The morphology and internal structure of dogwood (Cornus L.) endocarps in the taxonomy and phylogeny of the genus

Endocarp morphology

The seventeen morphological characters (eight qualitative and nine quantitative) of the woody stones were analysed (Table 2) for 3,846 stones collected. The terminology used to describe these 17 morphological characters of endocarps was from Woźnicka, Melosik & Morozowska (2015) (Table 2; Fig. 3). Since few of endocarps’ qualitative characters (ASH, apical shape; BSH, basal shape; SSH, shape in the vertical projection; and VBP, position of vascular bundles on the surface) were described in the different states, the percentage share of the particular states of the same character was determined for each species under study (Supplemental File 5). If the share of a given state was 100% of the occurrences, the character was regarded as uniform for the group, subgenera, or species. If the share of a character state was greater than or equal to 90%, it was regarded as typical for these taxa.

Endocarp cross-section

Four quantitative characters of the endocarp structure visible on the endocarp equatorial cross-section were analysed (Table 3, Fig. 1) for 317 stones. Longitudinal sections of the stones were also submitted for scanning electron microscopy (SEM). The endocarps were sectioned with a Leica CM18050 cryostat in a cryochamber at −15 °C. The ImageJ application (Rasband, 1997–2018) was used for measurements. A Zeiss Axioscope A1 stereoscopic microscope was used for photographic documentation.

Endocarp morphology

The comparative analyses of the results were made at three levels: groups (BW, CC, DW, and BB), subgenera (Kraniopsis, Mesomora, Cornus, Cynoxylon, Syncarpea, and Arctocrania), and species.

The statistical analyses were based on the average values of the measurements of 17 morphological characters obtained from 254 specimens considered in the present work. The quantitative characters of the stones were transformed logarithmically to obtain a normal or close to normal data distribution for statistical purposes (Sárnal et al., 1999; Howell, 2007; Tabachnick & Fidell, 2007).

The principal component analysis (PCA) aimed to estimate the diversity of specimens in terms of the quantitative morphological characters, identify the main trends in the diversity of the set of specimens, and select the quantitative characters that were most closely related to the observed gradient of diversity. The PCA was used for all quantitative characters, including apical cavity length (ACL) and width (ACW). The apical cavity was present in 12 out of 22 species and was a constant character only in three of them. Using PCA allowed us to determine whether such very strongly variable characters have a significant influence on the differentiation of the Cornus genus. If endocarps had no apical cavity, zero was entered for ACL and ACW. First, all data were standardised due to the wide range of values. The calculations were based on the characters’ correlation matrix and varimax rotation. The Kaiser criterion was used to select the principal components (V) that significantly explained the variability of the set. Therefore, the principal components whose eigenvalue exceeded or was close to 1.0 were left. The PCA enabled the identification of the characters strongly correlated (r ≥ 0.60) with previously selected principal components.

The multivariate analyses of variance (MANOVAs) based on the three PC scores were carried out to determine whether a separation of endocarps in terms of their belonging to groups, subgenera, and species was statistically significant. To evaluate this differentiation, each PC axis was compared using analyses of variance (ANOVAs) with the Bonferroni correction.

At the next stage of statistical analysis, the specimens from particular groups, subgenera, and species were assessed for significant differences in the morphological characters selected through the PCA method. The homogeneity of variances of characters was checked with Levene’s test. As some of the characters did not meet the assumption of the homogeneity of variance, the significance of differences was verified using one-way ANOVA with the Welch correction and Dunnett’s T3 post-hoc test.

The statistical analyses of the eight qualitative characters were based on individual measurements of 3,846 stones. Chi-square tests of independence were used to test the relationship between specific qualitative characters and the belonging of the dogwood specimens to particular groups, subgenera, and species. The Yates correction was used in the analyses with one degree of freedom (df = 1). With this correction, the discrete distribution of the characters was better approximated by the continuous chi-square distribution (Aczel, 2000). When the expected value was less than 5, the closest states of the characters were combined into one class (in tests with df > 1), or Fisher’s exact test was used (in tests with df = 1).

To check whether the studied morphological characters of the endocarps reflected the current relations within the Cornus genus (Murrell, 1993; Xiang et al., 2006), similarity coefficients were calculated. For eight log-transformed quantitative characters (one character was excluded during the course of analyses; see Results), the Manhattan distance converted into similarity was calculated. In the case of eight qualitative characters, the Jaccard similarity coefficient was used. The qualitative characters were common in several states, and the same character states were often observed in the four compared groups of dogwood. Therefore, the similarities between the groups concerned a similar frequency of occurrence of character states. For this reason, when calculating the Jaccard index, the number of endocarps sharing the same character states was considered as the intersection for each pair and was divided by the total remaining number of endocarps for that pair. For both quantitative and qualitative characters, Gower’s similarity coefficient was developed. The obtained similarity coefficients were in the range [0 1], with 1 denoting the highest similarity (the compared groups were identical) and 0 denoting the lowest.

Most statistical analyses were performed using the program STATISTICA 11 (StatSoft, Poland). ANOVAs with Welch’s correction and Dunnett’s T3 tests were calculated with IBM SPPS 21.0 (SPSS Inc., Chicago, IL, USA). The Bonferroni correction was found at http://quantitativeskills.com/sisa/calculations/bonfer.htm. The chi-square tests of independence with df > 1were performed at http://www.quantpsy.org/chisq/chisq.htm.

Endocarp cross-section

The discriminant function analysis (DFA) was used to check which of the species under study were the most strongly discriminated on the basis of three examined characters of the endocarps present in all groups studied (GVT, germination valve thickness; SMW, septum width; and WTP, thickness of the endocarp wall divided by the endocarp diameter, multiplied by 100). Factor structure coefficients, that is, simple correlations between the characters and discriminatory axes were used. Each character was tested statistically to verify the initial assessment of diversity obtained through the DFA. Some of the data did not have a normal distribution after the logarithmic transformation either. Therefore, nonparametric tests were used on nontransformed data. The Kruskal–Wallis H test was used for three combinations of categories (four groups, six subgenera, 22 species). Numerous Mann–Whitney U tests were also conducted to compare consecutive pairs within each category. Statistical analyses were performed using STATISTICA 11 (StatSoft, Poland).

Quantitative characters

The diversity of the quantitative morphological characters was initially tested using PCA. The first three principal components together explained 77.95% of the total variance (V1 = 35.08%; V2 = 31.98%; V3 = 14.74%). The scatter diagram of the first two components showed a clear cluster of specimens belonging to the same species and those belonging to the same subgenus and group (Fig. 2). The groups and subgenera were clearly separated, with the exception of the CC group and the Cynoxylon subgenus from the BB group, which overlapped. The first principal component mostly separated the DW and CC groups (Table 4). Four characters were negatively correlated with the first principal component: the endocarp thickness (ST), the apical cavity width (ACW), the apical cavity length (ACL), and the endocarp width (SW) (Table 5). The second component separates the above-mentioned groups from the species included in the BW group. The following characters were negatively correlated with the second component: the endocarp length (SL) and the endocarp length-to-width ratio (SL/SW). The following characters were positively correlated: the share of bifurcated vascular bundles (FV%) and the number of vascular bundles on the endocarp surface (VN). The third principal component noticeably separated the BB and DW groups (Table 4). The following characters were positively correlated: the apical cavity width (ACW) and the apical cavity length (ACL). The endocarp width-to-thickness ratio (SW/ST) was not significantly correlated with any of the principal components (Table 5). Therefore, this character was not included in further analyses.

MANOVAs based on PC scores (V1–V3), with the taxonomic affiliations as independent variables, confirmed that tested groups, subgenera, and species differed in the quantitative characters of endocarps (groups: Wilks’ lambda = 0.03, F_(3,9) = 203.67, P < 0.0001; subgenera: Wilks’ lambda = 0.00, F_(3,15) = 311.6, P < 0.0001; species: Wilks’ lambda = 0.00, F_(3,63) = 113.05, P < 0.0001). The statistically significant separation within taxonomic units occurred along all tree axes (ANOVAs for groups: F_(3,250) = 69.37, 555.46, and 32.94, respectively, P < 0.0001; ANOVAs for subgenera: F_(5,248) = 262.58, 479.63, and 122.72, P < 0.0001; ANOVAs for species: F_(21,232) = 150.51, 164.95, and 51.62, P < 0.0001).

The differentiation of the particular quantitative characters of the endocarps between individual groups, subgenera, and species as shown in Figs. 3–5. The endocarp length (SL) and thickness (ST) were the best diagnostic characters as they significantly differentiated each group from the others (Figs. 3A and 3D; post-hoc tests, p < 0.05). The length and thickness of the endocarps (SL and ST) were the greatest in the CC group and the smallest in the DW group. The endocarp width (SW) and length-to-width ratio (SL/SW) also assumed the highest values in the CC group and the smallest values in the DW group. However, these two characters did not differ significantly between the BW and BB groups (Figs. 3G and 4A). The highest number of vascular bundles (VN) was present on endocarps of species from the BW group. This character distinguished all groups, with the exception of BB and DW (Fig. 4D).

Considering the differentiation in quantitative characters within particular subgenera, the endocarp length (SL) also revealed a significant difference between subg. Mesomora and subg. Kraniopsis as well as between subg. Cynoxylon and subg. Syncarpea (Fig. 3B). Subgenera Mesomora and Kraniopsis also differed significantly according to endocarp thickness (ST) and width (SW) (Figs. 3E and 3H). The endocarp length/width ratio (SL/SW) significantly differed between subg. Cynoxylon and subg. Syncarpea (Fig. 4B). The number of vascular bundles (VN) did not differ between subgenera Mesomora and Kraniopsis, between Cornus and Cynoxylon, and between Cynoxylon, Syncarpea, and Arctocrania (Fig. 4E).

Most of the species clustered in the same groups and subgenera exhibited great similarities in the analysed quantitative characters, but with some exceptions. The endocarp length (SL) and SL/SW ratio of C. kousa (Syncarpea) were significantly lower compared to those of C. florida and C. nuttallii (Cynoxylon) (Figs. 3C and 4C; species nos. 18–20). The endocarp thickness (ST) of C. alba, C. bretschneideri, C. occidentalis, and C. sericea were significantly smaller than those of other BW species (Fig. 3F; species nos. 1, 5, 11, and 14). The endocarp width (SW) of C. canadensis was significantly smaller than that of C. suecica (Fig. 3I; species nos. 21–22). Some similarities between species clustered in different subgenera were also observed. C. kousa (Synacrpea) and both C. canadensis and C. suecica from subg. Arctocrania did not differ (Dunnett’s T3 test, p > 0.05) in endocarp length (SL) and SL/SW ratio (Figs. 3B, 3C, 4B, and 4C). In turn, the endocarp thickness (ST) of Cornus officinalis (Dunnett’s T3 test, p > 0.05) did not differ from those of species from subg. Cynoxylon and Syncarpea (Figs. 3E and 3F).

An apical cavity was observed in stones of 12 out of 22 species under study. However, for most of these species, it occurred sporadically (C. australis, C. bretschneideri, C. drummondii, C. foemina, C. macrophylla, C. obliqua, C. sanguinea, C. walteri, and C. officinalis). Thus, the apical cavity length and width (ACL, ACW) were analysed in detail only for C. alternifolia, C. controversa (Mesomora), and C. mas. Both species of subg. Mesomora differed from each other in ACL and ACW. Furthermore, their cavities were significantly longer and wider comparing to C. mas (Figs. 5A and 5B).

Qualitative characters

The results of the chi-square test of independence showed significant differences between examined dogwood groups, subgenera, and species in all the qualitative morphological characters of the endocarps (Table 6).

None of the examined characters were uniform within all groups, subgenera, or species. However, it was possible to distinguish uniform characters for particular groups, subgenera, and species (Table 7).

The qualitative characters were the least differentiated in the DW group, where four characters of endocarps were uniform: smooth surface (SSF), absence of apical cavity (ACP), flat vascular bundles (VBP), and unforked vascular bundles (FV). Additionally, two typical characters (>90% of the occurrences) were identified in DW: rounded base (BSH) and absence of distinctive furrow (DF), with 92.2% and 96.8% percentage of occurrence, respectively. In the CC group, four characters were considered uniform: spherical = globose endocarps (SSH), rounded or truncated apex (ASH), smooth endocarp surface (SSF) and absence of distinctive furrow (DF). In the BB group, two uniform characters were observed: absence of the apical cavity (ACP) and unforked vascular bundles (FV), while rounded or truncate apex (ASH) was recognised as a typical character, with a percentage of occurrence of 91.1%. The endocarps of BW species were the most heterogeneous, and neither uniform nor typical characters in this group were present.

Four characters were found to be uniform in the following subgenera: Mesomora [rounded or truncate apex (ASH), presence of apical cavity (ACP), sunken vascular bundles (VBP) and forked vascular bundles (FV)]; Cornus [spherical=globose endocarps (SSH), rounded or truncate apex (ASH), smooth endocarp surface (SSF), absence of distinctive furrow (DF)]; Arctocrania [smooth endocarp surface (SSF), absence of apical cavity (ACP), flat vascular bundles (VBP), unforked vascular bundles (FV)]. Three uniform characters were found in two subgenera: Cynoxylon [spherical=globose endocarps (SSH), smooth endocarp surface (SSF), unforked vascular bundles (FV)]; Syncarpea [irregular shape (SSH), absence of apical cavity (ACP) and sunken vascular bundles (VBP)] while in subg. Kraniopsis no uniform characters were present.

With reference to uniform characters recognised in particular species, the smooth endocarp surface (SSF) was observed in endocarps of 16 species (Figs. 6–7). In turn, flat vascular bundles (VBP) were present in endocarps of 12 species, the absence of distinctive furrow (DF) was uniform for seven species, a spherical=globose endocarp (SSH) was uniform for six species, a rounded or truncate apex (ASH) was uniform for five species, unforked vascular bundles (FV) were uniform for four species, sunken vascular bundles (VBP) and a rounded basal shape for the endocarp (BSH) were uniform in three species, and raised vascular bundles (VBP) were uniform for two species (Figs. 6–7, Table 7, Supplemental File 5).

Considering the obtained results based on the analyses of the uniform characters, no differences were found between the CC and DW groups in terms of smooth endocarp surface (SSF) and between the BB and DW groups in terms of unforked vascular bundles (FV). With reference to subgenera, no differences were found between Cornus and Cynoxylon in terms of spherical endocarps (SSH); between Mesomora and Cornus in terms of rounded or truncate apex (ASH); between Cornus, Cynoxylon, and Arctocrania in terms of smooth endocarp surface (SSF); between Mesomora and Syncarpea in terms of sunken vascular bundles (VBP); and between Cynoxylon, Syncarpea, and Arctocrania in terms of unforked vacular bundles (FV). The results concerning the occurrence of uniform characters in individual species are presented in Table 7.

The analysis results based on the examination of all endocarp characters concerning the differentiation between the studied groups, the subgenera and species showed that the endocarp shape (SSH), the apical shape (ASH), the basal shape of the endocarp (BSH), the position of vascular bundles on the endocarp surface (VBP), and the absence/presence of distinctive furrow (DF) enabled the differentiation of each group from the others (chi-square tests, p < 0.05) (Figs. 6A, 6D, 6G, 7D, and 7J). The absence/presence of apical cavity (ACP), and the absence/presence of forked vascular bundles (FV) separated the examined groups from each other, with the exception of BB and DW (Figs. 7A and 7G). The endocarp surface (SSF) enabled the differentiation of all the groups except the CC and DW (Fig. 6J).

None of the characters under study differentiated between all subgenera. However, the rounded or truncate apex (ASH) significantly distinguished between subg. Cornus and subg. Cynoxylon and Syncarpea, while subg. Cornus and Mesomora did not differ according to ASH (Fig. 6E). The rounded basal shape (BSH) was present in 68.7–92.2% of endocarps of subg. Kraniopsis, Mesomora, Cornus, Syncarpea, and Arctocrania. Subgenus Cornus did not differ between subg. Kraniopsis and Mesomora in terms of BSH, while the differentiation between subg. Syncarpea and Arctocrania was significant despite the observed similarity (Fig. 6H). The smooth endocarp surface (SSF) was predominant among subgenera Kraniopsis, Mesomora, Cornus, Cynoxylon, and Arctorania (78.3–100% of the occurrences within these subgenera), with the only exception of subg. Syncarpea; no differences were found between subg. Kraniopsis-Mesomora and Cornus-Cynoxylon-Arctorania. In subg. Syncarpea (C. kousa), 86.3% of endocarps had rough surfaces; thus, this subgenus differed significantly from all others in terms of SSF (Fig. 6K). The rough endocarp surface (SSF) was recognised also in 95.1% and 100% of endocarps of two Kraniopsis species: C. amomum and C. obliqua, respectively (Fig. 6L). The sunken vascular bundles (VBP) did not differ between subg. Mesomora and Syncarpea, while significant differences in VBP were found between all other subgenera (Fig. 7E). With reference to the presence of unforked vascular bundles (FV), there were no differences between subg. Cynoxylon, Syncarpea, and Arctocrania (Fig. 7H). The significant differentiation in FV was found within the subg. Cornus as unforked vascular bundles were present on 100% of Cornus mas and on 42.2% of C. officinalis endocarps. In turn, the unforked vascular bundles were present on 99.5% of C. kousa endocarps (Syncarpea) (Fig. 7I, Supplemental File 5). On endocarps of Kraniopsis and Mesomora, the forked vascular bundles were present in 77.7% and 99.8%, respectively (Fig. 7H). The absence of distinctive furrow (DF) was predominant on endocarps of subg. Mesomora (96.6%) and Arctocrania (96.8%), and no differences were found between these two subgenera. Same as above, the following pairs of subgenera: Mesomora-Cornus (96.6–100%) and Cornus-Arctocrania (100–96.8%) were very similar in terms of the absence of DF. However, they still had differences. The presence of DF was predominant among the endocarps of the Cynoxylon and Syncarpea specimens (90.7% and 86.3%), and no differences between these two subgenera were found (Fig. 7K).

Qualitative and quantitative characters

Similarity coefficients calculated for the morphological characters of the endocarps showed a high similarity between the CC, BB, and DW groups (Table 8). In all comparisons, the BW group had the lowest similarity to other groups. In terms of quantitative characters, the BB-DW groups were clearly the most similar (similarity coefficient S = 0.71). In the case of qualitative characters, the CC-DW groups were the most similar (S = 0.45), but the groups CC-BB were only slightly less similar (S = 0.42). Considering both quantitative and qualitative characters, the BB-DW groups were again most similar (S = 0.54).

Quantitative characters

Considering the taxonomic importance of the examined quantitative characters, the endocarp length (SL) and its thickness (ST) were identified as the best diagnostic characters, as they significantly differentiated each group from others. The number of vascular bundles (VN) was the character that significantly differentiated almost all BW species from the red-fruited species (CC, BB, and DW). It was shown that the specimens representing the Kraniopsis-Mesomora complex (BW) had significantly more vascular bundles than the specimens representing the subgenera Cornus, Cynoxylon, Syncarpea, and Arctocrania (Figs. 4E and 4F). Within the red-fruited dogwoods, the average number of vascular bundles (VN) differed significantly for the CC, BB, and DW species, with the exception of C. nuttallii (Cynoxylon, BB) (Figs. 4D, 4E, and 4F). Such results agree with phylogenies based on molecular evidence (Xiang et al., 2006). However, the similarity of C. mas and C. officinalis (subg. Cornus, CC) with C. nuttallii in terms of VN supports Murrell’s (1993) morphological phylogeny.

The results on the subgeneric level showed that there were no differences in the endocarp length (SL) between Syncarpea and Arctocrania species. Additionally, the same character significantly differentiated between subg. Cornus and subgenera Cynoxylon, Syncarpea as well Arctocrania (Fig. 3B). Exactly the same similarities/differences were found according to the SL/SW ratio (Fig. 4B). The differentiation of SL and SL/SW found among the above-mentioned subgenera indicates that the described results support the molecular (Xiang et al., 2006) rather than the morphological (Murrell, 1993) phylogenies.

Considering other studied quantitative characters, such as the endocarp width (SW) and its thickness (ST), the results were not so clear. The SW analyses showed that subgenera Cornus and Mesomora did not differ in that character, which may highlight the results of the parsimony analysis of matK, according to which CC-BW was the pair of sister clades (Xiang et al., 2006). In turn, the stone thickness (ST) did not differ significantly between subg. Cornus and subg. Cynoxylon and Syncarpea, which supports Murrell’s (1993) phylogeny.

Qualitative characters

Among the qualitative characters, the endocarp shape in vertical projection (SSH), apical shape (ASH), basal shape (BSH), smooth/rough endocarp surface (SSF), position of vascular bundles on endocarp surface (VBP), absence/presence of forked vascular bundles (FV), and absence/presence of distinct furrow (DF) were taxonomically important characters, as they allowed us to differentiate between particular species, subgenera, or groups.

The differentiation in the stone shape (SSH) allowed the selection of the species with spherical, intermediate, flattened, and irregular stones. Flattened fruit stones were quite often present for species from BW (33.7%) and DW (42.9%) groups, while spherical endocarps were uniform for the CC group. Irregular stones were uniform for Syncarpea and allowed a clear distinction of C. kousa from the rest of the examined dogwoods. The stone shape (SSH) together with the degree of flattening of the stone, i.e. its thickness (ST) and length-to-width ratio (SL/SW) were often used in the past to analyse the diversity of the endocarps of very similar and phylogenetically closely related species (e.g. C. alba, C. sericea, and C. occidentalis from the BW group) (Fosberg, 1942; Szafer & Pawłowski, 1959; Rehder, 1967; Bean, 1976; Seneta, 1994; Rutkowski, 2004; Xiang et al., 2006; Schulz, 2012; Zieliński et al., 2014; Woźnicka, Melosik & Morozowska, 2015). However, the results of recent studies concerning dogwoods’ morphology (Schulz, 2012; Zieliński et al., 2014; Woźnicka, Melosik & Morozowska, 2015) and our analyses of ST and SL/SW average values have shown that the ranges of these characters mostly overlap; thus, they do not explain sufficiently the species status of these closely related taxa. Additionally, the results concerning the flattened shape (SSH) of C. alba, C. sericea, and C. occidentalis endocarps (97.8%, 92.1%, and 94.4%, respectively) and the presence of distinctive furrow running longitudinally on the lateral faces of their endocarps (DF) (53.8%, 77.2%, and 87.9%, respectively) also indicated a close similarity between C. alba, C. sericea, and C. occidentalis (BW). Considering these three species, Schulz (2012) proposed the new taxonomic approach of C. alba s.l. with subsp. alba and subsp. stolonifera. He treated C. occidentalis at the rank of variety below subspecies stolonifera of C. alba. This new taxonomic approach was accepted by Zieliński et al. (2014) and Woźnicka, Melosik & Morozowska (2015) and supported by our results.

The lack of significant differentiation in the endocarp apical shape (ASH) and its basal shape (BSH) supports close relationships between particular species from the subg. Kraniopsis (BW) (e.g., North American species C. amomum, C. obliqua, C. foemina, and C. racemosa or European species C. sanguinea and C. australis). With reference to C. amomum and C. obliqua, some other morphological similarities between these two species [e.g. the rough endocarp surface (SSF) (95.1% and 100%, respectively) and the uniform raised vascular bundles (VBP)] were also found. Referring to C. sanguinea and C. australis, the flat vascular bundles (VBP) with 93.7% and 100% of the occurrences and unforked vascular bundles (FV) with over 91.5% of the occurrences indicated their close similarity. All the described results support Schulz (2012), who suggested that C. obliqua and C. australis should be classified as subspecies within C. amomum and C. sanguinea, respectively. The results also complement the earlier findings describing the morphological and molecular similarities of these species (Wilson, 1964; Bean, 1976; Ball, 2005; Xiang et al., 2006; Schulz, 2012; Woźnicka, Melosik & Morozowska, 2015).

The noticeable variation in the form of endocarp apical shape (ASH) allowed the distinction between examined Arctocrania species (DW). In turn, the rounded or truncated apex was recognised as a uniform character for C. mas and C. officinalis (CC). According to Bojnanský & Fargašová (2007), C. mas stones have a pointed apex, whereas C. officinalis stones have a blunt apex. It is most likely that the differences between the results of our study and the cited data were caused by the inverse description of the stone apex and the stone base by Bojnanský & Fargašová (2007). The rounded or truncated apex (ASH) was also recognised as a uniform character for alternate-leaf dogwoods C. alternifolia and C. controversa (subg. Mesomora). However, it should be noted that this character was somewhat combined with the presence of distinct ACP, which was present on the endocarps of these two species and was recognised as a taxonomically important character. It allowed for a significant separation of two alternate-leaf dogwoods C. alternifolia and C. controversa (subgen. Mesomora, BW) from species belonging to subg. Kraniopsis, which was in line with earlier results by other authors (Eyde, 1988; Xiang & Boufford, 2005; Schulz, 2011, 2012; Woźnicka, Melosik & Morozowska, 2015). Our results confirmed also the presence of noticeable, but clearly much smaller, apical cavities on the stones of several Kraniopsis species, which was described earlier by Woźnicka, Melosik & Morozowska (2015). Apart from the BW group, much smaller apical cavities were present on the endocarps of examined species from subg. Cornus. They were frequent on C. mas (99.9%) fruit stones and occurred sporadically on C. officinalis (16.1%) endocarps. We also observed a shallow depression on C. volkensii stones (unpublished data). Eyde (1988) did not clearly state whether cavities could be found in these species nor in what form. As far as the species from the group of large-fruited edible dogwoods (CC) were concerned, the author only described the apical cavity that was wide and shallow in C. volkensii; in C. chinensis, it had the form of a V-shaped incision and was absent on C. sessilis endocarps. According to Manchester, Xiang & Xiang (2010), the endocarps of species belonging to subg. Cornus differ in the expression of an apical cavity. These authors described a V-shaped apical notch in C. chinensis and a broad shallow cavity in C. volkensii. The other species were described as apically rounded without or with inconspicuous apical cavity. Atkinson, Stockey & Rothwell (2016) confirmed the presence of the apical cavity on stones of C. mas, C. officinalis, and C. volkensii, together with a few other living (C. chinensis, C. eydeana, and C. sessilis) and extinct species (C. piggae, C. ettingshausenii, and C. multilocularis). According to the results, the ACP in C. mas fruit stones was significantly smaller than in endocarps of C. alternifolia and C. controversa (Fig. 5). Based on the described results, the presence or absence of apical cavities is a taxonomically important character. However, according to already published data, it is a plesiomorphic character and thus have no value in the analysis of the phylogeny of the genus (Murrell, 1996).

The extensive and detailed study of Eyde (1988) also includes an analysis of the variation in a stone’s surface. The author emphasised that the stones of the herbaceous species (DW) were easy to distinguish because they were smoother than the stones in the BW group. Similar results were obtained earlier by Woźnicka, Melosik & Morozowska (2015), which were confirmed in the present study. The smooth endocarp surface (SSF) was found in all three groups of the red-fruited dogwoods (CC, BB, and DW), but with the exception of C. kousa (Syncarpea, BB) (Fig. 6L), which agrees with Eyde (1988, Figs. 13h and 13i, p. 276) who showed that C. kousa stones were irregular and rough. A uniform character of smooth endocarps present in three subgenera from different groups—Cornus (CC), Cynoxylon (BB), and Arctocrania (DW)—supports published phylogenies based on morphological and molecular data (Murrell, 1993; Xiang et al., 2006).

The vascular bundle position on the endocarp surface (VBP) differentiated significantly Kraniopsis and Mesomora species. The sunken vascular bundles were recognised as a uniform and taxonomically important character for subg. Mesomora. In endocarps of Kraniopsis species, mostly flat vascular bundles were present. The present results confirmed the results of Woźnicka, Melosik & Morozowska (2015). The significant differentiation in VBP within the subgenera of red-fruited dogwoods (CC, BB, and DW) also indicate systematic importance of that character. The presence of flat vascular bundles, which were uniform for the endocarps of a few species belonging to different morphological groups, like C. officinalis (CC), C. nuttallii (Cynoxylon, BB), C. canadensis, and C. suecica (DW), supports the phylogenies based on the morphological and molecular evidence (Murrell, 1993; Xiang et al., 2006). In turn, the presence of the unforked vascular bundles on endocarp surface (FV) showed a distinct similarity between two groups of red-fruited dogwoods (BB-DW), which supports the close phylogenetic relationship of these clades shown by Xiang et al. (2006). However, there was a close similarity between C. mas (subg. Cornus) and BB species according to the presence of the unforked vascular bundles, which supports Murrell’s (1993) morphological phylogeny.

The presence of distinctive furrow running longitudinally on the lateral faces of an endocarp (DF) was a strongly variable character. It was uniform only for the CC group, so no support for the published phylogenies was found according to that character. However, some of its taxonomic usefulness should be mentioned. The DF was present on 35.6% of Kraniopsis endocarps and only on 0.4% of Mesomora fruit stones (BW). Such results agree with the results of Fosberg (1942), Schulz (2012), and Woźnicka, Melosik & Morozowska (2015). In reference to the red-fruited dogwoods, fruit stones with furrows on sides were predominant among the C. florida, C. nuttallii, and C. kousa endocarps (BB) (86.3%–90.7%), while it was almost (96.8%) or completely (100.0%) absent on endocarps of CC and DW species. Such results indicate that the presence of DF on the endocarps of most of big-bracted species may be helpful in distinguishing them from other red-fruited dogwoods.

The calculated similarity coefficients allowed us to check whether the described morphological similarities between the studied groups and subgenera reflect their current phylogenetic relations. The similarity coefficients (Table 8) calculated jointly for the qualitative and quantitative characters showed that the endocarps of big-bracted dogwoods (BB) and dwarf dogwoods (DW) were the most similar. Such results indicated support for molecular phylogenies (Xiang et al., 2006). When only qualitative characters were considered, the closest similarity of groups CC-DW was found mainly due to the smooth endocarp surface (SSF). This character may possibly be considered a synapomorphy for CC and DW groups. However, that similarity does not directly reflect any of the published phylogenies (Murrell, 1993; Fan & Xiang, 2001; Xiang et al., 2006).