Crataegus monogyna Jacq. is a widespread species of Crataegus sect. Crataegus that occurs in much of Europe, northern Africa and western Asia. Within the area of its natural distribution it hybridizes with several other species of sect. Crataegus, e.g., Crataegus laevigata (Poir.) DC., Crataegus rhipidophylla Gand., Crataegus meyeri Pojark., Crataegus pentagyna Waldst. & Kit. ex Willd., Crataegus orientalis M. Bieb., and Crataegus azarolus L., as well as Crataegus nigra Waldst. & Kit. of sect. Sanguineae (Albarouki and Peterson 2007; Byatt 1975; Christensen 1983; Christensen 1992a, b, 1994; Christensen and Zielinski 2008; Dönmez 2004). In fact, Christensen (1992) applied the term “compilospecies” to Crataegus monogyna. This term, coined by Harlan and DeWet (1963), describes species that aggressively acquire genes from other species by introgressive hybridization, potentially explaining the “…great variability of Crataegus monogyna and also its wide distribution” in the Old World (Christensen 1992a). Crataegus monogyna was introduced to the U.S.A. and Canada by the early European settlers (Billings 1862; Douglas 1914; Kirk 1819; Provancher 1863). It has often escaped from cultivation and, e.g., in abandoned fields and woodlands with extensive hawthorn colonization, it may hybridize with native diploid species of Crataegus such as Crataegus punctata Jacq. (sect. Coccineae Loudon; Phipps pers. comm.; Wells and Phipps 1989) and Crataegus suksdorfii (Sarg.) Kruschke (sect. Douglasia Loudon; Dickinson et al. 2008; Love and Feigen 1978; Talent and Dickinson 2005). Because of the striking contrast in leaf shape between members of Crataegus sect. Crataegus and most North American Crataegus species, these hybrids are currently the best-known examples of diploid-diploid hybridization in the North American Crataegus flora. We provide names for these two nothospecies (as well as for the corresponding nothosections and nothoseries), referring to existing documentation in the literature for Crataegus ×ninae-celottiae K.I. Chr. & T.A. Dickinson (Crataegus monogyna × Crataegus punctata Jacq.; Wells and Phipps 1989). We also document variation in leaf shape for the second hybrid, Crataegus ×cogswellii K.I. Chr. & T.A. Dickinson (Crataegus suksdorfii × Crataegus monogyna), and provide new sequence data from ITS2 and chloroplast DNA barcoding loci that confirm the genetic relationships between the two nothospecies and their respective parents.

Sampling. Because the occurrence of Crataegus monogyna and its hybrids is sporadic, most of our samples are non-random, and merely attempt to document the co-occurrence of the parental species and (or) their hybrids (Table 1). Only in the case of the hybrid swarm found at the Cogswell-Foster Preserve in Linn Co., Oregon (site OR1), have we used either the throw of a pair of dice or ignorant person sampling (Ward 1974) in order to draw more unbiased samples, with the inevitable consequence that these samples reflect the greater frequency of the introduced species and its hybrids. To mitigate this, we have included individuals of mostly diploid Crataegus suksdorfii from other sites in order to reflect the variation found in this taxon.

Note that we distinguish the taxon referred to here as Crataegus suksdorfii from the other western North American black-fruited hawthorn with 20 stamens per flower, Crataegus gaylussacia A. Heller. This is because these two taxa are allopatric (Coughlan 2012 and unpubl. data), and differ in morphology and cytotype. Crataegus gaylussacia has shorter petioles and thorns that are thicker at their base than is the case with diploid Crataegus suksdorfii (Dickinson unpubl. data). Molecular data are consistent with Crataegus gaylussacia being an autotriploid derivative of diploid Crataegus suksdorfii (Zarrei et al. http://2012.botanyconference.org/engine/search/index.php?func=detail&aid=536 and unpubl. data; see also Lo et al. 2009). In contrast, the Crataegus suksdorfii complex has been shown to comprise, in addition to diploids, allotriploids and allotetraploids (Zarrei et al. http://2012.botanyconference.org/engine/search/index.php?func=detail&aid=536 and unpubl. data).

In order to increase our sample for molecular studies we have supplemented field collections of leaf tissue and herbarium vouchers with tissue removed from existing specimens in the ROM Green Plant Herbarium. Historical records of the distribution of Crataegus monogyna were collected from five herbaria across Canada (TRT, MTMG, MT, QFA and UBC). Online databases of Canadian and U.S. herbaria used included ACAD, the Invader Database System of the University of Montana (which contains information for five northwestern states: Idaho, Montana, Oregon, Washington, Wyoming), OSC, and WTU. Distribution maps were prepared from specimen locality data using SimpleMappr (Shorthouse 2010). Names of Crataegus sections and series used here follow those published by VASCAN (Brouillet et al. 2010), and are accepted names sensu FNA Ed. Comm. in prep.

Morphology. For this study we concentrated on capturing and analyzing leaf shape data, as described elsewhere (Dickinson et al. 2008). Many previous studies of hybridization involving Crataegus monogyna (Bradshaw 1953; Byatt 1975; Love and Feigen 1978), and of leaf shape variation in Crataegus generally (e.g. El-Gazzar 1980; Phipps and O’Kennon 2007), have attempted to quantify leaf lobing by means of a ratio of two measurements, \(x\) and \(y\), where \(x\) is the distance from the tip of a lobe (usually the most basal one) to the deepest point of the sinus between that lobe and the adjacent one above it, and \(y\) is a measure of leaf size, usually the parallel distance from the tip of the lobe to the midrib. This approach can be effective when comparisons involve only leaves that have some degree of lobing (e.g. studies of hybridization between Crataegus monogyna and Crataegus laevigata in Europe, or of the lobed leaves of many species belonging to North American Crataegus sect. Coccineae, such as Crataegus punctata). However, when lobing is absent altogether the necessary landmarks (lobe tip, deepest point of the sinus) are absent, and the distance \(x\) is undefined or is set to zero (Love and Feigen 1978). In this case, a better approach is to carry out multivariate analyses of additional measurements of leaves and other organs (Wells and Phipps 1989), or to quantify variation in the leaf outline as a whole. Elliptic Fourier coefficients obtained from digitized leaf outlines captured using MorphoSys (Meacham and Duncan 1991), or the Fourier amplitudes derived from them, provide a useful method for doing just this (Dickinson et al. 2008; McLellan and Endler 1998; Rohlf and Archie 1984).

Leaf outline data were collected from two overlapping samples: (1) short shoot leaf spectra (Dickinson and Phipps 1984) collected from a random sample of individuals at the Cogswell-Foster Preserve (comprising one Crataegus suksdorfii, seven Crataegus monogyna, and 12 putative hybrids), and (2) leaves on herbarium specimens from the Cogswell-Foster Preserve and other locations in the Pacific Northwest. In the latter the attempt was made to sample the leaf shape variation seen in Crataegus suksdorfii as widely as possible. In both cases, variation in the shape of the leaf blade (i.e. excluding the petiole) was summarized by means of 39 Fourier amplitudes, and displayed by means of principal components analysis.

For each leaf outline we also obtained the area (\(A\)) and perimeter (\(P\)), so as to calculate the inverse of the dissection index described by Kincaid and Schneider (1978), i.e. \(\mathrm{inv}(D.I.) = 2(A\pi)^{1/2}/P\), a parameter that has an upper bound of one for a perfect circle regardless of size, and approaches zero as the length of the perimeter increases with increased lobing of the outline (Dickinson 2003; Dorken and Barrett 2004). In addition to outline data we made linear measurements with which to index overall leaf shape: \(X\), leaf blade length above the widest point; \(Y\), leaf width; and \(Z\), leaf blade length below the widest point (Marshall 1978). On some of the flowering specimens in our sample we collected additional data on stamen number, style length and style number (in fruiting specimens, equivalently, pyrene number), and stigma width, in order to compare these with data collected by others from the introduced species and Crataegus punctata. After transformation to a common [0, 1] range these data were also summarized using principal components analysis. Analyses of variance were carried out on selected measurements. All data analyses described above were carried out using the R environment for statistical computing (R Core Team 2013). Significance of individual principal component axes was evaluated using the broken-stick criterion (Frontier 1976) with the help of R function evplot (Borcard et al. 2011).

Molecular methods. Four DNA barcodes (rbcL, matK, trnH-psbA, and ITS2; CBOL Plant Working Group 2009; Chase et al. 2007; Hollingsworth et al. 2011) were generated directly from genomic DNA for a worldwide sample of Crataegus (Dickinson et al. http://2011.botanyconference.org/engine/search/720.html; Zarrei et al. unpubl. data). The plastid origin of the markers was used to establish the maternal parentage of the hybrids. DNA was extracted and amplified from leaf tissue of individuals representing the two hybrids and their parent species (Table 2) using Canadian Centre for DNA Barcoding (CCDB) protocols (Ivanova et al. 2011; Kuzmina and Ivanova 2011a, b). This sample overlapped partially with the cloned ITS2 one (below), and provided an additional two Crataegus suksdorfii, 10 Crataegus monogyna, and five Crataegus punctata individuals, as well as one more of each of the two hybrids (Table 2).

We also analyzed data from another project (Zarrei et al. http://2012.botanyconference.org/engine/search/index.php?func=detail&aid=536 and unpubl. data) in which ITS2 was cloned for a sample of individuals that included 14 Crataegus suksdorfii, four Crataegus monogyna, three Crataegus punctata and two each of the two hybrids (Table 3). Methods for extracting total genomic DNA, marker amplification, cloning, DNA sequencing, and collapsing original sequences to unique sequences (ribotypes) are described elsewhere (Zarrei et al. http://2012.botanyconference.org/engine/search/index.php?func=detail&aid=536 and unpubl. data). Here we report on analyses of a total of 160 ribotypes (Table 3). A recombination test was performed using RDP4 Beta 4.14 (Martin et al. 2010). The Neighbor-Net analysis (Bryant and Moulton 2004) was undertaken using SplitsTree v.4.12.3 (Huson and Bryant 2006) to visualize incompatible splits in the network from uncorrected p-distances calculated with MEGA5 (Tamura et al. 2011). Bootstrap support (BS) was estimated using 1, 000 bootstrap pseudoreplicates (Felsenstein 1985) implemented in SplitsTree.

Flow cytometry. Flow-cytometric methods for quantifying nuclear DNA in embryo and endosperm followed Talent and Dickinson (2007a). Embryo DNA amounts of 1.48–1.70 pg were taken to indicate diploids, and an endosperm to embryo ratio of approximately 1.5 was taken to indicate sexual reproduction with meiosis.

Morphology. Despite differences in sample size, the Pacific Northwest hybrid, Crataegus ×cogswellii, appears more variable than either of its putative parents, Crataegus monogyna or Crataegus suksdorfii (Fig. 1). The hybrid is clearly intermediate with respect to both leaf lobing (the inverse Dissection Index; Fig. 1) and style number (STYLE; Fig. 1). Principal components analyses of leaf outlines from Pacific Northwest Crataegus monogyna, Crataegus suksdorfii, and their putative hybrid, demonstrate variation in leaf shape both within and between these three entities (Fig. 2A, B). The first principal component reflects the contrast between the unlobed leaves of Crataegus suksdorfii and the markedly lobed ones of Crataegus monogyna, as well as the intermediacy of the hybrid (Fig. 2A, B), much as illustrated earlier by Love and Feigen (1978; their Fig. 3), and by Wells and Phipps (1989) for the Ontario hybrid and its parents (their Fig. 4). The second principal component reflects variation in the relative overall lengths and widths of the leaf outlines (Fig. 2A).

DNA barcode loci. Analyses of both the directly sequenced and the cloned ITS2 ribotypes demonstrate the parentage of both putative hybrids (Fig. 3; Table 3); no signs of recombination were detected in the cloned ITS2 dataset. ITS2 sequences from the hybrids resemble either Crataegus monogyna or one of the native North American species. The way in which both parental ribotypes are maintained in each of the hybrids examined here is probably due to how recently the hybrids have been formed: less than 200 years ago in the case of the Ontario hybrids (Douglas 1914; Kirk 1819; Provancher 1863), and less than 100 years ago in the case of the Oregon ones (the earliest specimen of Crataegus monogyna was collected in 1914 in Douglas Co. Oregon; Phipps 1998). These time periods are evidently too short for genome homogenization (concerted evolution) to have taken place, even in diploids reproducing sexually. Our small sample of seed from the hybrids (Table 4) parallels earlier results (Talent and Dickinson 2007a) showing diploidy and sexual reproduction in both parental taxa.

Only two of the three chloroplast genome barcode loci showed sufficient variation for individuals from Crataegus section Crataegus to be distinguished from ones belonging to either Crataegus section Coccineae or Crataegus section Douglasia (Table 2). Sequence data from both rbcL-a and the trnH-psbA spacer region showed the same two clusters, Crataegus sections Coccineae and Douglasia (Cluster 1), and Crataegus sect. Crataegus (Cluster 2; Table 2). The way in which the hybrids fell into one of these clusters or the other demonstrates that, with one exception, Crataegus monogyna is the female parent of the Ontario hybrids with Crataegus punctata studied here, while Crataegus suksdorfii is the female parent of the Pacific Northwest hybrids.

These results corroborate earlier observations based on seed-set in artificial crosses between the parent species (Love and Feigen 1978; Wells and Phipps 1989). In reciprocal pollinations seed set was greatest (32–34%) when Crataegus monogyna stigmas received pollen from Crataegus punctata (Wells and Phipps 1989). Fruit set was most successful when Crataegus monogyna pollen was applied to the stigmas of Crataegus suksdorfii flowers (mean 42%, range 25–73%, compared to a 29% mean fruit set by Crataegus suksdorfii with open pollination; Love and Feigen 1978). However, all reciprocal crosses between Crataegus monogyna, Crataegus suksdorfii, and their hybrid yielded seeds (R. M. Love, personal communication).

Our use of data from DNA barcoding is not a test of the value of DNA barcoding in Crataegus, as this is discussed elsewhere (Dickinson et al. http://2011.botanyconference.org/engine/search/720.html; Zarrei et al. unpubl. data). Rather, we have taken advantage of our barcode sequence data from individuals unequivocally identifiable as Crataegus monogyna, Crataegus punctata, Crataegus suksdorfii and their hybrids in order to use sequence similarity to inform us about the hybridization process.

Hybridization. Since its introduction to North America during the late 18^th and the 19^th centuries (Kirk 1819; Provancher 1863; Douglas 1914), first on the east coast and then on the west, Crataegus monogyna has become widely naturalized in the U.S.A. (EDDMapS 2013) and Canada (Phipps and Muniyamma 1980; Phipps 1998; Lin 2009). Nevertheless, except for isolated occurrences in northern Delaware and adjacent Pennsylvania, as well as in Kentucky, Utah, and the San Francisco Bay area in California, Crataegus monogyna in North America is not found south of 40°N latitude. In Ontario, Crataegus punctata appears to be the only native diploid with a similarly late flowering period that is also frequently sympatric with Crataegus monogyna (Fig. 1 in Campbell et al. 1991; Fig. 4). Crataegus suksdorfii is the only native hawthorn in the Pacific Northwest known to include diploid individuals, and these are restricted to Oregon and adjacent California and Washington, west of the Cascades (Fig. 5; Lo et al. 2013). Where they co-occur, diploid Crataegus suksdorfii and Crataegus monogyna flower at the same time, the latter species much more abundantly than the former (Love and Feigen 1978).

Crataegus monogyna may never have been commonly planted in boundary hedges in Canada as it was in Europe. Fences and hedges appear to have been only rarely constructed in 17th Century Canada by European settlers to confine ruminant animals (Greer 2012); the animals were instead fed indoors, but allowed to roam the arable land for a short season after harvest, confined by the wall of surrounding forest. To this day, the hawthorn commonly growing along Ontario fence lines consists of native species, perhaps naturally occurring there. In Ontario forests we often encounter remnants of zig-zag post-and-rail fences, and these had the advantage over a hedge that they could be rapidly constructed as needed to mark property boundaries or to keep animals out of particular areas. In the United States hedging had its advocates in the early nineteenth century, but one of these described the superiority of native species like Crataegus crus-galli (“cockspur” or “Newcastle thorn”) and Crataegus marshallii (“parsley-leaved” or “Virginia thorn”) over introduced Crataegus monogyna (Kirk 1819; “to sow or plant without fencing, would (in this country) be a useless labour”).

Flow cytometry of seeds from both hybrids was consistent with diploid embryos and triploid endosperm, except that the embryos from Crataegus ×cogswellii show slightly higher than diploid measurements, higher than the 1.39–1.66 pg measurements previously obtained from leaf data (Table 4; Talent and Dickinson 2005). Whether the seeds involved would have germinated is unknown, but in contrast to the large healthy looking seeds from Crataegus ×ninae-celottiae, those from Crataegus ×cogswellii had smaller embryos and were variously misshapen. We noted that some individual trees of Crataegus ×cogswellii have a high degree of parthenocarpy—completely seedless fruit—and the seeds we collected may therefore have been supernumerary to any strongly viable seeds. We can only state that Crataegus ×cogswellii apparently carries out both meiosis and fertilization, as expected of other diploid Crataegus (Table 4; Talent and Dickinson 2007b).

In her examination of hybridization between Crataegus punctata and Crataegus monogyna in Ontario, Purich (2005) found that the styles of Crataegus punctata are significantly longer than those of Crataegus monogyna (mean_mono = 4.1 mm; mean_punc = 7.3 mm; sample sizes 5/52 and 7/116, individuals/styles). Differences between the two species in pollen grain diameter, hence volume, are not significant (Purich 2005). No such difference in style length is present when comparing Crataegus monogyna and Crataegus suksdorfii. These results suggest that in Ontario, at least, the longer styles of Crataegus punctata could act as a barrier to the successful penetration of Crataegus punctata ovules by pollen tubes from Crataegus monogyna pollen grains (Table 2). With style lengths and pollen grain diameters in Crataegus monogyna and Crataegus suksdorfii similar (Dickinson unpublished data), it may be that the more abundant flower production of Crataegus monogyna (Love and Feigen 1978) contributes to its role as the predominant pollen parent of Crataegus ×cogswellii. The exception to the summary above (TRT203 in Table 2; Crataegus punctata as the maternal parent) reflects the way in which differences in style length likely act to influence the direction of hybridization in a probabilistic rather than an absolute way.