The complexity of ovulate cones and their preservation as fossils makes them promising material for reconstructing the evolutionary history of gymnosperms, but phylogenetic analyses of cone morphological characters of Pinaceae have been inconclusive. We describe the ovulate cone anatomy of Pinus nelsonii, a rare and phylogenetically isolated pinyon pine endemic to Mexico, and add the species together with Pinus ponderosa and the fossil Pinus belgica to a recoded and expanded ovulate cone morphology matrix for fossil and extant Pinaceae. The cone anatomy of Pinus nelsonii conforms to previous generic concepts of Pinus. Despite its phylogenetically isolated position among the soft pines (Pinus subgenus Strobus) and thus potential for displaying plesiomorphic features, the cone of Pinus nelsonii is unlike the oldest Pinus fossil cones in possessing enlarged, functionally wingless seeds partially embedded in scale tissue, and in lacking sclerenchyma in the cortex of the axis, in the bract, and in the scale. Cladistic analysis of cone morphology characters recovers several Pityostrobus species in a clade with Pinus. Although the inferred relationships among living species do not coincide in several respects to molecular studies, adding taxa and further exploration of characters promise to clarify relationships.
Anatomically preserved fossil ovulate cones from the Cretaceous and Cenozoic have been extremely influential in understanding past Pinaceae diversity (Miller 1976a; Smith and Stockey 2001, 2002). Most pinaceous fossil cones from the Cretaceous are classified in the extinct organ genera Obirastrobus Ohsawa, Nishida & Nishida (two described species), Pityostrobus Nathorst emend. Dutt (a polyphyletic assemblage of at least 28 anatomically described species), and Pseudoaraucaria Fliehe (six species; Smith and Stockey 2002). Phylogenetic analyses of ovulate cone morphology characters suggest that these fossils are in the Pinaceae crown, but either few characters have been included (Alvin 1988; Shang et al. 2001) or strict consensus trees have been poorly resolved (Smith and Stockey 2001, 2002). Several strategies could be used to increase phylogenetic resolution. First, character coding decisions have an important influence on the results of phylogenetic analysis (Pleijel 1995; Hawkins 2000), and these decisions require further evaluation. Second, more fossil and extant taxa are available for inclusion in morphological analyses. Regarding extant pinaceous taxa, most have been treated as generic-level terminals, and consequently many characters have been scored as polymorphic (Alvin 1988; Smith and Stockey 2001, 2002; Gernandt et al. 2008). Dividing taxa with polymorphic character states into smaller units with monomorphic states (in this case, using species instead of genera) can reduce error in phylogenetic reconstructions (Nixon and Davis 1991). Third, adding characters can improve phylogenetic accuracy even when this requires including more missing data for some taxa (Wiens 2003). Fourth, through reciprocal illumination (Hennig 1966), hypotheses of homology based on morphological similarity can be assessed after phylogenetic analysis via congruence with other characters on a tree (Patterson 1982), and homoplasious characters can be identified for further study.
Pinus L. is by far the largest and most phylogenetically diverse genus in Pinaceae (Farjon 2005; Eckenwalder 2009). It is the only extant genus thought to be represented by anatomically preserved ovulate cones from the Cretaceous (Alvin 1960; Miller 1976a), and many ovulate cones from the Cretaceous assigned to the organ-genus Pityostrobus strongly resemble pines (Miller 1976a). Pinus and Pinaceae are of outstanding ecological and economic importance and relatively detailed morphological descriptions are available for all extant species (e.g. Farjon 2005; Eckenwalder 2009). Furthermore, molecular systematic studies have clarified relationships among extant genera and species (Wang et al. 2000; Gernandt et al. 2008). However, our knowledge of the anatomical variation in living species of Pinus is incomplete. Miller (1976a) proposed generic concepts for living and fossil pinaceous cones based in part on the earlier work on living conifer species by Radais (1894) and in part on his reference collection, which at the time included 65 species from nine of the 11 Pinaceae genera (sensu Farjon 2005; Eckenwalder 2009). Anatomical descriptions for ovulate cones of most living species either do not exist or are unpublished. In contrast, anatomical descriptions are available for most fossil cones assigned to Obirastrobus, Pityostrobus, and Pseudoaraucaria (Miller 1976a; Smith and Stockey 2001, 2002).
Molecular phylogenetic studies strongly indicate that Pinus includes two monotypic subsections, both in Pinus subgenus Strobus (the soft pines): Pinus subsection Krempfianae is represented by P. krempfii Lecomte, endemic to central Vietnam, and Pinus subsection Nelsoniae is represented by P. nelsonii Shaw, a rare pinyon pine endemic to high desert woodlands of the Sierra Madre Oriental of Mexico (Fig. 1A; Gernandt et al. 2001; Syring et al. 2005; Parks et al. 2009). Neither species was available to Miller (1976a) when he proposed diagnostic features that fossil cones must possess to be classified in Pinus rather than Pityostrobus. As a result, the cone morphologies of two major lineages remained unknown. The present study's objectives are to describe the ovulate cone anatomy of Pinus nelsonii and to explore additions and modifications to an existing cone morphology matrix (Smith and Stockey 2001, 2002) that could clarify relationships among extant and fossil taxa. We split the generic-level terminal “Pinus“ into three species-level terminals, divided several multistate characters into multiple characters, and added several external morphological characters for ovuliferous scales, bracts, and seeds.
MATERIALS AND METHODS
Branch, leaf, and cone collections of Pinus nelsonii were made from mature trees from a natural population in the municipio of Guadalcázar, San Luis Potosí on August 17, 2004 and December 13, 2009 (DSG Nos. 499 and 1095, vouchers deposited at MEXU). Two ovulate cones, one mature and the other immature, were rehydrated by soaking for 48–72 hrs in water. Scales from the basal, medial, and apical parts of the axis were removed from the mature cone, and a double-edged razor was used to make transverse and longitudinal sections of the immature cone. The material was soaked for one wk in 10% aqueous ethylene diamine at room temperature (Carlquist 1982), dehydrated in serial rinses of 70–100% ethanol, embedded in paraffin, and sectioned on a rotary microtome at 10–14 µm. Sections were stained with safranin and alcian blue (Ruzin 1999), cleared with CitriSolv (Fisher Scientific, Suwanee, Georgia), and mounted with Entellan rapid mounting medium (Merck, Whitehouse Station, New Jersey) for observation under a compound microscope at 5–100 × magnification. Means and standard deviations were calculated based on 25 measurements of tracheids and rays from serial sections of the mature cone. Images were captured with a Leica Firecam and stitched in Photoshop CS3 with the photomerge option. Original unmerged images are available from the first author upon request.
The cone morphology matrix of Smith and Stockey (2002) was reconstructed in Mesquite (Maddison and Maddison 2009), modified, and deposited in TreeBASE (study number 10553). The matrix included 50 taxa (Appendix 1). The outgroups used by Smith and Stockey (2001, 2002) were retained: two extant conifers, Cryptomeria japonica and Sciadopitys verticillata, and the fossil taxon Pararaucaria patagonica, the latter usually hypothesized as belonging to a conifer family outside Pinaceae (Stockey 1977). The taxon “Pinus”, for which seven of thirty-tree characters were scored as polymorphic by Smith and Stockey (2002), was replaced with the extant species Pinus nelsonii and Pinus ponderosa (Pinus subgenus Pinus, a hard pine), and the fossil species Pinus belgica. Character states for P. nelsonii were from this study, those for P. ponderosa were based on observations from the reference collection of C. N. Miller, currently housed at the University of Alberta, and those for P. belgica were from Alvin (1960).
The character matrix was expanded and modified from Smith and Stockey (2002; Appendix 2). Some multistate characters from Smith and Stockey (2002) included the state “absent” together with alternative character forms (e.g. resin canal position in the scale). We separated these into two characters, one specifying the structure's presence or absence, and the other specifying its alternative forms (Hawkins et al. 1997; Sereno 2007). Those taxa scored as absent for the first character state were scored as inapplicable for the second. The parsimony searches did not treat inapplicable states differently than missing states; we distinguished between inapplicable and missing characters to make the matrix scoring clearer. Also, Smith and Stockey (2002) coded the distribution of resin canals and sclerenchyma in the scale with the states “abaxial,” “adaxial,” “between,” and “both abaxial and adaxial.” The alternative states can be found in the same taxa and implies that they are not homologous (conjunction coding; Patterson 1982). We separated these multistate characters into separate presence or absence characters (characters 26–36). This effectively divided one of Miller's (1976a) features used to delimit Pinus and Pityostrobus (resin canals at the scale base restricted to the abaxial side of the vascular tissue) into two characters.
Nine characters were added to the matrix: bract apex shape (character 18; Alvin 1988), vascular trace transverse shape in the ovuliferous scale distal to the seed (character 24; Alvin 1988), ovuliferous scale shape in surface view (character 37; Frankis 1989; see also Alvin 1988; Mente and Brack-Hanes 1992; Farjon 2005), shape of the ovulate scale base (character 38; Frankis 1989; Farjon 2005), ovuliferous scale umbo presence or absence (character 40; treated as separate from ovuliferous scales thickening distally; Smith and Stockey 2001, 2002), ovuliferous scale umbo position (character 41), seed body shape in surface view (character 51; Frankis 1989), presence or absence of a vascular trace entering the seed (character 52; Alvin 1988), and nature of seed wing attachment (character 54; Frankis 1989; see also Alvin 1988). Five multistate characters represented counts or measures and were treated as ordered (Appendix 2). Newly included characters for fossil cones and the outgroups were scored from published descriptions (Radais 1894; Dutt 1916; Alvin 1953, 1957a, 1957b, 1960; Creber 1956, 1960, 1967; Miller 1972, 1974, 1976a, 1976b, 1977, 1978, 1985; Miller and Robison 1975; Robison and Miller 1977; Stockey 1977, 1981; Crabtree and Miller 1989; Takaso and Tomlinson 1989, 1991; Ohsawa et al. 1991, Ohsawa and Nishida 1992; Falder et al. 1998; Ratzel et al. 2001; Smith and Stockey 2001, 2002).
The matrix included 54 parsimony-informative characters (Appendix 2) for a total of 2,700 cells, of which 219 cells (8.1%) were scored as missing, 46 (1.7%) as polymorphic, and 65 (2.4%) as inapplicable. The extant ingroup taxa with the most missing, polymorphic, and inapplicable characters were Picea (nine polymorphic and one inapplicable), Cathaya argyrophylla (eight missing and one inapplicable), Larix (eight polymorphic and one inapplicable), and Abies (seven polymorphic and two inapplicable). Fossil ingroup taxa with the most missing and inapplicable characters were Pseudoaraucaria benstedi (18 missing), Pityostrobus shastaensis (12 missing), Pityostrobus cliffwoodensis (nine missing), and Pityostrobus macrocephalus (11 missing and one inapplicable). For the outgroups, Pararaucaria patagonica had five characters scored as missing, thirteen scored as inapplicable, and one as polymorphic, and Cryptomeria japonica had four characters scored as missing and eleven scored as inapplicable.
The matrix was analyzed in PAUP* (Swofford 2002) with equally weighted parsimony. Characters scored as polymorphic for a taxon were treated as polymorphic rather than uncertain (pset /mstaxa = polymorph). A heuristic search was performed with 100,000 random addition sequence replicates, tree-bisection reconnection (TBR) branch swapping, and saving 25 trees per replicate. The remaining search parameters were left on default. Branch support was estimated with 1,000 bootstrap replicates of a heuristic search with 50 random sequence addition replicates, TBR branch swapping, and saving 10 trees per random sequence addition replicate. Ancestral character states were reconstructed with parsimony in Mesquite.
General Features—The external morphology of the ovulate cones conformed to previous descriptions given for the species (Farjon and Styles 1997). Mature cones were cylindrical, measured 5.3 × 3.6–10.8 × 5.0 cm and were supported on elongated recurved peduncles 0.5–1.0 cm in diameter and up to 4.1 cm in length (Fig. 1B). Ovuliferous scales were helically arranged, ca. 70–80 per cone and oriented approximately at a right angle to the cone axis. Scales separated slightly from one another to expose the seeds; they did not spread widely as is typical for the genus.
Cone Peduncle—Transverse sections of the peduncle were 8.5–9.3 mm in diameter. The pith was 2.0 ± 0.2 mm in diameter and composed of isodiametric thick walled parenchyma cells; scattered idioblasts with dark organic contents were also present. Intercellular spaces were visible (Figs. 1C and 1D).
The vascular cylinder was 1.6 ± 0.1 mm thick (70–100 cells wide) from pith to cambium. The mature cone had three faint growth increments marked by a decrease in late wood tracheid radial dimensions but without an increase in wall thickness (Fig. 1E). Forty-four axial resin canals were present in transverse section, most occurring as two irregular rings in the early wood of the first growth increment, usually small (0.06 mm), and surrounded by approximately 7–10 thin-walled epithelial cells. No axial parenchyma was observed. Tracheids were 19.6 ± 2.1 µm in diameter and 840 ± 160 µm in height. The tracheid walls showed a transition from primary xylem with helical thickenings to metaxylem and secondary xylem with circular bordered pits with tori. The circular bordered pits were predominantly uniseriate (Fig. 1F), but occasionally biseriate and opposite.
Most xylem rays were uniseriate, but biseriate rays (with a radial resin canal) were also present (Fig. 1G). Uniseriate rays were 7.25 ± 6.71 (1–30) cells and 169 ± 155 µm tall and 16.3 ± 3.7 µm wide. Biseriate rays were 12.1 — 2.8 cells and 263 — 76 µm tall and 43 ± 9.0 µm wide. Rays were homocellular with pitted transverse walls. The cross-fields of tracheids with ray cells typically had two to four (much less frequently one or five) piceoid pits per cross-field (Fig. 1H).
The cambial-phloem zone was 0.25–0.6 mm (15 cells) thick with exclusively nonlignified cells. The cortex was 1.5 — 0.4 mm thick, composed of irregularly shaped parenchyma cells with walls of varying thickness. Distributed throughout the cortex were 52 axial resin canals varying in diameter from 0.06 × 0.06 mm to 0.8 × 0.4 mm and surrounded by 12–17 thin-walled epithelial cells (Fig. 1C). Two circular vascular bundles occurred in the cortex. Some cells in the cortex contained dark organic contents, and some of these had prismatic crystals (Fig. 1I). The phelloderm consisted of two to three rows of rectangular, sclerified cells, and exterior to them, a phellem of eight to 11 cell layers (Fig. 1C).
Cone Axis—The cone axis was 9.6 cm long, 8.5–10.0 mm in diameter at its base, slightly wider medially, and tapering apically. Near the cone base, the pith was 2.1 ± 0.1 mm in diameter and composed principally of isodiametric thick-walled parenchyma cells, occasionally with dark organic contents, crystals, or both (Fig. 2A). Scattered sclerenchyma and idioblasts were also present; resin canals were absent. The pith increased to 2.9 ± 0.1 mm at the medial part of the cone.
The vascular cylinder (Fig. 2B) was 1.3 ± 0.1 mm thick at the base and successively thinner apically (1.1 mm at the medial part of the cone). The vascular cylinder was continuous except where vasculature led to bract-scale complexes (Fig. 2C). Three faint growth increments were present in the xylem (visible at the axis base), delimited by two indistinct rings of resin canals. Each canal was surrounded by approximately 7–10 thin-walled epithelial cells (Fig. 2B). Tracheids were 15.7 ± 3.6 µm in diameter, with uniseriate and occasionally biseriate and opposite circular bordered pits (Fig. 2D). Uniseriate rays were 3.32 ± 4.16 (one to 17) cells and 90.8 ± 114 µm tall and 15.8 ± 4.8 µm wide. Rays were homocellular. Ray cells had walls with large pits (Fig. 2E). Near the primary xylem both procumbent and upright ray cells were present, giving way to procumbent cells in the later-formed secondary xylem. Two to four (rarely one or five) pits occurred in the cross-fields (Fig. 2F). These pits were interpreted as pinoid because the pit apertures lacked wide borders (IAWA Committee 2004).
The cortex was of uneven width, interrupted by vascular traces; midway between the cone base and apex its maximum width was 2.07 ± 0.12 mm (54 ± 5.3 cells; Fig. 2A). It was composed of parenchyma cells and resin canals. No sclerenchyma was observed. The resin canals decreased in number apically (there were 26 canals midway from the base to the apex), varied in diameter from 0.08–0.24 mm, and were surrounded by one to several layers of thin-walled epithelial cells. There were 12–17 cells in the innermost layer.
Cone Scale Complex—Vascular traces leading to the bractscale complex were cylindrical, fused at their origin in the inner cortex of the axis, and contained small resin canals (Fig. 2G). In the outer cortex, the traces separated into a scale and bract trace that were abaxially concave and circular in transverse section, respectively; these were accompanied abaxially by two resin canals (Fig. 2H). The bract and scale remained fused a short distance beyond the cone axis and the single scale trace diverged into three separate traces; a row of resin canals was abaxial to the scale traces and two resin canals were lateral to the bract trace (Fig. 2I). The scale and subtending bract separated first at their lateral margins.
The bracts were minute, 1.5–2.5 mm long, and 2.4–3.7 mm wide × 0.5 mm tall at their base (Fig. 2I). The bract apex was thin with a broad, round to slightly pointed margin. Bract cells were parenchymatous. A vascular trace entered the center of the bract and extended for more than half the bract length. Two small resin canals extended into the free part of the bract on both sides of the trace.
The ovuliferous scales were elongate, cuneate to obovate in surface view, and terminated in a prominent apophysis with a transversely keeled dorsal umbo and a minute mucro that was worn almost flat. The basal and apical scales were small and sterile, while those from the medial part of the cone measured ca. 1.8–2.5 cm long and 1.3–2.2 cm wide at their widest point near the apex. Scales bore two seeds in depressions on their adaxial surface separated by an overarching interseminal ridge. One seed was aborted in some scales.
Most cells in the scales were parenchyma. Near the scale base the vascular traces divided, forming a row near the adaxial surface (Fig. 3A). At seed level, the vascular traces were visibly curved on the phloem side (abaxially concave) in transverse section (Fig. 3B) and remained near the adaxial surface (Fig. 3C). Distal to the seeds, vascular traces were arranged in a row midway between the abaxial and adaxial surface and were more prominently curved (Fig. 3D). Resin canals were abaxial to (and inside) the traces at the scale base. At seed level they also occurred between, and adaxial to the vascular bundles; several also occurred in a prominent medial ridge that overarched the two seed cavities (Fig. 3C). Distally resin canals were distributed throughout the scale tissue, with larger canals near the lateral margins (Fig. 3D).
Seeds were obovoid, 10.5–14.8 mm long and 6.8–9.1 mm wide and surrounded by a thick (1.0–1.3 mm) seed coat. Effective seed wings were absent, but vestigial wing tissue surrounded the distal end and lateral sides of the seed in the position corresponding to wing attachment in other Pinus species. This tissue usually remained attached to the scale upon removing the seed, forming a 2–3 mm ridge around the seed cavity. Embryos were 8–9 mm long with 11–14 cotyledons.
Phylogenetic Analysis—The heuristic search recovered a single most parsimonious tree with a length (L) of 306 steps, a consistency index for informative characters (CI) of 0.37 and a retention index (RI) of 0.60. The shortest tree length was found in 11 of the 100,000 random addition sequence replicates. Only seven branches received bootstrap support greater than 50% (Fig. 4). Outgroup rooting with Sciadopitys and Cryptomeria recovered Pararaucaria (of uncertain familial affinity) within Pinaceae. The root was located between Larix + Pseudotsuga and the rest of Pinaceae. Pinaceae also included an “Abietoid” clade comprised of the extant genera Abies, Cedrus, Keteleeria, Pseudolarix, and Tsuga together with Pararaucaria, Pseudoaraucaria, and nine Pityostrobus species. A second “Pinoid” clade comprised the extant genera Cathaya, Picea, and Pinus together with 18 Pityostrobus species. All six species of the fossil genus Pseudoaraucaria were recovered as monophyletic and sister to Abies. Both species of Obirastrobus also formed a clade. The three Pinus species were paraphyletic to ten Pityostrobus species (P. andraei, P. argonnensis, P. bernissartensis, P. hautrageanus, P. jacksonii, P. kayeii, P. lynnii P. macrocephalus, P. palmeri, and P. shastaensis).
There were six characters with no homoplasy (consistency index = 1) on the most parsimonious tree (Fig. 4). Seed wings formed from sarcotestal tissue (character 47) united the outgroups Sciadopitys and Cryptomeria. The multistate character for bract length (character 16) united Sciadopitys and Cryptomeria (equal to scale length), and Larix and Pseudotsuga (longer than scale length). Bract length was optimized as shorter than scale length in the Abietoid and Pinoid clades, but the character was scored as polymorphic in several genera (both longer and shorter in Abies, Larix, Pseudolarix, and Tsuga). Had these taxa been divided into multiple species terminals the character would have exhibited homoplasy There were two synapomorphies for Pinaceae + Pararaucaria (character 46; seed wings formed from ovuliferous scale tissue and character 51; two seeds per ovuliferous scale, scored as polymorphic in Pararaucaria), a synapomorphy for the six Pseudoaraucaria species (character 50; an enlarged parenchyma pad or cushion present at the chalazal end of the seed), and a multistate character supporting the monophyly of part of the Abietoid clade including Abies and Keleeria, and of Pinus + 14 Pityostrobus species (character 54; seed wing attachment forming deep cups or claws, respectively). There were multiple equally most parsimonious reconstructions for this character; its location in Fig. 4 placed transformations as close as possible to the root. The most homoplasious characters on the trees were primarily anatomical rather than external morphological characters. Those with a consistency index less than 0.2 were presence or absence of bract sclerenchyma (character 19; CI = 0.14), abaxially concave versus straight vascular bundles distally in the ovuliferous scales (character 24; CI = 0.14), presence or absence of resin canals in the ovuliferous scales between the vascular tissue at the level of the seed body (character 29; CI = 0.12), presence or absence of resin canals in the ovuliferous scales adaxial to vascular tissue at the level of the seed body (character 30; CI = 0.14), presence or absence of sclerenchyma in the ovuliferous scales between the vascular tissue (character 35; CI = 0.091), and presence or absence of sclerenchyma in the ovuliferous scales adaxial to the vascular tissue (character 36; CI = 0.11).
Mapping character transformations onto the tree (favoring reversions with accelerated transformation) showed which characters united Pinus with thirteen Pityostrobus species (the node including Pityostrobus cliffwoodensis). There were three character transformations at that branch: bract and ovuliferous scale separating first medially (character 15; CI = 0.22, homoplastic in the same clade), resin canals to ovuliferous scales between vascular tissue abaxial at level of seed (character 29; CI = 0.14, homoplastic in the same clade), and seed wings attached by claws (character 54; CI = 1.0, scored as missing in several Pityostrobus species and with equally parsimonious reconstructions on other branches). The characters used by Miller (1976a) to separate Pinus from Pityostrobus had low consistency indices. The separation or fusion of bract traces at their origin (character 10; CI = 0.50) required four steps on the tree and would have required three more steps to account for polymorphic taxa (Fig. 5A). The fused state (as in Pinus) was inferred as derived seven times and lost once. The transverse shape of the scale vascular scale bundles distal to the seeds (character 24; CI = 0.14) required seven steps on the tree, with the abaxially concave state (as in Pinus) inferred as derived four times and lost three times (Fig. 5B). The presence or absence of resin canals at the scale base abaxial to the vascular traces (character 26; CI = 0.25) required four steps on the tree, with presence inferred as derived twice and lost twice (Fig. 5C). The presence or absence of resin canals at the scale base adaxial to the vascular traces (character 27; CI = 0.20) required five steps on the tree, with presence inferred as ancestral, being lost four times and regained once (Fig. 5D). The scale apex thinning or thickening distally (character 39; CI = 0.20) required five steps on the tree, with the thickening state inferred as derived four times and lost once (Fig. 5E).
We evaluated whether including taxa with many characters scored as missing or inapplicable affected the placement of species for which more information was available. Pararaucaria patagonica and Cryptomeria japonica lacked resin canals, which resulted in 10 characters scored as inapplicable. Deleting these two taxa resulted in a matrix with three variable but parsimony-uninformative characters and 50 parsimony-informative characters. The heuristic search recovered 24 trees (L = 292, CI excluding uninformative characters = 0.37, RI = 0.60). The strict consensus (not shown) was less resolved, with the Abietoid and Pinoid clades collapsing into a polytomy although the groups supported by bootstrap greater than 50% in Fig. 4 were recovered. The position of the root changed, with Pityostrobus pubescens sister to the rest of Pinaceae. We further deleted four ingroup taxa that were described from fossils with abraded scale apices (Pityostrobus cliffwoodensis, Pityostrobus macrocephalus, Pityostrobus shastaensis, and Pseudoaraucaria benstedii). The resulting matrix had three variable but parsimony-uninformative characters and 49 parsimony-informative characters. A heuristic search recovered 32 trees (L = 287, CI excluding uninformative characters = 0.38, RI = 0.59). In the strict consensus (not shown), Larix and Pseudotsuga were recovered in the same clade as Picea, Cathaya, Pinus, and 17 Pityostrobus species, and Pinus was paraphyletic to only two Pityostrobus species (P. hallii and P. villerotensis), although another nine Pityostrobus species occurred with them in a more inclusive clade. Relationships were less resolved among Cedrus, Abies, Keteleeria, Pseudolarix, Tsuga, Pseudoaraucaria, and other Pityostrobus species. The Pinaceae root again was located between Pityostrobus pubescens and the rest of the family.
Diagnostic Characters, Phylogeny, and Natural Groups—Miller (1976a, p. 114) used “four features that are highly characteristic of Pinus” ovulate cone to delimit the genus; these were confirmed as present in P. nelsonii: 1) inflated scale apices (character 39; Fig. 5E), 2) resin canals at the scale base restricted to the abaxial side of the vascular traces (coded as two characters, 26 and 27; Figs. 5C and 5D), 3) abaxially concave traces in the distal part of the scale (character 24; Fig. 5B), and 4) vascular traces to the bract and scale united at their origin (character 10; Fig. 5A). Miller (1976a) proposed that fossil permineralized cones must possess all four features to be included in the genus. The presence of these characters in Pinus nelsonii is consistent with Miller's delimitation of the genus and does not provide additional evidence that the concept of Pinus ovulate cones ought to be expanded to include particular fossils assigned to Pityostrobus.
The consensus tree reported in this study (Fig. 4) is more resolved and more robust (as measured by bootstrap values) than the strict consensus reported by Smith and Stockey (2002). Several of the same groups were recovered; for example the sister relationships among pairs of Pityostrobus species and the Pseudoaraucaria clade. There was also some agreement between our strict consensus tree and their majority-rule tree. In particular, Smith and Stockey (2002) recovered Pinus (as a single terminal) monophyletic with five Pityostrobus species (P. lynnii, P. cliffwoodensis, P. jacksonii, P. kayeii and P. virginiana). Except for Pityostrobus virginiana, the same Pityostrobus species (and others) grouped with Pinus here. The relationships we recovered for extant taxa were also more resolved, but they disagreed in several respects from those recovered in molecular studies (e.g. Wang et al. 2000; Gernandt et al. 2008). For example, Pseudotsuga and Larix are sister to Picea, Cathaya, and Pinus in most molecular studies. The results here suggest which Pityostrobus species will most profitably be included in future analyses of Pinus, and that further progress will be made once more generic-level terminals like Abies and Picea are divided into multiple species-level terminals.
The phylogenetic analysis (Fig. 4) supports Miller's (1976a) assertion that many but not all Pityostrobus species may be more closely related to Pinus than to any other living genus. Some fossil cones assigned to Pityostrobus have a superficial resemblance to Pinus but have been excluded from the genus because they lack one or more of the four ovulate cone features (Miller 1976a; Smith and Stockey 2001, 2002). Miller (1976a) recognized that three of the four features are homoplasious, occurring in phylogenetically disparate extant Pinaceae species. Among extant genera, only inflated scale apices are unique to Pinus (Fig. 5E). In contrast, the cone axis base in the living species Abies lasiocarpa (Hook.) Nutt. and Tsuga canadensis Carrière has bract and scale traces that are fused at their origin as in Pinus (Miller 1976a). Curved traces in the distal part of the scale were reported from four extant genera, Pinus, Larix, Pseudotsuga, and Keteleeria (Alvin 1988; Fig. 5B). Although Pinus and three other genera (Cathaya, Cedrus, and Keteleeria) have resin canals abaxial to the trace at the bractscale origin (Fig. 5C), Pinus is the only extant genus with the combination of abaxial resin canals present and adaxial resin canals absent (Figs. 5C and 5D). Widespread homoplasy in pinaceaous ovulate cone characters suggests that few monophyletic groups in Pinaceae will be characterized by unreversed synapomorphies.
Previous workers have suggested that at least one of the four ovulate cone features diagnostic for Pinus is actually absent in some species of the genus. In particular, Pinus bungeana Zucc. ex Endl. (subsection Gerardianae), and P. albicaulis Engelm. and P. lambertiana Douglas (subsection Strobus) were interpreted as having their vascular traces to the bract and scale separate at their origin (Miller 1992; Saiki 1996). If confirmed, these Pinus species would share the same three of four characters as Pityostrobus andraei, P. argonnensis, and P. kayeii. Pityostrobus argonnensis and a few other species currently assigned to Pityostrobus were originally described as Pinus. Many fossil species have been included in Pityostrobus rather than Pinus because they lack one of these characters, effectively treating the characters as homoplasious in Pityostrobus but not in Pinus. This practice arguably has promoted an artificial classification, and assignments to Pityostrobus have sometimes been made with workers expressing the need for a more natural classification (Alvin 1957b).
Potentially Informative Morphological Characters within Pinus—Variable macromorphological and anatomical characters in ovulate cones of extant and fossil pines are potential synapomorphies for infrageneric groups. A terminal scale umbo is a synapomorphy for Pinus subsection Strobus (Saiki 1996; Gernandt et al. 2005). And although homoplasy explains the shared presence of a prominent interseminal ridge overarching the seeds in Pinus nelsonii and the distantly related fossil genus Pseudoaraucaria, similar enlargement occurs in species of Pinus subsection Cembroides. Many pines rely on animals rather than wind to disperse their seeds, and an enlarged interseminal ridge may prevent the large, functionally wingless seeds from being passively released once the scales separate. The pinyon pine character syndrome, which includes a low habit, and cones with large, functional wingless, bird dispersed seeds, may be responsible for morphological similarities between Pinus nelsonii and its close relatives in Pinus subsection Cembroides. Such pinyon pine characters may have a similar distribution throughout the genus (Gernandt et al. 2005). Although subject to convergence, the characters are informative at lower levels as “local synapomorphies.”
Cortex sclerenchyma has been reported for species in the extant hard pine subsections Pinus, Contortae, and Australes, but it also occurs in the oldest fossil pines (Miller 1976a; Saiki 1996). Other variable characters identified in previous works (Smith and Stockey 2001, 2002) include presence or absence of pith sclerenchyma, the number of growth increments in the secondary xylem of the axis, presence or absence of marked cortex resin canal dilation at branching points to the bracts and scales, presence or absence of bract sclerenchyma, presence or absence of a bract vascular trace, and presence or absence of resin canals between and adaxial to the vascular traces in the scales at the level of the seeds. In general, these anatomical characters were found to be homoplasious, but further taxonomic sampling may reveal that they are informative at less inclusive levels.
Resin canals occur within the scale vascular traces in P. nelsonii and some other Pinus species. Two other species in subgenus Strobus, Pinus monophylla Torr, and Frém. and P. bungeana, also have resin canals in the vascular traces, as do some members of subgenus Pinus (e.g. P. radiata D. Don), suggesting that this is another homoplasious character. Two distinct rings of resin canals also occur in the traces of Pinus princetonensis Stockey from the Eocene of British Columbia (Stockey 1984), which is similar to section Pinus.
Regarding wood characters, growth increments observed in P. nelsonii, although faint, are characteristic of Pinus and the distantly related genus, Cedrus, both of which develop their cones over two or more years. Axial resin canals present in the wood of P. nelsonii are common in Pinaceae, having been reported from one or more species of all genera except Keteleeria, Pseudolarix, and Tsuga (Miller 1976a). Ray features have been emphasized for their diagnostic value in conifers (Greguss 1955, 1972) and described for Pinus nelsonii (Huerta Crespo 1963), but their taxonomic distribution in ovulate cones needs further study.
The present study fills an important gap in our knowledge of cone anatomical variation among Pinus subsections and identifies which fossil cones currently classified in the polyphyletic genus Pityostrobus may be closely related to Pinus. The expanded cone morphology data matrix continues to support the monophyly of Obirastrobus and Pseudoaraucaria, but their phylogenetic placement in Pinaceae remains unclear. A more robust phylogenetic hypothesis for fossil ovulate cones and extant genera may reveal new insights into the timing and sequence of evolution in Pinaceae. Further progress in understanding relationships between extant and fossil species is possible by including more taxa and characters in phylogenetic analyses. Dividing generic terminals into species terminals to avoid scoring characters as polymorphic is under way, as is adding characters not only from cones, but from other structural and molecular sources.
We thank Ruth Stockey and Ashley Klymiuk for making Charles Miller's Pinaceae reference collection available for comparison, and Manuel González Ledesma, Esmeralda Salgado Hernández, Denisse Téllez Mazzocco, Leticia Gómez, and Fred Gernandt for field assistance. We also gratefully acknowledge helpful suggestions on a previous version of this manuscript by Ruth Stockey, Gar Rothwell, and two anonymous reviewers. Financing was provided by the National Science Foundation (ATOL-0629508) to Sarah Mathews and DGAPA-PAPIIT (IN228209) to DSG.
Characters and their states. Characters and their states are from Smith and Stockey (2002), unless otherwise cited.