Knuth
56v P
EE366 EE366 Bolivia, La Paz, Ingavi, near Viacha O. bisfracta Turcz Clematodes Knuth — Q
PES1d EE125 unknown (‘‘S. Africa’’f) O. pes-caprae L. Cernuae Knuth 28m,r or 35t,q Ro
CRASS1d EE463 unknown (‘‘perhaps South America’’f) O. crassipes Urb. Articulatae Knuth — So
a Vouchers deposited at BH, with duplicates of wild Bolivian Oxalis deposited at LPB.
b Sections of Knuth (1930, 1935, 1936).
c Accessions of cultivated oca from Programa de la Investigacio´n de la Papa (PROINPA), Cochabamba, Bolivia.
d Plants purchased from Merry Gardens, Camden, ME.
e Native area as inferred from Smith, 1897; Lourteig, 1970, 1981b, and herbarium specimens at NY.
f Native areas as reported for these species in Hortus Third (Liberty Hyde Bailey Hortorium, 1976).
g New species proposed by A. Lourteig on specimen annotations, but not yet published (see Materials and Methods: Sampling).
h Combination as a species of Oxalis proposed by both Lourteig and Eiten on specimen annotations, but unpublished.
i de Azkue and Martı´nez, 1990. Other members of the ‘‘Oxalis tuberosa alliance’’ studied by de Azkue and Martı´nez but not sampled here (but
see Tosto and Hopp, 1996) are O. medicaginea, O. oblongiformis, O. subintegra, O. tabaconasensis, O. aff. villosula, and O. sp. (these with 2n 5
16), and O. lotoides (2n 5 32).
j Bru¨cher, 1969 (as O. pubescens H. B. K.).
k Favarger and Huynh, 1965.
l Huynh, 1965.
m Mathew, 1958 (as O. pubescens).
n Diers, 1961.
o Only partial sequences were determined for these plants (O. nubigena, ;470 bp; O. pes-caprae, ;300 bp; O. crassipes, ;210 bp).
p Heitz, 1927.
q Warburg, 1938.
r Marks, 1956.
s Naranjo et al., 1982.
t Franke, 1975
u Weller and Denton, 1975
v Hill, 1984.
tuberosa, although such work is in progress by Lourteig
(personal communication). The many conflicting deter-
minations of specimens in herbaria (E. Emshwiller, per-
sonal observations) indicate the need for basic work on
species delimitation. This lack makes identification of
specimens difficult, makes the identities of plants for
which there are published chromosome counts uncertain,
and has also complicated both sampling and intepretation
of results of this study (see below). Although progress
has been made in parts of the genus, Bru¨cher’s (1969)
comment that the systematics of genus Oxalis is still at
its beginning remains very much true today.
Cytological backgound—Base chromosome numbers
in Oxalis vary from x 5 5 to x 5 12, with x 5 7 most
frequent (Cronquist, 1981), and polyploidy is common in
the genus. Cytological work by de Azkue and Martı´nez
(1990) found a group of a dozen morphologically similar
Andean species that share a base chromosome number of
x 5 8, which is rare in Oxalis. This group includes oc-
toploid (2n 5 8x 5 64) O. tuberosa. There have been
conflicting reports of chromosome numbers in cultivated
oca (two reports of 2n 5 14, one of various euploid x 5
8 cytotypes [i.e., 2n 5 16, 24, 32, 48, 64], and others
from 2n 5 57–70: Heitz, 1927; Kostoff, Dogadkina, and
Tichonowa, 1935; Ca´rdenas and Hawkes, 1948; Gibbs,
Marshall, and Brunton, 1978; Talledo and Escobar, 1995;
Guama´n, 1997). However, over 100 accessions of culti-
vated oca have been found to have 2n 5 64 (Medina,
1994; Valladolid, Arbizu, and Talledo, 1994; Valladolid,
1996).
The species of the ‘‘O. tuberosa alliance’’ described
by de Azkue and Martı´nez (1990) (footnote i, Table 1)
belong to four of Knuth’s sections: Ortgieseae, Carnosae,
Clematodes, and Herrerea (Knuth, 1930, 1935, 1936).
With the exception of Herrerea these sections all include
not only the species with x 5 8, but other species with
published base numbers of x 5 5, 7, and 9 (Heitz, 1927;
Marks, 1956; Naranjo et al., 1982; de Azkue and Mar-
tı´nez, 1984). A few other species have published chro-
mosome counts with x 5 8 (Mathew, 1958; Bru¨cher,
1969; Federov, 1969). Most of these are also Andean
species that share some morphological similarities with
the alliance species, and all but one were placed by Knuth
into one of the same four sections (the very tiny O. nu-
bigena [2n 5 48–50: Diers, 1961] is a member of section
Capillares).
Earlier speculations by other workers concerning the
wild relatives of oca also have pointed to species of the
x 5 8 alliance. Bru¨cher (1969) wrote that oca’s progenitor

978 [Vol. 85AMERICAN JOURNAL OF BOTANY
was probably ‘‘a species similar to O. melilotoides,’’ an
x 5 8 species not included in the study of de Azkue and
Martı´nez (1990). Annotations by Eiten on specimens at
NY refer the x 5 8 species and other morphologically
similar species to either an ‘‘O. tuberosa complex’’ or an
‘‘O. scandens complex’’ (of unknown chromosome num-
ber). His only published reference to these species is a
brief mention of ‘‘a group of South American species
related to O. scandens,’’ distinguishing them from section
Corniculatae (Eiten, 1963, p. 290), and listing some of
the species (including Xanthoxalis flagellata, sampled in
this study; see below).
The x 5 8 alliance probably includes other species for
which cytological information is not yet available. Mac-
bride (1943) described several Peruvian named species as
being similar to (or perhaps conspecific with) some of
the alliance species. In addition to the morphologically
similar species mentioned by Macbride (1943) and Eiten
(1963), yet other taxa that resemble the x 5 8 species
were encountered in observations of herbarium speci-
mens and field collections (E. Emshwiller, unpublished
data), but these other species were not included in the
study by de Azkue and Martı´nez (1990), and do not yet
have published chromosome counts.
As part of a larger project investigating the systemat-
ics, genetic diversity, and ethnobotany of oca, we have
initiated molecular and morphological studies aimed at
establishing the identities of oca’s progenitor(s) and clar-
ifying the origins of polyploidy in the species. In this
paper we report on the application of nucleotide sequence
data from the internal transcribed spacer of nuclear ri-
bosomal DNA (nrDNA ITS; reviewed in Baldwin et al.,
1995) to the question of the relationships and origins of
Oxalis tuberosa.
MATERIALS AND METHODS
Sampling—DNA was isolated (Doyle and Doyle, 1990) from either
fresh leaves of greenhouse-grown plants, or leaves dried in silica gel in
the field (Chase and Hills, 1991) and kept at 48C after return to the
laboratory. DNA isolations were made from single individual plants in
all cases except EE345, O. nubigena. The plants sampled for ITS se-
quence (Table 1) included Oxalis from the following categories.
1) Cultivated O. tuberosa accessions kindly provided by the germplasm
bank of PROINPA (Cochabamba, Bolivia). Individuals of three ge-
notypes were selected that differed in tuber pigmentation and col-
lection locality.
2) Members of the x 5 8 alliance and species of unknown chromosome
number but resembling x 5 8 species. Most of these occur in the
cloud forest regions of the eastern slopes of the Andes, although
some are found in drier habitats at higher elevations.
3) Oxalis species that were found in similar Andean habitats (e.g.,
cloud forests) as the alliance species, but differed from that group
morphologically and lacked chromosome number data.
4) Species that belonged to the same four sections as alliance species.
5) Oxalis species that were found growing in areas where oca is cul-
tivated, even if they were not thought to be particularly closely re-
lated to oca (e.g., acaulescent bulbous species of section Ionoxalis,
weedy species of section Corniculatae).
6) Some divergent Oxalis species, from the Andes or elsewhere, in-
cluded to assess the levels of diversity in ITS sequence in the genus
and to serve as outgroups.
Determinations of some of the collections (Table 1) must still be
considered tentative, due to the taxonomic uncertainties in the genus
and the lack of good keys that cover these species. Identifications were
made using a combination of keys and descriptive information in Knuth
(e.g., 1930, 1935, 1936), Macbride (1943), and Eiten (1963), as well as
the original species descriptions, and plants were compared with her-
barium specimens, including types of some of the species.
In two cases the plants collected matched specimens at NY annotated
by Lourteig with yet-unpublished names (Table 1). Isotypes (Bang 315)
were examined of one of these new species, here designated as O. sp.
‘‘O.’’ The second species, here designated as O. sp. ‘‘R,’’ matches
Buchtien 620 p.p. (right-hand side; left side is type of Xanthoxalis un-
duavensis Rusby), annotated by Lourteig with a new specific epithet
honoring Rusby.
DNA amplification and sequencing—Amplifications were per-
formed in either 100 mL or 25 mL volumes. Amplification reactions
contained 20 mmol/L Tris-HCl, 50 mmol/L KCl, 2.5 mmol/L MgCl2,
0.2 mmol/L each dNTP, 0.6 mmol/L each of primers ITS4 and ITS5
(White et al., 1990), and 2.5 units Taq polymerase (Gibco BRL, Be-
thesda, MD, or Promega, Madison, WI) per 100 mL reaction. Ther-
mocycling profiles were set at an initial 1 min at 978C, followed by 39
cycles of 1 min at 978C, 1 min at 488C, 3 min at 728C, with a final
incubation of 7 min at 728C. Amplification products from one or more
reactions were then purified by cutting bands out of a 1% agarose gel,
with excess primers and dNTPs removed by Wizardy PCR Preps (Pro-
mega, Madison, WI) or QIAquicky Gel Extraction Kit (QIAGEN, Ba-
sel, Switzerland) columns.
Direct sequencing of double-stranded products followed the standard
Sequenaset version 2.0 (US Biochemical Corp., Cleveland, OH) pro-
tocols and labeling with 35S-dATP, with the following modifications as
recommended by Nickrent (1994): (1) a higher concentration of primer
(20 mmol/L); (2) dilution of the label mix by only 3.753 instead of 53;
and (3) modification of the annealing step to include subjecting the
primer/template mixture to 2 min at 998–1008C, followed by quick chill-
ing in an ethanol bath at 2808C for 2 min, adding the reaction buffer,
and incubating for 20 min at 378C. Sequencing reactions usually were
separated in two different acrylimide gels to maximize readable se-
quence: the first (6% acrylimide) for ;2 h, the second (4.5 to 6%
acrylimide) for ;6 h. Fuji x-ray film was exposed to the dried gels for
a minimum of 4 d.
Nucleotide sequences were determined for both strands, by use of
primers ITS2, ITS3, ITS4, and either ITS5 or ITS1 (White et al., 1990;
Fig. 1). In addition, another primer (‘‘ITS3b’’) was designed near the
39 end of the 5.8S coding region to facilitate reading the sequence of
the ITS-2 region. This primer anneals at a similar but not identical
position to primer ‘‘ITS3a’’ published by Downie and Katz-Downie
(1996) and has the sequence: 59-AGGGCACGYCTGCCTGGGTGTC-
39. Areas of compressions or other ambiguities presumably caused by
secondary structure could usually be resolved by comparison with the
complementary strand. The middle portion of the 5.8S gene was not
read in all taxa, and was invariant in all that were read, so it was not
included in the analysis. Downstream of this portion was a region that
was only read in the forward direction in most cases (the 39 end of the
5.8S and 59 end of the ITS-2), but this segment was free of ambiguities
and was the most conserved part of the ITS region, with few characters
contributed to the main analysis. Boundaries of ITS regions with the
coding sequences were determined by comparison with sequences in
Yokota et al. (1989).
Alignment and analysis—Sequences were aligned using Clustal V
(Higgins and Sharp, 1989) as implemented in the DNASTAR package,
using several different multiple sequence gap and gap length penalties.
Because of problems in alignment of the more divergent sequences (see
below), two separate sets of analyses were performed: (1) the primary
analyses of sequences that were readily alignable with those of the x 5
8 group (‘‘main analyses’’), using characters from the ITS-1, 5.8S, and

July 1998] 979EMSHWILLER AND DOYLE—ORIGINS OF CULTIVATED OXALIS TUBEROSA
Fig. 1. Internal transcribed spacer (ITS) region of nuclear ribosomal DNA, showing the anomolously amplified segment which appeared in
some amplification reactions of Oxalis DNA and indicating its position relative to the primers ITS1–5 (White et al., 1990) and the additional primer
used in this study. Asterisk indicates ITS4 primer in an unusual position and orientation.
ITS-2 regions; and (2) analyses of all sequences (‘‘outgroup analyses’’),
using only characters from the ITS-2 and 39 end of the 5.8S, the results
of which were used only for rooting the main analyses.
Each alignment was used in equally weighted parsimony analyses as
implemented in PAUP version 3.1.1 (Swofford, 1991). Branch and
Bound searches were conducted with MULPARS on and TBR branch
swapping, and included only one or two representatives of identical
sequences. Gaps were treated as missing data, and later mapped onto
the resulting most parsimonious trees.
RESULTS
Amplification products and sizes of ITS region—Am-
plifications of the ITS/5.8S region from Oxalis species
normally resulted in a single product of ;700 bp (base
pairs), typical in size for angiosperms (Baldwin et al.,
1995). In some amplifications a second product appeared
as well, ;60 bp larger than the first. When this second
amplification product was sequenced, it was found to be
identical to the smaller product but included an additional
segment at its 59 end (Fig. 1). Comparison of the longer
sequence with published sequences of the 18S nrRNA
genes of Daucus carota and Vicia faba (Yokota et al.,
1989) showed similarity in the first 40 bp upstream of
the ITS5 primer, with the 59-most 20 bp of the Oxalis
product being complementary to the ITS4 primer. Ap-
parently this larger amplification product is produced
when the ITS4 primer anneals in the forward direction,
upstream of the position of the ITS5 primer. Thus this
product is not, apparently, due to the presence of a di-
vergent paralogous locus in some taxa or to a contami-
nant in some samples.
The ITS-1 region was 200 bp in Oxalis tuberosa and
close relatives and varied from 198 to 202 bp in the spe-
cies included in the main analyses. Sizes of ITS-1 in out-
group taxa were more variable, up to 226 bp in O. mar-
tiana (exact sizes unknown for some species due to un-
sequenced segments). The ITS-2 region was 225 bp in
oca, varying from 213 to 226 bp in the other Oxalis sam-
pled. Sequences are deposited in GenBank as accessions
U74260–U74290.
In alignments that included sequences from all sam-
pled Oxalis species (the outgroup analyses), a core group
of very similar sequences (Table 1, sequence types A
through M) aligned nearly identically and with relatively
few gaps, regardless of alignment parameters used. This
core group was then used in the main analyses. Sequenc-
es from other species were difficult to align with those
of this core group (i.e., those of O. regnellii, O. latifolia,
O. martiana, O. bisfracta, and partial sequences from O.
pes-caprae and O. crassipes [Table 1, sequence types N
through S]). They were found to differ from the core
group in the position of a conserved sequence in ITS-1
(Liu and Schardl, 1994). Large (or many) insertion/de-
letion (indel) differences between the ITS-1 sequences of
these taxa and those of the x 5 8 species were suggested
by the differences in the length of the segments 59 and
39 of that conserved sequence. Length differences also
occurred for these same taxa in the region at the 39 end
of the ITS-2 designated ‘‘variable region 6’’ by Hersh-
kovitz and Zimmer (1996). Some sequence ambiguities
existed in other parts of the ITS-2 as well for these di-
vergent sequences, leading to differences among the re-
sults under different alignment parameters, especially
with respect to the most divergent sequence, that of O.
bisfracta. Sequences of these divergent species were not
included in the main analyses, and only portions of them
were used for the outgroup analyses, for the purpose of
rooting the closely related taxa (see ‘‘Phylogenetic anal-
ysis,’’ below).
ITS sequence types—When possible, sampling within
the x 5 8 group included several populations thought to
belong to the same species (Table 1). In some of these
cases accessions that were identified as the same species
had different sequence types (usually differing at a single
nucleotide position). If not artifactual, this could be due
either to intraspecific polymorphism or in some cases,
given the taxonomic uncertainties in Oxalis, to these ac-
cessions actually belonging to different species (see Dis-
cussion).
In other cases, plants identified as different species,
some of these quite distinct morphologically, had iden-
tical ITS sequences. A notable example is a group of
species that share the same ITS sequence with oca (Table
1). Among species identified as members of the x 5 8
alliance, sequences had no or very few differences across
the entire ITS region. Within the ‘‘x 5 8 clade’’ (see
below), sequences differed by a maximum of seven sub-
stitutions and no indels, while pairwise divergence be-
tween taxa included in the main analyses reached 35 sub-
stitutions and seven indels of from one to seven base
pairs.
All three cultivated oca accessions appeared at first to
have a single sequence type (that designated as ‘‘A’’ in

980 [Vol. 85AMERICAN JOURNAL OF BOTANY
Fig. 2. Phylogeny of Oxalis nrDNA ITS sequences. One of two equally most parsimonious trees found in the ‘‘main analyses,’’ which differs
from the alternative tree only by the optimization of character ‘‘d’’ (asterisk indicates the node that collapses in the alternative tree). Species in
boldface are classified as members of sect. Ortgieseae by Knuth (1930). Numbers in parentheses are collection numbers (see Table 1). Rectangles
are characters: black 5 nonhomoplasious changes; white 5 homoplasious characters; horizontal lines 5 insertions or deletions, not used as characters
in this analysis. Lowercase letters refer to particular characters (see text); uppercase letters denote ITS types. The more divergent ITS sequences of
O. bisfracta, O. crassipes, O. latifolia, O. martiana, O. pes-caprae, and O. regnellii were not unambiguously alignable with those of the taxa shown
above, so they were not included in the main analyses. EE294, a suspected hybrid that had both sequence types ‘‘A’’ and ‘‘B,’’ is not included in
the figure, nor is the incomplete sequence of EE345 (O. nubigena). The individual characters shown on the branch to O. pachyrrhiza may be either
synapomorphies of the ingroup or autapomorphies of O. pachyrrhiza.
Fig. 2), but on closer examination one of the accessions
showed faint secondary bands at the three positions
whose transformations are designated as characters ‘‘a,’’
‘‘b,’’ and ‘‘c’’ in Fig. 2. Assuming that only two sequence
types exist, (i.e., that no recombination has occurred), this
would indicate that oca has not only sequence type ‘‘A,’’
but also sequence type ‘‘E,’’ with the plesiomorphic nu-
cleotides at these three positons (see Sang, Crawford, and
Stuessy, 1995, for a similar case of ITS additivity in
Paeonia). Other than O. tuberosa, the only plant among
those included in the main analyses that had more than
one sequence type in an individual (EE294, Table 1) was
one suspected of being a hybrid a priori, on the basis of
its intermediate morphology. Among the more divergent
taxa, two polyploid species (O. pes-caprae and O. mar-
tiana) also had multiple sequence types within an indi-
vidual.
As indicated in Table 1, only a partial sequence was
obtained for Oxalis nubigena, so its sequence type is not
indicated. However, the regions that have been sequenced
are consistent with the placement of this species with the
other members of the x 5 8 alliance.
Phylogenetic analysis—Parsimony analyses of ITS-2
sequences (the outgroup analyses) nearly always identi-
fied the same topology for the subset of the sequences
that were of primary interest here and that were used
subsequently in the main analyses (see below) regardless
of alignment. Although the placements of more divergent
sequences varied, the O. pachyrrhiza sequence appeared
as sister to this core group of sequences in all but one
analysis (one among the several in which the highly di-
vergent and very difficult to align O. bisfracta sequence
was included). Therefore, O. pachyrrhiza was used as the
outgroup for the main analysis. This rooting also seems
to be in agreement with morphology, although no formal
cladistic analysis of morphological data yet exists for
these species.
Analyses of alignments produced using different pa-
rameters all resulted in the same two topologies for the
main analysis (Fig. 2). These differed only due to two
possible optimizations for one character (character ‘‘d’’
in Fig. 2). The Consistency Index (CI) and Retention In-
dex (RI) were calculated with all duplicate sequences re-
moved; the CI excluding uninformative characters was
0.870, and the RI was 0.914. Two gaps were potentially
informative in the main analyses, but were not used in
tree searches; neither these nor other gaps showed ho-
moplasy when mapped onto the most parsimonious trees
(Fig. 2).
DISCUSSION
Levels of ITS divergence and utility in Oxalis—Lev-
els of nucleotide substitution and length variation in ITS
sequence in the genus as a whole appear to be quite high,
such that it proved impossible to produce reliable align-
ments of sequences of some of the more divergent Oxalis
species with members of the x 5 8 group that were the
focus of this study. A comparison of the more divergent
sequences indicated that areas of similarity had many
small indel differences. In addition, the presence of other,
larger indels were suggested by (1) segments of sequence
that did not appear to have homologues in sequences of
other taxa, (2) the shifted position of a conserved se-
quence in ITS-1 (Liu and Schardl, 1994), and (3) differ-

July 1998] 981EMSHWILLER AND DOYLE—ORIGINS OF CULTIVATED OXALIS TUBEROSA
ences in length as well as sequence in the 39 portion of
ITS-2 designated as ‘‘variable region 6’’ by Hershkovitz
and Zimmer (1996). Characters from the more conserved
areas (the 5.8S and 59 region of ITS-2) could be aligned
and analyzed, suggesting that these portions of the se-
quence could be used in an analysis across the genus, or
perhaps even throughout the family Oxalidaceae. How-
ever, the level of variation in sequences of the complete
ITS region seems too high for practical use at these high-
er taxonomic levels (but see Hershkovitz and Lewis
[1996] for discussion of ‘‘deep-level’’ utility of ITS se-
quences).
Although these alignment problems precluded the use
of several divergent ITS sequences in the main analysis,
some of these sequences were quite similar to each other,
suggesting that ITS data could be used in future studies
to investigate phylogenetic relationships in other species
groups of Oxalis. For example, Oxalis native to southern
Africa have tunicate bulbs (e.g., O. pes-caprae), whereas
some species of the Americas have scaly bulbs (e.g., O.
martiana, O. latifolia), and some others have scaly rhi-
zomes (e.g., O. regnellii). The similarity and alignability
of ITS sequences of the few examples of these habits
included in this study suggest that further sampling could
test the speculations of Denton (1973) on the possibly
polyphyletic origin of the bulbous habit in Oxalis.
Xanthoxalis Small is one of several segregate genera
proposed by Small (1903, 1907) and adopted by some
other workers (e.g., Rose, 1906; Rusby, 1920; Holub,
1973). Some members of the x 5 8 alliance are included
in this segregate genus along with creeping weedy spe-
cies that are placed by others in Oxalis section Corni-
culatae (e.g., Knuth, 1930; Eiten, 1963; Lourteig, 1979).
Although they share some overall morphological similar-
ities, the very high levels of divergence in ITS sequence
between the members of the x 5 8 alliance and the weedy
species sampled so far (the Andean O. bisfracta and a
partial sequence of O. stricta L. for which data are not
presented here) suggest that they are not closely related.
Further sampling of ITS data could test these hypothe-
sized relationships.
ITS support of x 5 8 ‘‘Oxalis tuberosa’’ alliance vs.
traditional taxonomy—Phylogenetic analysis of ITS se-
quences supports the monophyly of the x 5 8 ‘‘Oxalis
tuberosa alliance’’ of de Azkue and Martı´nez (1990), in
that all plants identified as members of the alliance were
found within a single clade on the ITS tree (Fig. 2 and
Table 1, footnote i). Chromosome numbers of other spe-
cies whose ITS sequences fall within this clade have not
yet been determined, but all of these species are morpho-
logically similar to the x 5 8 group (see below), so it
seems likely that they also will be found to be based on
x 5 8, or at least to have been derived from ancestors
with this base number. On the other hand, the results of
the analysis are not consistent with Knuth’s (1930, 1936)
sectional classification, because sequences from members
of different sections are scattered across very different
parts of the tree, and the clade that includes the x 5 8
alliance also includes members of several other sections
(Fig. 2; Table 1). One notable example is that O. ortgiesii,
the type species of the section (Ortgieseae) into which
Knuth placed oca, not only has x 5 7 instead of x 5 8
(Heitz, 1927; Warburg, 1938; Marks, 1956), but its ITS
sequence is much less closely related to that of oca than
are those of several x 5 8 species (e.g., O. peduncularis,
O. herrerae, O. mollissima) that are placed by him in
different sections. However, with the exception of O. bo-
liviana (of unknown chromosome number), other mem-
bers of the section Ortgieseae that were sampled are sup-
ported by the ITS tree as closely related to oca.
Within the x 5 8 clade there is very little variation in
ITS sequence, yet the species are diverse morphological-
ly. Although they do share some features of floral and
inflorescence morphology, and are succulent, with large,
adnate stipules, the species vary in habit, indument, and
leaflet shape. They are especially variable in size, ranging
from a few centimeters tall (e.g., O. nubigena), to thick
sprawling vines of several meters (e.g., aff. O. distincta).
The morphological variation is partially due to high lev-
els of phenotypic plasticity, yet preliminary observations
from ‘‘common garden’’ greenhouse experiments indicate
that genetic polymorphism among different populations
is also important (E. Emshwiller, unpublished data). The
alliance species are ecologically diverse as well, being
found from tropical cloud forest levels in the eastern An-
dean foothills, through higher and drier environments, up
to almost the limits of vegetation at high altitudes.
Sister to the clade that includes sequences from the
species known to be based on x 5 8 is a clade of ITS
sequences of several other species collected in cloud for-
est regions on the eastern slopes of the Andes in Bolivia
(e.g., O. andina, O. dolichopoda). These species do not
have published chromosome counts, so it is unknown
whether they share a base number of x 5 8 with O. tu-
berosa and its allies. They share some morphological
similarities with the known alliance species (e.g., corolla
yellow with red veins, relatively large stipules), but differ
in lacking succulence, instead bearing leathery leaves on
thin, somewhat woody stems. Whether or not they share
the same base chromosome number, there is no evidence
from ITS data that they were involved in the origins of
oca (see below).
Origins of Oxalis tuberosa—Investigations of origins
of polyploidy and domestication are often complicated by
the number of different processes that may be at work to
one extent or another. In addition to the questions about
whether a polyploid arose from a single progenitor or
more than one species or population and whether it had
a single or multiple origin (reviewed in Soltis and Soltis,
1993) is the question of whether domestication occurred
before or after formation of the polyploid. The area of
domestication may have been localized or diffuse (Har-
lan, 1971), and predominantly vegetatively propagated
crops may still have some sexual recombination. Feral
escapes may be indistinguishable from possible progen-
itors, and crop–wild gene flow may add additional com-
plexity. Duplicated and diverged paralogous loci, which
in the case of ITS might even remain within a single
ribosomal DNA array, can confound the identification of
homeologous loci.
The study of the origins of domesticated O. tuberosa
is particularly challenging because of the rudimentary
state of Oxalis systematics. The unresolved problems of
species delimitation make it uncertain whether some local

982 [Vol. 85AMERICAN JOURNAL OF BOTANY
variants belong to separate species or ecotypes of a highly
polymorphic species. For example, O. spiralis (some-
times called O. pubescens H. B. K., an illegitimate name)
is a very widespread and morphologically variable spe-
cies complex for which both diploid and hexaploid cy-
totypes have been described (Mathew, 1958; Favarger
and Huynh, 1965; Huynh, 1965; Bru¨cher, 1969; de Azkue
and Martı´nez, 1990), and it is uncertain whether all the
plants identified here as O. spiralis are truly conspecific.
Species delimitation is also a problem with O. peduncu-
laris, in which case there have been various species rec-
ognized by Knuth and other taxonomists that may or may
not be distinct from this species. Regardless of the spe-
cies concept used, these problems may not be resolved
for many years, if ever. Our emphasis at this time is on
identifying populations that may have contributed to the
origins of oca, rather than on determining which popu-
lations are or are not conspecific.
The morphological diversity within the x 5 8 group
discussed above contrasts with the low levels of ITS se-
quence variability and the sharing of the same sequence
by different species. This sequence similarity may be due
to low levels of divergence or could indicate that inter-
specific gene flow is widespread. The latter possibility is
suggested by the existence of some individuals with mor-
phologies intermediate between species, which could be
hybrids (e.g., EE294, as well as seedlings that have ap-
peared in greenhouse collections). Crossing relationships
among species of the x 5 8 group are currently under
investigation (A. Valladolid, International Potato Center,
personal communication), yet it is well known that it can
be difficult to distinguish the extent to which hybridiza-
tion as opposed to primary divergence has affected the
pattern of variation among natural populations (reviews
in Heiser, 1973; Rieseberg and Wendel, 1993). Despite
the challenging situation, the ITS data have been able to
contribute to the understanding of oca’s origins.
The congruence of the x 5 8 alliance with a mono-
phyletic group identified by the phylogenetic analysis of
ITS sequences confirms the cytologically defined group
as a natural assemblage. The ITS data, therefore, can be
used to predict species likely to share this base chromo-
some number. Both sources of data support this group of
species as being the closest relatives of cultivated oca.
However, the ITS variation among species in this clade
is low, with only three informative characters, so that it
provides minimal resolution of relationships among the
sequence types found within the group. The primary se-
quence type ‘‘A’’ from oca is found within this group
(Fig. 2), as is the second sequence type ‘‘E’’ that ap-
peared faintly in one individual of oca. This indicates that
one, and possibly two, genome donors of cultivated O.
tuberosa may be found within this x 5 8 group. Alter-
natively, given only the ITS data, this faint sequence type
might be interpreted as gene flow from wild Oxalis, or
duplicated and diverged paralogous repeat types, also
from within the x 5 8 clade. In any case, there is no
evidence of origins of oca genomes from outside of this
group.
The two sampled populations of wild Oxalis with tu-
bers had the same sequence type as O. tuberosa (‘‘A,’’
the primary sequence in oca). Although neither popula-
tion showed any signs of the second sequence type, this
second type was not always observed in oca either, so its
absence cannot be taken as definitive. Wild populations
with tubers could be truly wild or could merely be es-
capes from cultivation. Both situations appear to occur in
our sample, based on information from stylar morphol-
ogy. Oca is cultivated as a clonal crop, and each cultivar
includes only one of the three possible floral morphs of
this tristylous species. Although oca can produce viable
seed (Alandia, 1967; Gibbs, 1976; Vallenas, 1992; Car-
rio´n, 1995), seedlings are rarely observed in field con-
ditions (Hill, 1939; E. Emshwiller, personal observa-
tions). A population formed by an escape from cultiva-
tion might be founded by a single tuber, and if so would
be expected to include only one floral morph. In some
populations (e.g., EE284), plants bore relatively large tu-
bers and the population was nearly monomorphic for the
mid-styled morph. This suggests that plants of this pop-
ulation are propagating clonally and probably were es-
capes from cultivation. In contrast, all three floral morphs
were present in plants of the wild population with small
tubers (e.g., EE260), indicating that these were probably
reproducing by seed, and thus were possibly truly wild.
Sequence type ‘‘A’’ was also shared by other wild Ox-
alis: three plants (EE168, 291, and 351) identified as O.
unduavensis, as well as EE308, which matches specimens
annotated by Lourteig with a yet-unpublished name (here
designated O. sp. ‘‘O’’). O. unduavensis is similar enough
to oca morphologically that it is considered by Lourteig
(unpublished annotations of herbarium specimens, e.g.,
Buchtien 621 and 3250) to be a variety of Oxalis tu-
berosa, although it lacks tubers and differs in the form
of the inflorescence. O. sp. ‘‘O,’’ however, is clearly a
distinct species from oca, differing in height, indument,
inflorescence morphology, and flowering phenology. The
fact that the same ITS sequence is shared not only by
cultivated oca and the wild populations with tubers, but
also by O. unduavensis and especially O. sp. ‘‘O,’’ sug-
gests that the nrDNA ITS is not variable enough to fulfill
the goals of identifying with exactitude the progenitor or
progenitors of oca and of indicating how many times the
polyploid arose. The ITS data do, however, support some
of the x 5 8 species as being more likely candidates than
others to have contributed to the polyploid genome of
Oxalis tuberosa.
The second sequence type ‘‘E’’ in oca was inferred by
assuming that no recombination had occurred among the
three sites that showed faint signs of a second nucleotide.
No recombinational types were observed among the three
characters (‘‘a,’’ ‘‘b,’’ and ‘‘c,’’ Fig. 2) in the sequences
sampled in this study. However, the ITS sequence for O.
tuberosa submitted to Genbank by D. S. Tosto and E. H.
Hopp (accession Z66546; Tosto and Hopp, 1996) differs
from sequence type ‘‘A’’ at three sites, one of which is
character ‘‘b,’’ where they report the plesiomorphic state.
Thus this sequence could represent a recombined type;
alternatively, it could be a case of simple homoplasy.
When this sequence is included in the cladistic analysis,
it causes the collapse of the branches supported by char-
acters ‘‘a’’ and ‘‘b’’ (Fig. 2).
The sequence type ‘‘E’’ in unrecombined form was
found in O. herrerae in this study and is also identical
to the single ITS sequence type published by Tosto and
Hopp (1996; Genbank accession Z66547) for O. villosula

July 1998] 983EMSHWILLER AND DOYLE—ORIGINS OF CULTIVATED OXALIS TUBEROSA
Knuth, O. tabaconasensis Knuth, O. oblongiformis
Knuth, and O. peduncularis. This sequence type also ap-
pears on the ITS tree within the x 5 8 group, indicating
that a second progenitor of cultivated oca might be found
in a different part of the ‘‘x 5 8 clade’’ (Fig. 2). This
placement is consistent with our preliminary results from
sequences of the chloroplast-expressed isozyme of glu-
tamine synthetase (ncp-GS), a single-copy nuclear-encod-
ed locus. More than one sequence type of ncp-GS has
been found within individual plants of oca, perhaps rep-
resenting homeologous loci of an allopolyploid, each
contributed by a different member of the x 5 8 group (E.
Emshwiller and J. J. Doyle, unpublished data).
The demonstration of interlocus concerted evolution in
allopolyploid Gossypium (Wendel, Schnabel, and Seel-
man, 1995) suggests one possible explanation for the in-
consistent and faint appearance of the second sequence
type in O. tuberosa. This process has the potential to
‘‘erase’’ one of the parental sequence types in an allo-
polyploid. If concerted evolution has acted to homoge-
nize sequence types across loci, but has not yet done so
completely, there may still be some ITS repeats remain-
ing of the alternative type. In this case the parental repeat
types would no longer be present in numbers reflecting
the proportion of genomes contributed by each parent.
Alternatively, repeat number in the highly variable and
dynamic nrDNA may differ among sequence types even
when concerted evolution does not act across loci. The
fact that only one sequence type was detected in some
oca sequencing reactions suggests that the number of re-
peats of the second type, if this repeat type is present in
these individuals at all, may be much lower than that of
the primary type. It is possible that ‘‘PCR drift’’ (Wagner
et al., 1993) may be favoring the repeat type that is al-
ready present in greater numbers, further enhancing its
levels over the second type. This could be true whether
the original difference in numbers of the two sequence
types were due to interlocus concerted evolution or sim-
ply to fewer repeats at the second locus. Baldwin et al.
(1995) also report instances in which sequence types
found in cloned DNA were absent in direct sequencing
of PCR products from the same individual. With the cur-
rent data it is not possible to distinguish the extent to
which the weak presence of the second sequence type
represents its actual proportion or is an artifact of PCR.
Thus, within a genus with high levels of divergence in
ITS sequence overall, ITS data support a group of spe-
cies, including Oxalis tuberosa, as a natural lineage, with-
in which the levels of divergence are quite low. This
group is congruent with the x 5 8 ‘‘Oxalis tuberosa al-
liance’’ of de Azkue and Martı´nez (1990) and with mor-
phology, but includes additional species not yet charac-
terized cytologically. The sectional classification of
Knuth (e.g., 1930, 1935, 1936), however, is not supported
by these data. More importantly, ITS sequence data iden-
tify a pool of species closely related to oca, from any of
which one or more of its genomes may have been de-
rived. Other sources of data will be necessary to deter-
mine more precisely the progenitor(s) of octoploid O.
tuberosa and the number of times it arose, both because
of the low variability of ITS sequences in this group, and
because of the possibility that some parental ITS types
may not be observed. However, the ITS data have already
eliminated some taxa from consideration as progenitor
candidates in the absence of cytological data, and they
have distinguished among some members of the x 5 8
alliance those more likely to have been genome donors
in the formation of the octoploid. Thus, these data can
continue to be used to screen yet unsampled Oxalis spe-
cies from other areas of the Andes for other possible pro-
genitors of oca.
LITERATURE CITED
ALANDIA B., S. 1967. Produccio´n de semilla sexual en oca. Sayan˜a
(Bolivia) 2: 12–15.
BALDWIN, B. G., M. J. SANDERSON, J. M. PORTER, M. F. WOJCIECHOWSKI,
C. S. CAMPBELL, AND M. J. DONOGHUE. 1995. The ITS region of
nuclear ribosomal DNA: a valuable source of evidence on angio-
sperm phylogeny. Annals of the Missouri Botanical Garden 82:
247–277.
BRU¨ CHER, H. 1969. Poliploidı´a en especies sudamericanas de Oxalis.
Boletı´n de la Sociedad Venozolana de Cie´ncias Naturales 28(115/
116): 145–178.
. 1989. Useful plants of Neotropical origin and their wild rela-
tives. Springer-Verlag, Berlin.
CA´ RDENAS, M., AND J. G. HAWKES. 1948. Numero de cromosomas de
algunas plantas nativas cultivadas por los indios en los Andes. Re-
vista de Agricultura Bolivia (Cochabamba) 5(4): 30–32.
CARRIO´ N P., S. 1995. Biologı´a reproductiva de la oca (Oxalis tuberosa
Molina) y sus implicaciones para la produccio´n de semilla botanica.
Tesis de Licenciada en Ciencias Biolo´gicas, Pontifica Universidad
Cato´lica del Ecuador, Quito, Ecuador.
CHASE, M. W., AND H. H. HILLS. 1991. Silica gel: an ideal material for
field preservation of leaf samples for DNA studies. Taxon 40: 215–
220.
CRONQUIST, A. 1981. An integrated system of classification of flowering
plants. Colombia University Press, New York, NY.
DE AZKUE, D., AND A. MARTı´NEZ. 1984. Variacion del cariotipo, volu-
men nuclear y contenido de ADN en siete especies de Oxalis. Dar-
winiana 25(1–4): 267–277.
, AND . 1990. Chromosome number of the Oxalis tuber-
osa alliance (Oxalidaceae). Plant Systematics and Evolution 169:
25–29.
DE CANDOLLE, A. P. 1824. Oxalideae. Prodromus Systematis naturalis
1: 689–702.
DENTON, M. F. 1973. A Monograph of Oxalis, section Ionoxalis (Ox-
alidaceae) in North America. Publications of the Michigan State
University Museum, Biological Series 4: 455–615.
DIERS, L. 1961. Der Anteil an Polyploiden in den Vegetatsgu¨rteln der
Westkordillere Perus. Zeitschrift fu¨r Botanik 49: 437–488.
DOYLE, J. J., AND J. L. DOYLE. 1990. Isolation of plant DNA from fresh
tissue. Focus (Gibco BRL) 12: 13–15.
DOWNIE, S. R., AND D. S. KATZ-DOWNIE. 1996. A molecular phylogeny
of Apiaceae subfamily Apioideae: evidence from nuclear ribosomal
DNA internal transcribed spacer sequences. American Journal of
Botany 83: 234–251.
EITEN, G. 1963. Taxonomy and regional variation of Oxalis sect. Cor-
niculatae 1. Introduction, keys and synopsis of the species. Amer-
ican Midland Naturalist 69: 257–309.
FAVARGER, C., AND K. HUYNH. 1965. Chromosome number reports IV.
Taxon 14: 86–92.
FEDEROV, A. A. [ed.]. 1969. Chromosome numbers of flowering plants.
Academy of Sciences, Leningrad (English translation, 1974).
FRANKE, R. 1975. U¨ ber das Auftreten von unreduzierten Ganeten bei
Angiospermen. Arch. Zu¨chtungsforsch, Berlin 5: 201–202.
GIBBS, P. E. 1976. Studies on the breeding system of Oxalis tuberosa
Mol. Flora, Bd. 165, S. 129–138.
, D. MARSHALL, AND D. BRUNTON. 1978. Studies on the cytology
of Oxalis tuberosa and Tropaeolum tuberosum. Edinburgh Royal
Botanical Garden-Notes 37: 215–220.
GUAMA´ N C., S. 1997. Conservacio´n in situ caracterizacio´n y evaluacio´n
de la biodiversidad de oca (Oxalis tuberosa) y papalisa (Ullucus
tuberosus) en Candelaria (Chapare) y Pocanche (Ayopaya). Tesis

984 [Vol. 85AMERICAN JOURNAL OF BOTANY
de Ingeniero Agro´nomo, Universidad Mayor de San Simo´n, Co-
chabamba, Bolivia.
HARLAN, J. R. 1971. Agricultural origins: centers and noncenters. Sci-
ence 174: 468–473.
HEITZ, E. 1927. Ueber multiple und aberrante Chromosomenzahlen.
Ablandl Naturwississenschaften Vereins Hamburg 21: 47–57.
HEISER, C. B. 1973. Introgression re-examined. Botanical Review 39:
347–326.
HERMANN, M. 1992. Raı´ces y tube´rculos andinos: prioridades de inves-
tigacio´n para un recurso alimentario pospuesto. Centro Internacion-
al de la Papa, Lima.
HERSHKOVITZ, M. A., AND L. A. LEWIS. 1996. Deep-level utility of the
rDNA-ITS region. Molecular Biology and Evolution 13: 1276–
1295.
, AND E. A. ZIMMER. 1996. Conservation patterns in angiosperm
rDNA ITS2 sequences. Nucleic Acids Research 24: 2857–2867.
HIGGINS, D. G., AND P. M. SHARP. 1989. Fast and sensitive multiple
sequence alignments on a microcomputer. CABIOS 5: 151–153.
HILL, A. W. 1939. The oca and its varieties. Royal Botanic Gardens,
Kew, bulletin of miscellaneous information 4: 169–173.
HILL, L. M. 1984. Chromosome number reports LXXXIV, Taxon 33:
536–539.
HOLUB, J. 1973. New names in Phanerogamae 2. Folia Geobotanica et
Phytotaxonomica 8: 155–179.
HUYNH, K. L. 1965. Contribution a` l’e´tude caryologique et embryolo-
gique des Phane´rogames du Pe´rou. Denkschriften der Schweizer-
ischen Naturforshenden Gesellschaft. Me´moires de la Socie´te´ Hel-
ve´tique des Sciences Naturelles 85: 1–178.
JACKSON, R. C., AND J. CASEY. 1982. Cytogenetic analyses of auto-
polyploids: models and methods for triploids to octoploids. Amer-
ican Journal of Botany 69: 487–501.
JACQUIN, N. J. 1794. Oxalis monographia, iconibus illustrata. 1–120.
Vienna.
KING, S. R., AND H. H. C. BASTIEN. 1990. Oxalis tuberosa Mol. (Ox-
alidaceae) in Mexico: an Andean tuber crop in Meso-America. Ad-
vances in Economic Botany 8: 77–91.
KNUTH, R. 1930. Oxalidaceae. In A. Engler [ed.], Das Pflanzenreich.
IV 130 (Heft 95) Leipzig.
. 1931. Oxalidaceae novae, post editionem monographiae meae
(a. 1930) detectae. Repertorium Specierum Novarum Regni Vege-
tabilis 29: 213–219.
. 1935. Oxalidaceae 2. Repertorium Specierum Novarum Regni
Vegetabilis 38: 194–199.
. 1936. Oxalidaceae 3. Repertorium Specierum Novarum Regni
Vegetabilis 40: 289–293.
. 1940. Oxalidaceae 4. Repertorium Specierum Novarum Regni
Vegetabilis 48: 1–4.
KOSTOFF, D., H. DOGADKINA, AND A. TICHONOWA. 1935. Doklady Aca-
demiia Nauk URSS 3(8): 401 (cited in C. D. Darlington and A. P.
Wylie, 1955, Chromosome atlas of flowering plants. London).
LEO´ N, J. 1967. Andean tuber and root crops: origin and variability. In
E. A. Tai, W. B. Charles, E. F. Iton, P. H. Haynes, and K. A. Leslie
[eds.], Proceedings of the International Symposium on Tropical
Root Crops held at University of the West Indies, vol. 1, 118–123.
Department of Crop Science, University of the West Indies, St.
Augustine, Trinidad.
LIBERTY HYDE BAILEY HORTORIUM. 1976. Hortus third; a concise dictio-
nary of plants cultivated in the United States and Canada. Mac-
millan, New York, NY.
LIU, J. S., AND C. L. SCHARDL. 1994. A conserved sequence in internal
transcribed spacer 1 of plant nuclear rRNA genes. Plant Molecular
Biology 26: 775–778.
LOURTEIG, A. 1970. The correct name for Oxalis ‘‘siliquosa’’ [Oxali-
daceae]. Baileya 17: 38–39.
. 1975. Oxalidaceae extra-austroamericanae. I. Oxalis L. Sectio
Thamnoxys Planchon. Phytologia 29: 449–471.
. 1979. Oxalidaceae extra-austroamericanae. II. Oxalis L. Sectio
Corniculatae DC. Phytologia 42: 57–198.
. 1981a. Oxalidaceae extra-austroamericanae. IV. Oxalis L. Sec-
tio Articulatae Knuth. Phytologia 50: 130–142.
. 1981b. Oxalidaceae in Flora of Panama. Annals of the Missouri
Botanical Garden 67: 823–850.
. 1983. Oxalida´ceas. Flora Ilustrada Catarinense. Santa Catari-
na, Brasil.
. 1994. Oxalis L. subge´nero Thamnoxys (Endl.) Reiche emend.
Lourt. Bradea, Boletim do Herbarium Bradeanum 7: 1–199.
MACBRIDE J. F. 1943. Flora of Peru. Field Museum of Natural History,
Botanical Series, vol. XIII, part III, no. 1, Publication 531. 544–
608.
MARKS, G. E. 1956. Chromosome numbers in the genus Oxalis. New
Phytologist 55: 120–129.
MATHEW, P. M. 1958. Cytology of Oxalidaceae. Cytologia 23: 200–210.
MEDINA H., T. C. 1994. Contaje cromoso´mico de la oca (Oxalis tuberosa
Molina) conservada in vitro. Tesis de Ingeniero Agro´nomo, Uni-
versidad Nacional del Centro del Peru´, Huancayo, Peru.
NARANJO, C. A., L. M. MOLA, L. POGGIO, AND M. MULGURA DE ROMERO.
1982. Estudios citotaxonomicos y evolutivos en especies herbaceas
sudamericanas de Oxalis (Oxalidaceae). I. Boletı´n de la Sociedad
Argentina de Botanica 20: 183–200.
NATIONAL RESEARCH COUNCIL. 1989. Lost crops of the Incas: little-
known plants of the Andes with promise for worldwide cultivation.
National Academy Press, Washington, DC.
NICKRENT, D. L. 1994. From field to film: rapid sequencing methods
for field-collected plant species. BioTechniques 16: 470–475.
RIESEBERG, L. H., AND J. F. WENDEL. 1993. Introgression and its con-
sequences in plants. In R. G. Harrison [ed.], Hybrid zones and the
evolutionary process, 70–109. Oxford, New York, NY.
ROSE, J. N. 1906. Studies of Mexican and Central American plants—
No. 5. Contributions from the National Herbarium 10: 79–117.
RUSBY, H. H. 1920. Descriptions of three hundred new species of South
American plants, with an index to previously published South Amer-
ican species by the same author. Published by the author, New York.
SALTER, T. M. 1944. The genus Oxalis in South Africa, a taxonomic
revision. Journal of South African Botany (suppl. 1): 1–355.
SANG T., D. J. CRAWFORD, AND T. F. STUESSY. 1995. Documentation of
reticulated evolution in peonies (Paeonia) using internal tran-
scribed spacer sequences of nuclear ribosomal DNA: implication
for biogeography and concerted evolution. Proceedings of the Na-
tional Academy of Sciences, USA 92: 6913–6817.
SMALL, J. K. 1903. Flora of the southeastern United States. Published
by the author, New York.
. 1907. Oxalidaceae. In North American Flora 25: 25–58.
SMITH, J. D. 1897. Undescribed plants from Guatemala and other Cen-
tral American republics. XVIII. Botanical Gazette 23: 234–251.
SOLTIS, D. E., AND P. S. SOLTIS. 1993. Molecular data and the dynamic
nature of polyploidy. Critical Reviews in Plant Sciences 12: 243–273.
SWOFFORD, D. L. 1991. PAUP: phylogenetic analysis using parsi-
mony, version 3.1.1. Computer program distributed by the Illinois
Natural History Survey, Champaign, IL.
TALLEDO, D. AND C. ESCOBAR. 1995. Citogene´tica de Oxalis tuberosa:
ciclo celular y nu´mero cromoso´mico. Biotempo (Universidad Ri-
cardo Palma, Lima, Peru) 2: 33–46.
TOSTO, D., AND H. F. HOPP. 1996. Sequence analysis of the 5.8S ribo-
somal DNA and internal transcribed spacers from five species of
the Oxalis tuberosa alliance. DNA Sequence 6: 361–364.
VALLADOLID, A. 1996. Niveles de ploidı´a de la oca (Oxalis tuberosa
Mol.) y sus parientes silvestres. Master’s thesis. Universidad Na-
cional Agraria La Molina, Lima, Peru.
, C. ARBIZU AND D. TALLEDO. 1994. Niveles de ploidı´a de la oca
(Oxalis tuberosa Mol.) y sus parientes silvestres. Agro Sur (numero
especial): 11–12.
VALLENAS, M. 1992. Fructificacio´n, produccio´n y viabilidad de semilla
sexual en oca (Oxalis tuberosa Mol.) In: D. Morales and J. J. Vach-
er [eds.], Actas del VII congreso internacional sobre cultivos an-
dinos, 255–258. Instituto Boliviano de Tecnologı´a Agropecuaria,
La Paz, Bolivia.
WAGNER A., N. BLACKSTONE, P. CARTWRIGHT, M. DICK, B. MISOF, P.
SNOW, G. P. WAGNER, B. J. BARTELS, M. MURTHA, AND J. PENDEL-
TON. 1993. Surveys of gene families using polymerase chain re-
action: PCR selection and PCR drift. Systematic Biology 43: 250–
261.
WARBURG, E. F. 1938. Taxonomy and relationship in the Geraniales,
Part II. New Phytologist 37: 189–210.
WELLER, S. G., AND M. F. DENTON. 1976. Cytographic evidence for
the evolution of distyly from tristyly in the North American

July 1998] 985EMSHWILLER AND DOYLE—ORIGINS OF CULTIVATED OXALIS TUBEROSA
species of Oxalis section Ionoxalis. American Journal of Botany
63: 120–125.
WENDEL, J. F., A. SCHNABEL, AND T. SEELMAN. 1995. Bidirectional in-
terlocus concerted evolution following allopolyploid speciation in
cotton (Gossypium). Proceedings of the National Academy of Sci-
ences, USA 92: 280–284.
WHITE, T. J., T. BRUNS, S. LEE, AND J. TAYLOR. 1990. Amplification and
direct sequencing of fungal ribosomal RNA genes for phylogenet-
ics. In M. Innis, D. Gelfand. J. Sninsky, and T. White [eds.], PCR
protocols: a guide to methods and applications 315–322. Academic
Press, San Diego, CA.
YOKOTA, Y., T. KAWATA, Y. IIDA, A. KATO, AND S. TANIFUJI. 1989. Nu-
cleotide sequences of the 5.8S rRNA gene and internal transcribed
spacer regions in carrot and broad bean ribosomal DNA. Journal
of Molecular Evolution 29: 294–301.
ZUCCARINI, J. G. 1825. Monographia der amerikanischen Oxalis-Arten.
Denkschriften der Ko¨niglichen Akademie der Wissenschaften Mu¨n-
chen ser. 1 9: 129–184.
. 1831. Nachtrag zu der Monographie der amerikanischen Ox-
alis-Arten. Denkschriften der Ko¨niglichen Akademie der Wissen-
schaften Mu¨nchen ser. 2 1: 181–276.