Iuiti J~I~I;I S

Home , Staminode

Corrections Proc. Nati. Acad. Sci. USA 89 (1992) 12207 Eveduton. In the article "Oldest fossil-flowers of hamael- ber 1, 1992, of Proc. Nati. Acad. Sci. USA (89, 8986-89), idaceous affinity, from the Late Cretaceous of New Jersey" the reproduction of Fig. 1 was not satisfactory (due to a by William L. Crepet, Kevin C. Nixon, Else Marie Friis, and printer's error, a splotch appeared between E and H). The John V. Freudenstein, which appeared in number 19, Octo- figure and its legend are reproduced below. IUItI j~i~I;I S-

0 F I ,...A .--

:.i; :1 :4

FIG. 1. Scanning electron micrographs of fossil inflorescences, flowers and organs. (A) Capitate staminate inflorescence showing florets, sepal cap lobes (sc). Each flower with four triangular stamen connectives (c), visible within sepal cup. (x23.) (B) One floret from a staminate inflorescence showing slightly separated anther connectives (c) opposite sepal cup lobes (sc). (x60.) (C) A lateral view of a partially broken staminal tube (st), within a sepal cup (sc), illustrating the thin-walled nature ofthe staminal tube, its hollow interior, and apical staminodes with adaxially bulbous extremities (bs). (x70.) (D) Floret with anthers removed, showing sepal cup and four sagittate staminode heads. (x55.) (E) Floret with two whole stamens removed revealing the incurved, contiguous staminodes (s) and one anther (a) sheared off illustrating cellular preservation. (x70.) (F) Closer view ofdistal staminode showing sagittate head (arrow), similar to shape offertile stamens. (x 140.) (G) Stamen removed from a floret showing the sagittate distal connective and thecae with slightly open valves. (x60.) (H) A single in situ pollen grain. (x4000.) (I) Pistillate inflorescence showing the somewhat abraded carpels and sepal lobes. (x34.) (J) Distal adaxial stamen connective epidermis showing three anomocytic stomata-one with broken guard cell walls. (x560.) (K) Close-up of a staminate floret with stamens removed to show the partially abraded sagittate heads of two staminodes (sh) and the broken stamen filament (sf) that appears to be interior to the two staminodal stalks. Note the broken sepal cup (arrow) for perspective. (x40.) (L) One tangentially sheared pistillate floret showing the bicarpellate syncarpous ovary inside of a sepal cup. (x105.) Downloaded by guest on October 1, 2021 Proc. Nati. Acad. Sci. USA Vol. 89, pp. 8986-8989, October 1992 Evolution Oldest fossil flowers of hamamelidaceous affinity, from the Late Cretaceous of New Jersey (Anglospermae/Hamaeiddae/pofna on/phylogeny) WILLIAM L. CREPET*, KEVIN C. NIXON*, ELSE MARIE FRlISt, AND JOHN V. FREUDENSTEIN* *L. H. Bailey Hortorium, Cornell University, Ithaca, NY 14853; and tNaturhistoriska Riksmuseet, Department of Palaeobotany, S-104 05 Stockholm, Sweden Communicated by Peter H. Raven, June 23, 1992 (receivedfor review November 4, 1991)

ABSTRACT Exceptionally well-preserved staminate inflo- cellular and epidermal features. It has been suggested that rescences, pistillate inflorescences, and detached stamens with this type of preservation is the result of rapid charcoalifica- important phylogenetic and paleoecological implications have tion during ancient flash forest fires (4). Fossils are isolated been discovered from the Turonian (ca. 88.5-90.4 million from the matrix by dissolving the matrix in water and sieving, years B.P.) Raritan Formation of New Jersey. The fossils have followed by tedious sorting oftiny fragments ofcharcoalified a combination of floral and pollen characters found in various material. In addition to a large and growing number of other genera ofmodern entomophilous and anemophilous Hamamel- angiosperm taxa, a series of fossil staminate inflorescences, idaceae and anemophilous Platanus (Platanaceae). The floral pistillate inflorescences, and detached stamens of apparent characters of the fossils, induding a sepal cup, stamil tube, hamamelidaceous/platanoid affinity have been discovered and apparently nectariferous staminodes, indicate that this by these methods. These fossils were all removed from a taxon was probably insect pollinated. The juxtaposition of relatively small sample (ca. 15 kg of matrix). One striking character complexes in an extinct taxon from disparate modern aspect of these fossil flowers is the extremely small size (see taxa provides an interesting phylogenetic perspective on the below), a feature of virtually all of the fossil flowers isolated origins of Iamamelidaceae and is a striking example of a fossil from these sites to date. that is a mosaic of familial level characters relative to modern Fossil staminate heads are ca. 1-2 mm in diameter, with taxa. Of even broader interest, however, is the occurrence of about 15 sessile, tightly packed florets (Fig. 1A). Individual stMinodal nectaries that have structural characters interme- florets are between 0.5 and 0.75 mm in width. Staminate diate between the fossil's functional stamens and modern florets have a sepal cup that is distally four-lobed (Fig. 1 B and hamamelidaceous petals. This transitional staminode morphol- D). In young inflorescences, the sepal lobes are slightly ogy in the context of the other fossil characters suggests a imbricate. Each flower has four functional stamens that are staminodal origin of petals in the hamamelid-rosid lineage. free from the calyx and inserted opposite the calyx lobes (Fig. This hypothesis is supported by the apparent staminode posi- 1B). The most striking features of the stamens are the tion within the fossil flowers where petals are found in modern massive incurved connectives (Fig. 1B). These appear trian- genera. The character complex ofmorphologically transitional gular in surface view, have a nonpapillar, smooth surface staminodes, a staminal tube, and sepal cup can be viewed as with anomocytic stomata (Fig. 1J), and have small auricular prehypanthial, lacking only fusion of the staminal tube to the extensions resulting in a sagittate outline (Fig. 1G). Stamens sepal cup. The appearance of the character complex embodied have very short filaments with almost sessile anthers that in these flowers during the late mid-Cretaceous may signal the have two locules at maturity (Fig. 1G). Each anther locule of has a single elongate slit that is bifurcate at one or both ends, early stages the relationship between specialized pollinators, suggesting that anther dehiscence was valvate. Four stami- such as bees, and the h a elid-rosid-asterid lineage of nodes are present in a single whorl alternate with the fertile angiosperms, arguably one of the most important events in stamens and inside the calyx (Fig. 1 E, F, and K). These angiosperm radiation. staminodes are strongly incurved, and their distal heads actually touch or overlap slightly in the central portion of the One of the dilemmas facing paleobotanists is the rarity of flower (Fig. 1 E and K). Below the anthers, the filaments of fossil taxa that are phylogenetically informative-i.e., that both staminodes and stamens are united into a hollow, are mosaics ofcharacters found in separate modern taxa. The cylindrical/hourglass-shaped staminal tube (Fig. 1C). The fossils described below, from the Turonian (88.5-90.4 million staminal filaments appear to originate inside of the whorl of years B.P.; ref. 1) Raritan Formation ofNew Jersey (2, 3), are staminodes (Fig. 1K). Although the traces that lead to the a notable exception. These fossils have a character mosaic of staminodes also appear to be inserted slightly outside of the the modern hamamelidid families Platanaceae and Hamamel- stamen traces in specimens where the staminal tube is idaceae and are therefore of considerable phylogenetic in- broken, this filament tube is extremely thin and it is difficult terest. They also provide evidence about the origin of petals to be certain of the relative positions of the whorls that form in a major lineage of flowering plants, the Hamamelididae- it, presumably by congenital fusion. Each staminode has a Rosidae-Asteridae lineage, which is now associated with stalk inserted at the rim of the tube and a sagittate head specialized insect pollinators. The exceptional quality of shaped like an anther connective (Fig. 1 C, E, F, and K). preservation makes it possible to detect minute, but highly Abaxial surface cells of the expanded staminode head are significant, structural features that typically would not be larger than those ofthe stalk (Fig. iF). Adaxially, staminodes found even in very well-preserved fossils. are expanded into hemispherical protuberances composed of The majority of fossils at the Raritan site are charcoalified large thin-walled cells that project downward into the stam- and three-dimensional (not noticeably compressed), with inal tube (Fig. 1C). Even at high magnification with scanning moderate to extraordinary preservation of details such as electron microscopy, there is no evidence of gynoecia or of pistillodes at the bases of the staminal tubes in any of the The publication costs of this article were defrayed in part by page charge fossils examined; therefore we interpret these as completely payment. This article must therefore be hereby marked "advertisement" unisexual inflorescences. Staminodes are routinely observed in accordance with 18 U.S.C. §1734 solely to indicate this fact. in fossil flowers that appear to have been fully mature at the 8986 Evolution: Crepet et al. Proc. Natl. Acad. Sci. USA 89 (1992) 8987

FIG. 1. Scanning electron micrographs of fossil inflorescences, flowers, and organs. (A) Capitate staminate inflorescence showing florets, sepal cup lobes (sc), Each flower with four triangular stamen connectives (c), visible within sepal cup. (x23.) (B) One floret from a staminate inflorescence showing slightly separated anther connectives (c) opposite sepal cup lobes (sc). (x60.) (C) A lateral view of a partially broken staminal tube (st), within a sepal cup (sc), illustrating the thin-walled nature of the staminal tube, its hollow interior, and apical staminodes with adaxially bulbous extremities (bs). (x70.) (D) Floret with anthers removed, showing sepal cup and four sagittate staminode heads. (x55.) (E) Floret with two whole stamens removed revealing the incurved, contiguous staminodes (s) and one anther (a) sheared off illustrating cellular preservation. (x 70.) (F) Closer view of distal staminode showing sagittate head (arrow), similar to shape offertile stamens. (x 140.) (G) Stamen removed from a floret showing the sagittate distal connective and thecae with slightly open valves. (x60.) (H) A single in situ pollen grain. (x4000.) (I) Pistillate inflorescence showing the somewhat abraded carpels and sepal lobes. (x 34.) (J) Distal adaxial stamen connective epidermis showing three anomocytic stomata-one with broken guard cell walls. (x560.) (K) Close-up of a staminate floret with stamens removed to show the partially abraded sagittate heads of two staminodes (sh) and the broken stamen filament (sf) that appears to be interior to the two staminodal stalks. Note the broken sepal cup (arrow) for perspective. (x140.) (L) One tangentially sheared pistillate floret showing the bicarpellate syncarpous ovary inside of a sepal cup. (x105.) time of preservation and were past anthesis judging from (Fig. 1F), sagittate head (Fig. 1 F vs. G), and incurvature. pollen maturity and dehiscence of anther sacs. The stamin- Large surface cell size and association with a well-developed odal nature of these appendages is also suggested by their staminal tube and sepal cup are consistent with a nectarifer- morphological similarity to stamens, particularly in the stalk ous function for the staminodes. 8988 Evolution: Crepet et al. Proc. Natl. Acad. Sci. USA 89 (1992) Pollen (Fig. 1H) was found in situ in attached and detached pollen sculpture, are not found in combination in any single anthers. Pollen grains are relatively small (12 ,um x 9 ,tm), genus of modem Hamamelidaceae. Given the relationship moderately prolate (polar/equatorial ratio = 1.33), and con- between Platanaceae and Hamamelidaceae, these characters sistently tricolpate. Colpi extend almost the entire polar are best interpreted as retained primitive platanoid features in length ofthe grains. There is no indication ofan endoaperture the fossil taxon (a more extensive discussion of detailed in any of the grains examined with light microscopy (whole phylogenetic context will be provided in a specialty journal grains or 1- to 2-pum sections) or scanning electron micros- article that also formally names the fossil taxon). Certain of copy. Pollen wall structure is reticulate with lumina varying the platanoid features found in the fossils, then, have been considerably in size. subsequently lost or further modified in the various clades Pistillate inflorescences are small (to ca. 2.7 mm in diam- represented by modem genera ofthe Hamamelidaceae. Con- eter), capitate heads of sessile flowers that are similar in size sideration ofthe character mosaic ofthe fossils will undoubt- and gross morphology to staminate inflorescences (Fig. 1!). edly force a reevaluation of current thinking on the primitive Based on floral structure and local cooccurence of pistillate floral type in Hamamelidaceae and the relationships of the and staminate inflorescences, as well as on the lack of other subgroups within the family. appropriate pistillate fossils, we interpret these to represent Based primarily on structure, we interpret the strap-shaped the same taxon as the staminate fossils. In the available petals in modem Hamamelidaceae (e.g., Hamamelis) as material, pistillate inflorescences are not as well preserved as homologous to the fossil staminodes. The fossil staminodes staminate inflorescences. All pistillate fossils appear to be are morphologically intermediate between functional sta- physically abraded to some degree, precluding determination mens and modem strap-shaped petals of Hamamelidaceae. of whether stamens or staminodes might have been present Petals of modem Hamamelidaceae (e.g., Hamamelis) often in younger flowers. However, based on the evident unisex- have a similar incurved vernation and, at least in some uality of staminate flowers, it is likely that the pistil-bearing genera, have basal adaxial nectaries (4, 6-8). The fossil flowers were also unisexual or functionally unisexual. In staminodes also appear to occupy the same position in the some inflorescences, what appear to be syncarpous pistils are flower as the petals of modem taxa, but, as we have already partially inferior and included in a four-lobed sepal cup noted, tight juxtaposition of the whorls makes it almost similar to that seen in staminate fossils. Sometimes pistillate impossible to verify this impression. Because Platanaceae or inflorescences are preserved with the flowers sheared off any other reasonable outgroup to Hamamelidaceae are apet- tangentially on a plane through the ovary, revealing two alous, and the fossils retain many platanoid features, it is not locules and an axile placenta, and illustrating the excellent, parsimonious to interpret staminodes as derived phylogenet- cell by cell quality of preservation (Fig. 1L). The ovary wall ically from petals. Instead, the apparent structural and po- is five to eight cells thick. sitional homology between the apparently nectariferous sta- Fossil inflorescences exhibit a mosaic of characters found minodes ofthe fossils and nectariferous petals of the modem in modem Hamamelidaceae and Platanaceae. In addition to Hamamelidaceae may be indicative ofan intermediate step in the form and unisexuality of the inflorescences, the reticu- the character transformation from apetalous to petalous late-tricolpate pollen removed from in situ stamens, although flowers in the hamamelidid-rosid lineage. Under this inter- generalized, is suggestive of modern Platanaceae, showing pretation, the petals of hamamelidaceous and probably most somewhat less similarity to pollen of modem members ofthe rosid/asterid taxa are phylogenetically derived from a whorl family Hamamelidaceae (4, 5). Stamens are tightly packed in of staminodes such as those found in the fossil flowers. This the fossil flowers, and the four broad triangular apical con- fossil evidence is compatible with the neontologically based nectives virtually fill the floral envelopes (Fig. 1 A and B), as hypotheses of Walker and Walker (14) on petal origins in the in modem and fossil Platanaceae (4). Other floral features, hamamelid-rosid-asterid clade. including the consistent presence of a well-developed sepal The complex of nectariferous staminodes and sepal cup, cup, staminal tube, and a cycle of staminodes that were suggestive of insect pollination, in an apetalous flower is a apparently nectariferous, are typical of some modem Ha- plausible precursor to flowers in which the staminodes have mamelidaceae (6) and various Rosidae. Also typical of mod- been further modified and serve the dual functions of nectar em Hamamelidaceae and Rosidae are the syncarpous/ production and pollinator attraction, similar to the petaloid bicarpellate flowers ofthepistillate inflorescences in contrast appendages in the modem genus Hamamelis (6). It is likely to the apocarpous flowers of Platanaceae (4, 6, 7). that such a transition from nectar-bearing staminodes to more Because the fossil taxon possesses derived features found complex petals occurred earlier in the Cretaceous, because in Hamamelidaceae that are not found in Platanaceae, we rosid flowers with petals (representing taxa that are nested consider it to be representative of an extinct "hamamelida- within the hamamelidid lineage) of slightly greater age have ceous" taxon. The specific combination of features in the been described by Basinger and Dilcher (15). The origin of fossils (capitate spheroidal inflorescences, stamens with ex- petals in the hamamelidid-rosid-asterid lineage is of great panded connectives, staminal tube, nectariferous stami- interest because ofthe strong relationship between rosid taxa nodes, tricolpate reticulate pollen, functional unisexuality, a and insect pollinators. Also significant, in this context, is the well-developed sepal cup, and a syncarpous bicarpellate phylogenetic origin of the hypanthium-nectary complex em- ovary) is not found in any extant hamamelidaceous genus. bodied by these fossils. Although the Cenomanian rosid Traditionally, Platanaceae have been regarded as closely flower (15) apparently has nectaries, the preservation of that related to Hamamelidaceae (8). Recent, cladistic analyses (9) fossil is not sufficient to determine if they are of staminodal further suggest that Hamamelidaceae are a sister group of origin. It also does not provide insight into the nature and Platanaceae, which have an extensive early and mid Creta- origin of petals in the rosid-asterid lineage because its floral ceous fossil record (4, 10-12). In addition, numerous features parts are apparently already well-differentiated. These two of Hamamelidaceae support the conclusion that the modem fossils taken together do provide insight into the early diver- family has a basal relationship to the higher hamamelidids gence of this clade and suggest that the sepal cup/nectary/ (oaks, walnuts, birches, Casuarina, etc.) as well as to the petal complex may have had an even earlier origin. entire rosid/asterid clade (13), one of the most speciose and The occurrence of petal homologues and a nectar cup sets diverse groups ofangiosperms. Some features that the fossils a minimum age on one of the most important steps in the share with Platanaceae, such as functionally unisexual flow- history of biotic pollination-the relationship between spe- ers, capitate spheroidal inflorescences, tightly packed sta- cialized insect pollinators and angiosperms. This has been mens with expanded contiguous connectives, and platanoid invoked often as a partial explanation for the spectacular Evolution: Crepet et al. Proc. Natl. Acad. Sci. USA 89 (1992) 8989 diversity of the flowering plants relative to other vascular We thank Peter R. Crane (Department of Geology, Field Museum plants (16-21). The sepal cup and/or hypanthium, a feature of Natural History, Chicago), David L. Dilcher (Florida State found throughout the vast majority of families of the sub- Natural History Museum, Gainesville), and Peter K. Endress (In- classes Rosidae and Asteridae, probably first evolved in an stitut fur Systematische Botanik der Universitat, Zurich) for their extinct group close to Hamamelidaceae. The combination of helpful comments and suggestions. Field work was supported by the sepal cup and staminal tube in these fossils suggests a National Science Foundation. Laboratory work and scanning elec- plausible sequence of evolutionary steps leading to the origin tron microscopy fees were supported by the College of Agriculture ofthe hypanthium-fusion of the staminal tube and sepal cup and Life Sciences, Cornell University. being the next logical and, given the equivalence of stamin- 1. Harland, W. B., Armstrong, R. L., Cox, A. V., Craig, L. E., odes and petals, final step in the evolution ofthe hypanthium. Smith, A. G. & Smith, D. G. (1989) A Geologic Time Scale These fossils represent a taxon that is phylogenetically near (Cambridge Univ. Press, Cambridge, U.K.). the base of the hamamelidid-rosid-asterid clade (to be re- 2. Brenner, G. J. (1963) Md. Dept. Geol. Mines Water Resour. ported elsewhere), one of the most diverse angiosperm Bull. 27, 1-215. lineages and one in which the hypanthium-nectary character 3. Doyle, J. A. & Robbins, E. I. (1977) Palynology 1, 43-78. complex is of fundamental significance in the relationship 4. Friis, E. M., Crane, P. R. & Pedersen, K. R. (1988) Biol. Skrift. with hymenopteran pollinators. 31, 1-55. Apidae (bees), the most significant pollinating Hy- 5. Bogle, A. L. & Philbrick, C. T. (1980) Contrib. Gray Herb. 210, menoptera, have an earlier origin than previously believed, 29-103. based on direct fossil evidence from the Late Cretaceous (22) 6. Endress, P. K. (1977) Plant Syst. Evol. 1, 321-347. 7. Endress, P. K. (1989) Plant Syst. Evol. 162, 193-211. as well as on inference from the fossil records of other 8. Cronquist, A. (1981) An Integrated System ofClassification of Hymenoptera and flowers (19, 20, 23, 24). The Turonian Flowering Plants (Columbia Univ. Press, New York). occurrence of the sepal cup-staminal tube-nectary complex 9. Hufford, L. D. & Crane, P. R. (1989) in Evolution, Systematics is compatible with an earlier origin of bees and is consistent and Fossil History of the Hamamelidae, eds. Crane, P. R. & with the evidence for radiation of derived flies (25) that may Blackmore, S. (Oxford Univ. Press, Oxford), pp. 175-192. also have been significant early pollinators adapted to a 10. Endress, P. K. (1987) Trends Evol. 2, 300-304. hypanthium-nectary complex (26). 11. Friis, E. M. & Endress, P. K. (1990) Adv. Bot. Res. 17, 99-162. The tiny size of the flowers, one of the most distinctive 12. Dilcher, D. L. (1979) Rev. Palaeobot. Palynol. 27, 291-328. features of these fossils, cannot escape consideration in the 13. Dickison, W. C. (1989) in Evolution, Systematics and Fossil context of pollination biology. The fossilized flowers are History ofthe Hamamelidae, eds. Crane, P. R. & Blackmore, significantly smaller than any modem taxa of Hamameli- S. (Oxford Univ. Press, Oxford), pp. 47-74. daceae or Platanaceae. Overall flower size suggests that 14. Walker, J. W. & Walker, A. G. (1984) Ann. Mo. Bot. Gard. 71, 464-521. effective pollinators were exceptionally small. In modem 15. Basinger, J. F. & Dilcher, D. L. (1984) Science 224, 511-513. taxa associated with specialized hymenopteran pollinators, 16. Regal, P. J. (1977) Science 196, 622-629. reservation of pollen and nectar rewards for appropriate 17. Mulcahy, D. L. (1979) Science 206, 20-23. pollinators is an important aspect of the pollination syn- 18. Burger, W. C. (1981) BioScience 31, 572-581. drome. Long perianth/tepal tubes, complex morphology, 19. Crepet, W. L. & Friis, E. M. (1987) The Origins of An- and hidden rewards are characteristic oftaxa with specialized giosperms and Their Biological Consequences, eds. Friis, insect pollinators today (26). In these tiny fossil flowers, E. M., Chaloner, W. G. & Crane, P. R. (Cambridge Univ. exposed nectar and pollen would have been accessible to Press, Cambridge, U.K.), pp. 180-201. almost any insect. It is likely that the contiguous incurved 20. Friis, E. M. & Crepet, W. L. (1987) The Origins of An- anther connectives served to limit access to floral rewards to giosperms and Their Biological Consequences, eds. Friis, E. M., Chaloner, W. G. & Crane, P. R. (Cambridge Univ. specific pollinators that had small enough mouthparts to Press, Cambridge, U.K.), pp. 145-179. insert between the anthers (a protective function has also 21. Crepet, W. L., Friis, E. M. & Nixon, K. C. (1991) Philos. been suggested for such connectives in Platanaceae; ref. 27). Trans. R. Soc. London B 333, 187-195. Pollen size might also reflect size limits on pollen ingestion by 22. Michener, C. D. & Grimaldi, D. A. (1988) Am. Mus. Novit. these early pollinators. 2498, 1-10. Finally, these fossil inflorescences are significant because 23. Carpenter, J. M. & Rasnitsyn, A. P. (1990) Psyche 97, 1-20. they constitute the earliest fossil flowers with close affinities 24. Jarzembowski, E. A. (1984) Mod. Geol. 9, 71-93. to modern Hamamelidaceae (28). In the Raritan deposits, 25. Grimaldi, D. A. (1990) Bull. Am. Mus. Nat. Hist. 195, 164-183. there is an abundance of platanoid leaf compressions but no 26. Procter, M. & Yeo, P. (1973) The Pollination of Flowers (Taplinger, New York). reports of leaf compressions similar to modem Hamameli- 27. Endress, P. K. (1986) Bot. Jahrb. Syst. 96, 1-44. daceae. In conjunction with the mosaic of platanoid and 28. Endress, P. K. & Friis, E. M. (1991) Plant Syst. Evol. 175, hamamelidaceous floral and pollen features, this suggests 101-114. that the well-known platanoid leaf complex of the Late 29. Doyle, J. A. & Hickey, L. J. (1976) Origin and Early Evolution Cretaceous (29) might include cryptic or unrecognized taxa ofAngiosperms, ed. Beck, C. B. (Columbia Univ. Press, New more closely related to modern Hamamelidaceae. York), pp. 139-206.