| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Invited Special Papers |
228 Plant Science Building, L. H. Bailey Hortorium, Department of Plant Biology, Cornell University, Ithaca, New York 14853-4301 USA
Received for publication February 2, 2004. Accepted for publication June 17, 2004.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
Key Words: angiosperms • fossil history • minimum age • node dating
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
). This special issue of the American Journal of Botany reveals that progress is being made in understanding these problems and that there is imminent hope for more detailed and accurate understanding of these events. It also illustrates that three broad areas in particular have contributed to our improved understanding of angiosperm history and relationships: new data and methods for molecular genetics (Palmer and Zamir, 1982
; Matthews and Donoghue, 1999
), renewed interest in paleontology (Crane et al., 1989
; Crepet and Nixon, 1996
; Gandolfo et al., 1998a
; Zhou et al., 2001
), and new developments in the analysis of phylogenetic relationships (Nixon, 1996
, 1999
; Goloboff, 1999
). There is also an invigorated interest in methods for dating of diversification events (Sanderson et al., 2004
). What is distinctive about this moment and also raises the prospect that problems in understanding angiosperm history and relationships will ultimately be resolved is that there is increasing synthesis of these disparate approaches and resultant heightening of insights into the "abominable mystery." This special issue illustrates the momentum of this growing synthesis. Our contribution is intended to provide a perspective on one aspect of the potential importance of paleontological data: timing.
The fossil record, as a body of knowledge that incorporates morphology, biogeography, and ecology in a temporal context, has both informative potential and the capacity to serve as a corroborative mechanism for a variety of hypotheses about evolution, development/homology, relationships, biogeography, and generally about the evolutionary play in the ecological theater (sensu Hutchinson, 1965
). In addition, the fossil record, as representative of the history of life, holds the potential for clarifying relationships among extant taxa by revealing extinct mosaic taxa that link modern ones, in addition to providing the general pattern of evolution of taxa through time. Historically, and in the context of evolutionary biology, the fossil record has played both informative and corroborative roles and continues to be called upon to do both. However, the advent of modern methodologies for comparative studies of extant taxa invites a reassessment of the primacy and scope of the fossil record in addressing questions of evolution and systematic relationships.
The fossil record post evolutionary theory: utility and limitations in phylogenetic studies
Whereas in a "pre-evolutionary" world scientists struggled to understand the meaning of the fossil record and were ultimately informed by it and post evolutionary-theory scientists sought fossils to confirm and understand the process of evolutionary change and to clarify relationships among extant taxa, in this genomics era, scientists have an additional tendency to look to the fossil record to confirm increasingly accurate hypotheses of evolutionary relationships based on comparative studies of extant taxa. These relationships have become increasingly clear and accurate through analyses of multiple gene sequences made possible principally through pioneering innovations by, for example, Palmer and Zamir, 1982
; Palmer et al., 1983
, 1985
; Downie and Palmer, 1992
, who later, through multi gene analysis, suggested along with others that Amborella and Nymphaeales were among basal extant angiosperms (e.g., Parkinson et al., 1999
; Matthews and Donoghue, 1999
; Qiu et al., 1999
; Soltis et al., 1999
; Chaw et al., 2000
; Graham and Olmstead, 2000
; Soltis et al., 2000
; and Barkman et al., 2000
, who in contrast to the others, placed Amborella + waterlilies as a sister group to the remainder of angiosperms; also see Zanis et al., 2002
, and Soltis and Soltis, 2004
, for a detailed discussion). Notable among these analyses is the most inclusive, based on numbers of taxa simultaneously analyzed and made possible via various powerful algorithms that allow the rapid analysis of large matrices representing numerous taxa (Goloboff, 1999
; Nixon, 1999
) supporting the basal placement of Amborella and Nymphaeales (Soltis et al., 2000
). These analyses contribute to a growing consensus about the major relationships in angiosperms (e.g., Graham and Olmstead, 2000
; Savolainen et al., 2000
; Zanis et al., 2002
, 2003
; Borsch et al., 2003
; Hilu et al., 2003
; Magallón and Sanderson, 2002
; Chase, 2004
; Judd and Olmstead, 2004
; Soltis and Soltis, 2004
). Such comprehensive cladograms (e.g., Soltis et al., 2000
) provide additional opportunities for evaluating the fossil record or vice-versa. Corroborative use of fossil evidence in supporting such hypotheses can be as straightforward as observing broad correlations between the fossil record and ordinal sequences implied by hypotheses of phylogeny (Friis et al., 2001
) or more complicated and sophisticated as in statistically checking the fit of the fossil record to model-based expectations derived from phylogenies based on comparative analyses of extant taxa (e.g., Benton, 2001
). These latter approaches and others (e.g., Norell and Novacek, 1992
) have been somewhat controversial and variously have additional qualifications (e.g., those that limit the field of the comparisons to make it more compatible with available fossil data; Benton, 1998
).
Another approach is to use fossil data to calibrate molecular rate substitution models (often popularly called "molecular clock" models—e.g., Soltis and Soltis, 2003
; Hedges and Kumar, 2004
; and see Sanderson et al., 2004
) that are then used to date evolutionary events with any one of several methods including most commonly "rate smoothing" or maximum likelihood. When estimates based on substitution models differ from observed fossil ages, the molecular estimates are typically accepted as "better" (more accurate) due to the inherent underestimation and variability of an incomplete fossil record. This has resulted in extremely disparate and sometimes outlandish estimates of the ages of major clades (see Graur and Martin, 2004
; but also Hedges and Kumar, 2004
, for a response). The purpose of this paper is not to review such methods in detail (see Sanderson et al., 2004
, for relevant discussion); the reader is also referred to Graur and Martin (2004
; but also see the response to it by Hedges and Kumar, 2004
) for cogent criticisms of both the methods and some interesting and celebrated specific applications and for the responses. Both of these publications point out two important factors that have relevance to this paper and will be discussed further: weaknesses in the fossil record and the overriding importance of accurate fossil identifications. Absent the alleged hyperbole of Graur and Martin (according to Hedges and Kumar, 2004
), from our perspective, molecular rate substitution methods, although recently extremely popular (Wikström et al., 2001
; Davies et al., 2004
; Yoo et al., 2004
), are complicated by numerous factors including but not limited to assumptions about constant substitution rates among different lineages, problems with calibration due to incomplete fossil records, and/or the complexity of models that attempt to compensate for substitution variability (e.g., discussion in Sanderson, 1997
; Bremer, 2000
; Soltis and Soltis, 2003
; Sanderson et al., 2004
) and what have been called deceptively "errorless" numbers (Graur and Martin, 2004
). These methods also usually suffer from the need to simplify calculations by using a single divergence point (i.e., often a single fossil or related fossils; see Wikström et al., 2001
; Davies et al., 2004
; Yoo et al., 2004
; see Sanderson et al., 2004
, for relevant discussion) for calculating rates that are then applied throughout the tree, compounding the potential problem of incorrect calibration and problems with error due to an incomplete fossil record (Graur and Martin, 2004
). With all of these potential problems, given the present state of the art, even very careful and comprehensive analyses can be dramatically inaccurate and misleading. A relevant and striking example of a failure of molecular rate estimation even with an apparently careful analysis is provided by a recent preliminary study of Nymphaeaceae using three of the most common estimation methods: rate smoothing (NPRS), penalized likelihood (PL), and a "Bayesian" approach (Yoo et al., 2004
). Although their study estimated the age of modern ("crown-group") Nymphaeaceae to be ca. 40 million years old, extremely well-preserved fossils that nest within modern Nymphaeaceae based on careful phylogenetic analysis are now known from ca. 90 million-year-old deposits (see Microvictoria discussed later; Gandolfo et al., 2004
). In this case, the failure of molecular substitution rate methods is probably at least in part from poor calibration based on an incomplete understanding of the fossil record, emphasizing the need for precisely identified fossils in this context. Projecting such models over larger trees based on single calibration points (e.g., Wikström et al., 2001
), therefore, is highly suspect and raises questions about the efficacy of using calculated timing of angiosperm radiation events in assessing ecological factors in their radiation (e.g., Davies et al., 2004
).
Thus, attempts to use molecular substitution rate analyses to assess timing in angiosperm evolution (Sanderson, 1997
; Magallón et al., 1999
; Wikström et al., 2001
; Davies et al., 2004
; Yoo et al., 2004
), controversial in and of themselves (e.g., Graur and Martin, 2004
; see discussion in Sanderson et al., 2004
), are, in any case, unavoidably compromised by the inherently imperfect nature of the fossil record manifest in:
Given the limitations noted, the fossil record is likely to be biased, both ecologically and phylogenetically relative to modern diversity. Thus, in addition to the potential dangers in attempting to use the fossil record to calibrate rates of evolution based on molecular data [especially in cases where the choice of fossil(s) for calibration might be problematical], there is no reasonable expectation that available fossils will accurately reflect the evolutionary history of every group, and certainly, the fossil records of some groups (e.g., vertebrates, shelled mollusks) are likely to be much more reliable than others. Therefore, we believe that efforts to "test" the congruence of the fossil record with cladograms of modern and/or fossil taxa (e.g., Norell and Novacek, 1992
) are of ambiguous value. If such a test revealed that the fossil record was congruent (or consistent) with the cladograms of modern taxa, the possibility would remain that such congruence was illusory due to sampling error. Alternatively, if the test reveals incongruence, it would be unwise to reject cladograms based on molecular and/ or morphological evidence of extant, observable taxa due to incongruence with an admittedly incomplete and thus possibly biased fossil record. Furthermore, using this "congruence" approach to discriminate between alternative extant taxon-based hypotheses of phylogeny might be useful in certain instances today (e.g., Benton, 1998
), but one might speculate that, as analyses of modern taxa become more inclusive and complete, there will be fewer instances of conflicting hypotheses until there is general agreement, as seems to be the trend in estimating angiosperm relationships (see papers in this volume and references therein e.g., Parkinson et al., 1999
; Chaw et al., 2000
; Savolainlen et al., 2000
; Soltis et al., 2000
; Hilu et al., 2003
).
We believe that a straightforward and informative approach to combining fossil evidence with hypotheses of phylogeny based on comparative studies of extant taxa is available through the simple exercise of minimum age node mapping. We use this methodology herein to assess timing in the history of the angiosperms and to explore some of the broad implications of the results.
Goals—minimum age mapping—a means for the fossil record of angiosperms to retain the potential to be both informative and corroborative
By mapping estimated minimum ages for all possible clades on a large comprehensive angiosperm cladogram, we intend to provide estimates for some of the more important angiosperm clades based on the best available mesofossil (middle-sized fossils), macrofossil and, in a few cases, palynological evidence. We have taken a selective approach, using only the fossils that have either been assigned through direct cladistic analysis or have clearly identifiable characters to place them "unequivocally" (i.e., with high confidence) in an identifiable clade of modern angiosperms. As will be discussed in Materials and Methods, we have excluded some fossils from our analysis, following the caveats raised by the original authors related to ambiguity, conflicting placements, or low confidence in their taxonomic assignments. This does not mean that these excluded fossils should be ignored in future analyses; indeed, many are merely in need of careful analysis in a phylogenetic context in order to place them in a broader phylogeny of angiosperms.
Because the evidence for both extant and fossil taxa is much more complete for certain clades, we devote sections to some of the more important ones (assessed subjectively) and deal with these and the fossils that represent them in greater detail.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
; Lawrence, 1951
), had with occasional exceptions (Conwentz, 1886
) been essentially unstudied from the perspective of fossils because they were generally regarded as too rare and/or poorly preserved to make substantial contributions to our understanding of angiosperm history and relationships. Now they are the principal foci of many research laboratories, and have been since the beginning of a concerted and continuing effort on investigations of fossil flowers (Crepet et al., 1974
), which have been followed by a sustained and continuing line of investigations based on fossil flowers by the same authors and others (e.g., Crepet et al., 1975
, 1980
; Crepet and Dilcher, 1977
; Tiffney, 1977
; Crepet, 1978
, 1979a
, b
, 1984a
, b
, 1989
; Crepet and Stuessy, 1978
; Crepet and Daghlian, 1980
, 1982a
, b
; Daghlian et al., 1980
; Zavada and Crepet, 1981
; Krassilov et al., 1983
; Crepet and Taylor, 1985
, 1986
; Crepet and Friis, 1987
; Friis and Crepet, 1987
; Taylor and Crepet, 1987
; Stockey, 1987
; Crepet and Feldman, 1991
; Stockey and Pigg, 1991
; Pedersen et al., 1994
; Stockey and Rothwell, 1997
; Hernández-Castillo and Cevallos-Ferris, 1999
; Boucher et al., 2003
).
A very significant development within fossil flower studies has been the investigation of charcoalified fossils preserved in three dimensions. These fossils have been of particular importance because their detailed three-dimensional preservation captures numerous characters that are often unavailable in compressed specimens and because their ages—ranging from Early to Late Cretaceous—represent a critical interval in angiosperm radiation. Since their first discovery on Martha's Vineyard, Massachusetts, USA by Tiffney (1977)
and later in southern Sweden by Friis and Skarby (1981)
, such fossils have been the object of numerous significant continuing investigations as have similar fossils from North America and Portugal (e.g., Friis, 1983
, 1984
, 1990
; Friis et al., 1986
, 1988
, 1992
, 1994
, 1997a
, b
; Crane et al., 1989
; Drinnan et al., 1990
, 1991
; Crepet et al., 1991
, 1992
; Herendeen et al., 1993
, 1994
; Nixon and Crepet, 1993
; Crepet, 1996
, 1999
, 2000
, 2001
; Crepet and Nixon, 1996
, 1998a
, b
; Eklund et al., 1997
; Gandolfo et al., 1998a
, b
, c
, 2000
, 2002
, 2004
; Magallón-Puebla et al., 1997
; Zhou et al., 2001
; Eklund, 2003
; Hermsen et al., 2003
; and others). In aggregate, these fossils have made substantial contributions to our understanding of Cretaceous angiosperm history and they continue to be the objects of sustained inquiries. Yet, it is important to note that compressed and even petrified flowers have also continued to make contributions to our understanding and constitute a significant element of angiosperm paleobotany to this day (e.g., Dilcher et al., 1976a,
b,
1978
; Doyle and Hickey, 1976
; Crane, 1981
, 1989
; Basinger and Dilcher, 1984
; Crane and Stockey, 1985
; Dilcher and Kovach, 1986
; Dilcher and Manchester, 1988
; Dilcher and Basson, 1990
; Call and Dilcher, 1992
; Manchester, 1992
; Mohr and Friis, 2000
; Sun et al., 2002
; Taylor and Hickey, 1990
, 1992
).
Flower fossil studies would have failed to reach their full potential without the second major development—that of algorithms for evaluating phylogenetic relationships based on character arrays. The general topic of phylogenetics and fossils has been addressed by Nixon (1996)
and is discussed by Crane et al. (2004)
, but the need for precise and reliable identification of fossils that often represent extinct taxa requires that they be identified through a phylogenetic context. Such an approach makes the process of identifying a fossil transparent and includes an explicit illumination of the characters attributed to it. This process adds objectivity to identifications that may be inherently difficult, as in cases where the fossils might be mosaics relative to extant taxa. Overall phylogenetic approaches to understanding plant relationships by means of modern cladistic methods were initiated by Parenti (1980)
followed by Young and Seigler (1981
; angiosperms), Hill and Crane (1982)
, Crane (1985)
, Doyle and Donoghue (1986)
, Loconte and Stevenson (1990)
, Nixon et al. (1994)
, Rothwell and Serbet (1994)
and by subsequent reanalyses by the same authors. Most recently, there have been analyses of molecular-based or combined data sets, which are noted later (e.g., Soltis et al., 2000
). Use of phylogenetic analysis to identify fossil flowers has a separate, but more or less parallel history and issues specific to including fossil evidence in phylogenetic analysis have been discussed by Nixon (1996)
. Early applications of phylogenetic context for identification of fossils included Crane and Manchester (1982
, fruits of Juglandaceae) and Crepet and Nixon (1989a
, b)
on Tertiary Fagaceae. And, this methodology has now become standard in identifying charcoalified flowers as in the Turonian taxa Tylerianthus and Dressiantha (Gandolfo et al., 1998a
, b
) and others (e.g., Sims et al., 1999
). There is additional and interesting potential for illuminating the fossil history of plants through the use of phylogenetic analysis as seen in Gandolfo et al. (1997)
—that of assembling the dispersed organs of a taxon through their relative placements in phylogenetic context.
One of the major problems with attempts to rationalize the fossil record with phylogenies based on modern taxa [either by various means of direct comparison testing for congruence (e.g., Norell and Novacek, 1992
; Benton, 2001
)] or by using fossils as a basis for timing the rate of substitution in lineages and extrapolating that to other groups to assess the pattern on timing in angiosperms clades (Bremer, 2000
) is the circumscription of the set of fossil taxa that will be used in the analysis. Indeed, this selection dramatically alters the outcome of any analysis, given that fossil reports of almost any existing angiosperm taxon can be found in the paleobotanical literature. Many taxa have been misidentified over the years as revealed by new analytical methodologies and investigatory tools. In other cases, perfectly reasonable but too general assessments of affinities have been constrained by the nature of the fossil evidence. Because selection of fossil taxa that are reliably identified is critical to this analysis and to the utility of fossil evidence in related applications (e.g., Yoo et al., 2004
), we have been conservative in selecting taxa to be included in determining the minimum ages of modern clades (Table 1). Consequently, we restrict our analysis to taxa whose presence in the Cretaceous is supported by the strongest fossil evidence. We define the latter as having affinities based upon the outcome of phylogenetic analysis that includes the fossils or alternatively, the presence of a suite of unequivocally demonstrated fossil characters that uniquely matches the taxonomic unit to which the fossil is assigned.
|
|
; Harley and Zavada, 2000
), we have used such evidence sparingly in this analysis. In cases where the pollen identification seems relatively strong, based on unique combinations of characters that are found only in one modern clade, we have accepted determinations and used these records for our estimates. The caveats and problems associated with literature references to ostensibly identified fossils outlined earlier apply to fossil pollen as well, due to fewer distinctive characters. A cogent example of this phenomenon is the early record of supposed monocot pollen, but difficulties in pollen identification extend to various modern dicotyledonous families.
Criteria for inclusion of fossils for determining the ages of modern clades (i.e., through minimum age mapping)
(1) Fossil identifications are dependent on either cladistic morphological analyses or the presence of well-documented synapomorphies. (2) Accurate aging of fossil sites is required. We selected sites whose geology and ages are well known and are not controversial and have interpreted the ages conservatively by assigning the youngest of the suggested range of ages. (3) A high quality cladogram of modern taxa is required. We used the widely accepted Soltis et al. (2000)
three-gene 567-taxon cladogram as the basis for mapping major clades of angiosperms. This cladogram was selected because it provides sufficient detail and resolution to map the most important Cretaceous fossils of angiosperms that can be reasonably identified.
Magnoliid groups
We have found it most challenging to delimit the number of fossil taxa included in magnoliid groups (sensu Soltis et al., 2000
, as opposed to the more restrictive use of "magnoliid" in APG II, 2003). These taxa are significant because they are relevant to early branches in the angiosperms, thus the timing of their history ramifies significantly throughout the rest of the angiosperms. They are difficult because they include some of the earliest fossil evidence of the angiosperms, and these fossils are often imperfectly or incompletely preserved, making them the most difficult group to identify and thus the most likely group to be inadequately identified. The monocots too are nested within the broadly defined magnoliid group (sensu Soltis et al., 2000
), and although many claims have been made regarding Early Cretaceous monocot fossils, their early fossil record is generally problematical (Gandolfo et al., 2000
). Although the actual number of magnoliid/monocot fossils that we include in this analysis based on our criteria is relatively low, available evidence suggests that a large portion of the modern diversity within the "magnoliid" families, including the basal monocots, had developed by about 90 million years ago (mya). There have been a remarkable number of significant fossil discoveries of mesofossils or compressions of flowers of Cretaceous age during the past 25 years (e.g., Dilcher and Crane, 1984
; Crane et al., 1993
; Nixon and Crepet, 1993
; Crepet and Nixon, 1994
, 1998a
, b
; Friis et al., 1997a
, b
). Even more recently, the number of reports of Early Cretaceous angiosperm mesofossils of flowers has increased. We have been very careful in accepting purported affiliations for the fossils using the criteria delineated because it is our intention to make the dating as reliable as possible (for example, all records of Early Cretaceous monocots have been excluded based on the limited characters available; Gandolfo et al., 2000
). Several reports of Early Cretaceous angiosperms that support fossil record congruence with phylogenies based on nucleic acid sequences in key aspects or that are relevant in the timing of monocot history have not been included in our analysis and deserve additional discussion.
Included/excluded fossils
We have included as many Cretaceous fossils as possible that conform with the criteria delineated (Table 1). Fossil reports have been excluded if, based on the available evidence (including caveats raised by the original authors regarding their affinities), their taxonomic placements are too ambiguous to provide minimum age estimates with sufficient confidence. The sum of excluded + included fossils is not exhaustive relative to all known Cretaceous angiosperms, and in many cases where younger fossils occur for a clade, we have neither accepted nor rejected these because, in either case, they would not affect the estimated age of the nodes ancestral to them. We provide next a brief discussion of the most prominent of the explicitly excluded fossils (predominantly fossils that have been assigned to the magnoliids/monocots). Note that it is quite possible that some of these taxa do indeed have affinities suggested for them in the literature (and in fact, as will also be noted, our methodology has the potential of providing some indirect support for these assignments), but the number of characters present in the fossils precludes definitive assignments by our criteria, thus it would be potentially misleading to include them in our analysis. Moreover, in some cases, as will be noted, uncertainty about their affinities is expressed in the original descriptions.
Lesqueria elocata Crane and Dilcher (1984)
Mid Cretaceous. This is a fruiting axis with magnoliid affinities, but could represent a taxon that would be placed on a lower node of the cladogram. The authors do not place it in any particular family. They agreed that this fossil taxon has some characters of the Magnoliidae; however, they pointed out that "we know of no Recent plant that has the combination of numerous helically arranged, dehiscent follicles with bifid tips borne in a swollen head at the apex of a long receptacle bearing numerous, persistent, spirally arranged, laminar structures" (p. 398).
Archaeanthus linnenbergeri Dilcher and Crane (1984)
Uppermost Albian or lowermost Cenomanian. This also represents an infructescence of magnoliid affinity but cannot be placed in any particular clade without further analysis. In the original paper, the authors concluded, "Archaeanthus is clearly most similar to extant Magnoliidae and share features with supposedly primitive members of the Hamamelidae and Dilleniidae... However, we do not believe that Archaeanthus usefully can be assigned to any extant family; Archaeanthus is a unique and extinct genus of fossil angiosperms" (p. 378).
Prisca reynoldsii Retallack and Dilcher (1981)
Mid Cretaceous. Another fructification is, Prisca reynoldsii, a raceme of unisexual multifollicles. The authors did not place these fossils within any modern family and suggested the erection of the family Priscaceae in order to place the fossils ("Subclass and order incertae sedis, Family Priscaceae fam. nov. This family is intended for fructifications like Prisca reynoldsii and other organs, such as leaves, pollen organs, staminate inflorescences and woods, which appear related to and distinctive of plants with such fructifications," p. 107). Later, Drinnan et al. (1990)
tentatively referred Prisca to Lauraceae, based on gross similarities to the fossil Mauldinia, but provided no evidence of diagnostic features of Lauraceae in Prisca, nor did they provide any synapomorphies or cladistic analysis.
Protomonimia kasai-nakajhongii Nishida and Nishida (1988)
Turonian (Late Cretaceous). This is a reproductive structure similar to Archaeanthus and Lesqueria. However, the authors did not place it within any family and considered that Protomonimia represents a magnolioid ancestor from which some modern Magnoliales may have been derived. In the diagnosis, the authors said that it is an "Angiosperm fructification consisting of a globose head and a woody peduncle; apocarpous and apistillate." They also pointed out, "Judging from the polycarpelous floral structure, presence of oil cells and relatively primitive wood structure composed mainly of scalariform vessels with steep scalariform perforation plates, the specimen can be related to the Magnoliales (sensu Cronquist, 1968)" (p. 419), and "... the direct connection between Protomonimia and Monimiaceae is still based more on speculation than on unequivocal evidence" (p. 422). This fossil requires a cladistic placement prior to inclusion for dating nodes.
Elsemaria kokubunii H. Nishida (1994)
Coniacian–Santonian. An angiosperm permineralized fruit that has affinities to some extant taxa formerly placed in the Dilleniidae, but the lack of stamens and other floral parts prevent placement within a family. It is in need of cladistic reevaluation. Nishida presented the following systematics: Subclass Dilleniidae, Order and Family incertae sedis, and indicated that "However, the lack of other floral parts in Elsemaria and the possible polyphyletic nature of multilocular capsules make it difficult to attribute Elsemaria to a particular order or family" (p. 133).
Appomattoxia ancistrophora Friis et al. (1995)
Early or Middle Albian. Well-preserved fossilized fruits and pollen are known, but these have never been placed in a cladistic analysis. The pollen bears similarity to modern Piperales, but the authors discuss other possible relationships and do not conclude with a definite taxonomic placement. The authors stated, "Detailed evaluation of the phylogenetic position of Appomattoxia will require a much clearer understanding of relationships among extant ‘basal’ angiosperms than is currently available, as well as additional information on the floral and vegetative structure in the fossil material" (p. 294).
Anacostia Friis et al. (1997b)
Albian? Although the pollen found adhering to fossilized unicarpellate fruits has some features usually associated with monocotyledons (finer reticulum at the ends of the grains), the combined features of the fruits and pollen are not found in any extant monocotyledonous family, and the authors leave open the possibility that this fossil is a magnoliid (=dicotyledonous). Anacostia comprises four species, A. marylandensis, A. virginiensis, A. portugallica, and A. teixeirae; all are new species. Explicitly, the authors remarked, "The character combination in the fossil material, with unicarpellate fruiting units containing a single anatropous seed and trichotomocolpate/monocolpate pollen, indicates a relationship between Anacostia and extant Magnoliidae or monocotyledons" (p. 241). They also pointed out, "The available characters for Anacostia are not sufficient to assign the fossil material to any extant family. The weight of the evidence from the fruiting units and seeds, as well as the pollen indicates a relationship with the magnoliids, but affinity with extant monocotyledons cannot be ruled out" (p. 242). Until the fossil is placed in a cladistic analysis, its taxonomic position cannot be evaluated.
Couperites Pedersen et al. (1991)
Mid-Cretaceous. Although these fossils are likely chloranthaceous, they were characterized as incertae sedis by the authors and need to be evaluated in a cladistic context. In the original paper, the authors presented doubts about where this fossil taxon should be placed taxonomically as indicated by the following quotation: "The chloranthaceous affinity of Couperites is also strongly supported by similarities in fruit and seed structure. However, differences in seed organization preclude unequivocal assignment of Couperites to the family (table 1)" (p. 588). Couperites is compared by the authors with selected families of the Laurales, Magnoliales and Piperales (table 1, p. 587).
Silvanthemum Friis (1990)
Late Cretaceous, Santonian–Campanian. Silvanthemum was originally considered to have strong affinities with Escalloniaceae, but that placement was not verified by subsequent analyses (Backlund, 1996
). Because the reanalysis did not include all taxa included here and there are significant differences in the topologies, it is not possible to place this fossil definitively. Evaluation in a broader, more complete analysis is necessary. In the original paper, Friis (1990, p.15)
pointed out, "The great similarity between Silvianthemum and modern Escalloniaceae further suggests that the woody taxa of the saxifragalean complex constitute an ancient stock," and also mentioned, "the close relationship to members of the Escalloniaceae, particularly to species of Quintinia, is documented not only by morphological and organizational similarities, but also by anatomical features including cellular details of the trichomes" (p. 15). However, when Silvianthemum was included in a phylogenetic analysis for the Dipsacales, it did not appear closely related to Quintinia (Backlund, 1996
). Backlund (1996)
clearly stated, "Analysis of the matrix shows that Silvianthemum occupies a stable but not strongly supported position as somewhat more advanced than the extant Quintinia (see fig. 3), and just outside the basal node of the Dipsacales" (p. 18).
Nymphaeaceae
The putative key phylogenetic position within angiosperms of the Nymphaeales and Amborella, as a sister group or sister groups of the remainder of angiosperms in DNA-based hypotheses of angiosperm phylogeny including the 567-taxon three-gene tree (Soltis et al., 2000
) and others (e.g., Parkinson et al., 1999
; Chaw et al., 2000
), makes the fossil record of Nymphaeales potentially important to understanding angiosperm origins and the nature of the earliest flowering plants. Thus, the occurrence of early nymphaeaceous fossils is critical in dating the lowest nodes in the angiosperm phylogeny, so it is important for the date to be based on unequivocal fossil evidence.
Archaefructus, an Early Cretaceous angiosperm from China, has been interpreted both as sister group to the rest of the angiosperms (Sun et al., 1998
, 2002
) or as an aberrant member of the Nymphaeales, morphologically distant from any modern member of that group (Friis et al., 2003
). The latter interpretation is based on a phylogenetic reanalysis of the original matrix of Sun et al. (2002)
that includes an additional taxon (Cabombaceae) and has several changed and unconventional character codings. As in the Sun et al. analysis, the results of Friis et al. (2003)
place Archaefructus as the sister group to the remainder of the angiosperms in some of the most parsimonious trees, but Archaefructus is placed with Cabomba in the Nymphaeales in other trees of the same length. We are skeptical of the latter interpretation because it was executed with one empirically verifiable character miscoding (in their revised matrix, Friis et al. coded Cabomba as having only dichotomously veined dissected leaves, while in fact it also has entire floating leaves with numerous clearly anastomosing veins). When the Friis et al. matrix is modified to reflect the actual variation in leaf form within Cabomba by changing only the single matrix cell from dichotomous to polymorphic, Archaefructus again becomes a sister group to the angiosperms in 100% of the shortest trees and is never placed with Cabomba or in the Nymphaeales, thus conforming with the results of Sun et al. (2002)
. This restoration of the outcome of the original analysis by Sun et al. (2002)
has been achieved by making only that one change even though other matrix modifications by Friis et al. (2003)
, including a somewhat unorthodox coding of a perianth as possibly present in Ginkgo (which has naked stalked ovules intermixed with leaves), were left in the matrix. The resultant stability of the phylogenetic position of Archaefructus as a sister taxon of modern angiosperms is thus ultimately reinforced, not refuted, by a careful examination of the Friis et al. (2003)
analysis. While Friis et al. (2003)
appealed to the possibility that any morphologically simple taxon could owe its structure to evolutionary reduction (and thus to the notion that relationships of structurally simple aquatics can at times be difficult to determine based on morphology alone), there is absolutely no evidence to support this possibility. The fossil is indeed unusual relative to modern taxa and thus remains enigmatic, but even with available state-of-the-art techniques for identification, analysis only provides one result. Further insights into the affinities or support for the phylogenetic position indicated by analysis of the morphological characters now available from the fossils will most likely come from additional fossil evidence and its analysis. In addition, placement within the Nymphaeales as suggested by Friis et al. requires a number of ad hoc hypotheses to explain the dramatic morphological contrast between Archaefructus and flowers of modern Nymphaeales, but such speculation is not necessary given the results of the reanalysis. Thus, based on the best available evidence, we do not accept the notion that Archaefructus is an extinct nymphaealean, consistent with our criterion that it shares no identifiable synapomorphies with any extant member of that clade.
Another report of fossil Nymphaeaceae (sensu Les et al., 1999
) from the Early Cretaceous (approximately 100–125 mya) sediments in Portugal (Friis et al., 2001
), while not definitive, is still significant in dating angiosperm history. The actual character suite demonstrated in the fossil is, upon phylogenetic reanalysis that includes additional relevant taxa (Gandolfo et al., 2004
), equally compatible with Illiciaceae as well as with other angiosperm families. The incomplete preservation of the fossil in combination with the absence of definitive synapomorphies of Nymphaeaceae argue against its precise placement in any particular clade of angiosperms. We have reflected this ambiguity by including it as an estimator of the youngest node that subtends both Nymphaeaceae and Illiciaceae in the three-gene tree, aged at 113 mya (Fig. 16). Because these two clades are sister taxa of the remainder of angiosperms (excluding Amborella), it provides the oldest record of the extant clade, or "crown group" of angiosperms. The most conclusive, earliest fossil evidence for Nymphaeaceae fossil flowers is from the Turonian (90 mya) of the USA, and these fossils have floral structures almost identical to those of modern Victoria (Nymphaeaceae). These fossil flowers have been placed as unequivocally nested within modern Nymphaeaceae by explicit phylogenetic analysis (Gandolfo et al., 2004
).
|
). Monocots have a traditionally sparse Cretaceous fossil record, and recent critical analysis of Lower Cretaceous fossils suggests that none are definitively monocotyledonous (Gandolfo et al., 2000
). There are relatively few reports of Upper Cretaceous monocots that conform to the criteria we employ for this study. There have been several reviews of the fossil record of the monocotyledons (Horwood, 1912 a
, b
; Doyle, 1973
; Daghlian, 1981
; Muller, 1981
; Herendeen and Crane, 1995
; Gandolfo et al., 2000
), and in each case, selected fossils have been helpfully discussed. However, as mentioned in all of these reviews, it is sometimes difficult to recognize fossils with monocotyledonous affinities, particularly, we now realize, due to a frequent lack of monocot synapomorphies in supposed monocot fossils. Without such synapomorphies, it is impossible to distinguish some monocots from a number of magnoliid dicots, compromising the early fossil record of monocots and defeating the purpose and utility of a review of the early monocot record. Thus, as with other fossil evidence used in this analysis, we excluded reports of fossil taxa that could not be unequivocally related to monocotyledons based on the presence of diagnostic synapomorphies. For example, in the latest review by Gandolfo et al. (2000)
, the genera Liliacidites Couper (pollen taxon) and Acaciaephyllum Fontaine (leaves) were included in a cladistic analysis because these two fossil taxa had previously been considered to be the oldest examples of monocots. However, both fossils actually lack monocot synapomorphies, and the most parsimonious placement of these purported early monocotyledonous taxa was among dicot clades.
Recently, Bremer (2000)
drew attention to a subset of Cretaceous fossils with purported monocotyledonous affinities by using them to date nodes on a plastid rbcL-based phylogeny of monocot taxa, providing the means for calculating the rate of change (base substitution rate) in particular lineages. The results of such a combined approach, while potentially informative, depend on the quality of the identification of the fossils. While it was certainly appropriate for Bremer to choose relevant fossil evidence from the paleobotanical literature, we are using a different and more stringent set of standards for fossil identification and do not include the same fossil taxa in this analysis for reasons discussed next.
Dicolpopollis Pflanzl was used to provide the minimal age for the family Tofieldiaceae (Chmura, 1973
) and Milfordia Erdtman for the age of the Restoniaceae-Flagellariaceae-Joinvillaceae complex. These assignments, however, were made in the absence of comparative ultrastructural data necessary to definitively identify these taxa on the basis of palynological evidence alone (Zavada, 1984
; Harley and Zavada, 2000
).
As with many other fossil assignments that have not been included in the present analysis, lack of synapomorphic characters does not necessarily preclude the possibility that the suggested affinities may be accurate (as in the suggested affinity of Dicolpopollis to Tofieldiaceae), but it would be premature and potentially misleading to use such unsupported identifications for dating particular clades. Indeed, in the absence of additional characters, credibility of these identifications can ultimately be assessed best in the context of a temporal framework provided by the methodology used herein.
Another very important complex of monocots has been placed temporally on the basis of the fossil palynomorph Milfordia, but Linder (1986)
indicated that there are three pollen morphs of Maastrichtian-Oligocene age related to Restoniaceae and Poaceae. They are Monoporites annulatus Hammen, Milfordia hungarica (Kedves) W. Kr., and Milfordia incertae (Th. & Pf.) W. Kr. None are adequately characterized (they lack comparative ultrastructural features); thus it is difficult and complicated to determine their affinities with certainty.
Pistia corrugata Lesquereux
The oldest confirmed fossil record of P. corrugata dates from the Paleocene of southwestern Saskatchewan, Canada (McIver and Basinger, 1993
); earlier fossils assigned to Pistia are suspect (Hickey, 1991
).
Cymodocea Koenig
The record of Cymodocea presented by Bremer is based on the fossil Thalassocharis bosqueti Debey ex Miquel collected from Upper Maastrichtian sediments of the Kunrade region, Netherlands (Voigt and Domke, 1955
). However, Voigt (1981
, p. 283) notes that "Thalassocharis seems to show some affinities with the recent genus Cymodocea but because flowers and fructifications are not known the exact phylogenetic and systematic position of Thalassocharis is still uncertain," a reservation shared by Daghlian (1981)
.
Typha L
Bremer (2000)
used the fossil seeds T. ochreacea Knobloch and Mai and T. protogaea Knobloch and Mai, from the Maastrichtian of Eisleben, (Knobloch and Mai, 1986
) to date the node for bur-weeds and cat-tails. However, these fossils have no unique characters to unequivocally relate them to the extant genus Typha and have not been evaluated in a cladistic analysis.
As in the cases of the putatively monocotyledonous palynomorphs, the preceding assessments do not exclude the possibility that any or all of these early fossils are monocots or even that they belong to the groups used to support Bremer's analysis, but they have not been shown to have sufficient characters to justify such specific taxonomic assignments. Until the relationships are confirmed, they cannot reliably be used as a basis for estimating the ages of monocot clades.
Thus by default, the oldest unequivocal evidence for fossil monocots is embodied in charcoalified fossil flowers from the Turonian of New Jersey that bear a combination of features found today only in the achlorophyllous Triuridaceae. These fossil flowers were described and placed phylogenetically in extensive detail (but after the publication of Bremer, 2000
), and their relationships within the monocots have been well established through parsimony-based phylogenetic analysis (Gandolfo et al., 2002
). Based on these analyses, they are well nested within the Triuridaceae as the sister group to the clade that encompasses the tribe Triurideae. Although these flowers constitute the earliest unequivocal record for monocots, the group is well nested within the monocot clade, and there is no reason to assume that triurids are morphologically similar to the basal monocot condition. The lack of other verified meso—and macrofossils of monocots from similarly aged or older Cretaceous sediments is likely due to preservational bias, assuming that early monocots were herbaceous, as discussed later. The Late Cretaceous also includes several fossils of Maastrichtian (Upper Cretaceous) age with unmistakable monocot synapomorphies (Hickey and Petersen, 1978
; Daghlian, 1981
; Herendeen and Crane, 1995
; Harley, 1996
; Herendeen et al., 1999
). These will be analyzed in greater detail in another paper (Gandolfo et al., unpublished manuscript).
Methods
Fossil data from both the literature and unpublished sources were entered into a relational database structure, and the program Winclada (Nixon, 2003
) was utilized to automatically map minimum ages for each node on the consensus of the 567-taxon three-gene tree (Soltis et al., 2000
). The procedure for optimizing node ages was simple: each node was assigned the greatest minimum age of any node that had been assigned above it. Ages for modern taxa with no fossil record were initially assigned an age of zero. Fossils were positioned on the 567-taxon tree according to their most likely identification based on assignment either by published cladistic analyses or the occurrence of definite synapomorphies and/or of unique character combinations.
The resulting cladogram with nodes assigned minimum ages was then reduced by collapsing nodes to show a more tractable view, emphasizing major clades for which fossil evidence is available. Estimated minimum ages are only mapped in these trees for nodes that have at least one fossil that can be associated with at least one descendant branch (0 age estimates are not shown).
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
.
|
; Bremer, 2000
). It is not a method per se for evaluating congruence of the hypothesis of phylogeny of angiosperms and the fossil record; nonetheless, it does place the phylogeny in a temporal framework and does have the potential to reveal inconsistencies between the fossil record and the best estimates of phylogenetic relationships based on comparative studies of nucleic acid sequences of living taxa. It also allows comparison of the temporal framework implied by the technique with the background fossil record of organs that are inconclusively assignable to particular taxa but that may give broader indications of the general pattern of timing in angiosperm diversification. Thus for example, although our analysis illustrates that the node that precedes the divergence of Nymphaeaceae and Illiciaceae is at a minimum 113 million years old, this is not based on a model of gene substitution, nor does it suppose an unbiased sample. This estimate is dependent only on the identification of the fossil that is the basis for the estimate and of course the veracity of the phylogeny.
The estimates provided here are consistent with a major diversification of several important angiosperm clades by the Turonian (
90 mya) that included numerous magnoliids, the monocots, and several tricolpate clades. Extensive Early Cretaceous monocot diversification<
