A numerical taxonomic study of the family Zygophyllaceae from Egypt

(A numerical taxonomic study of the family Zygophyllaceae from Egypt). A systematic study of 29 taxa belonging to 7 genera of subfamilies Balanitoideae, Zygophylloideae, Peganoideae, Tribuloideae, Seetzenioideae and Tetradiclidoideae of Zygophyllaceae from Egypt was carried out by means of numerical analysis based on sixty-one morphological characters, including vegetative parts, pollen grains and seeds. On the basis of UPGMA clustering and PCO analysis, six main groups are recognized. Representatives of these groups are clustered together based on characters with high factor loadings in the PCO analysis. Th e results indicate that Zygophyllaceae are heterogeneous, including Peganum harmala which has been proposed to belong in a separate family. Zygophylloideae, Tribuloideae, Tetradiclidoideae and Seetzenioideae are the most homogeneous groups. Sections and groups in both Tribulus and Fagonia seem artifi cial.


Introduction
Zygophyllaceae is a widespread family of some 27 genera and 285 species subdivided into fi ve subfamilies (Sheahan & Chase 1996;2000).It consists of herbs, shrubs and trees growing in arid and semi-arid areas in the tropics and subtropics.Earlier studies place the Zygophyllaceae in diff erent orders, e.g.Sapindales, Rutales, Polygalales, Linales, and Geraniales (Cronquist 1968;Takhtajan 1969;1980;1983;1986;Th orne 1992).Soltis et al. (2000) put the Zygophyllaceae and Krameriaceae together in their own order Zygophyllales within the eurosid I group, and this position is changed in APG III (2009).Th ey put it within Fabids group.Delimitation of taxa within the family has repeatedly changed over time, because of their diversity in structural detail, particularly in Balanites, Nitraria, Peganum and Tetradiclis.For example, Engler (1896a;1931) divided the family into seven subfamilies, 8 tribes and 4 subtribes: Peganoideae, Tetradiclidoideae, Chitonioideae, Augeoideae, Zygophylloideae, Nitrarioideae and Balanitoideae.He considered that Zygophylloideae (including the tribuloid genera) formed the central typical group, with Augeoideae based on the characters of the leaves and fruits.
Pollen morphology of the family has been also examined by Erdtman (1952), Shimakura (1973), Kuprianova & Alyoshina (1978), Yunus & Nair (1988) and Perveen & Qaiser (2006).Sheahan & Cutler (1993) investigated the vegetative anatomy of 37 species in 19 genera, and illustrated that there is anatomical evidence to exclude of Balanites into a separate family and they recommended that the tribuloid genera Tribulus, Kallstroemia and Kelleronia should be separated from the Zygophylloid genera at least at subfamily level.Sheahan & Chase (1996) examined the phylogenetic analysis of Zygophyllaceae based on morphology, anatomy and rbcL DNA sequence.Zygophyllaceae form a clade.They divided Zygophyllaceae into five subfamilies, whereas Nitraria and Peganum were separated from Zygophyllaceae.Eventually, Sheahan & Chase (2000) analysed both rbcL and trnL-F sequences from 36 members of Zygophyllaceae.The results confirmed the previous division of Zygophyllaceae into five subfamilies, and Zygophylloideae were strongly supported as monophyletic, whereas Zygophyllum turned to be a polyphyletic genus.Beier et al. (2003), using non-coding trnL plastid data for 43 species of Zygophylloideae investigated the phylogeney of the family, retrieving a monophyletic subfamily Zygophylloideae, and the genus Zygophyllum, together with the genera Augea, Tetraena and Fagonia, turned to be a paraphyletic.In the fl ora of Egypt Zygophyllaceae is represented by six genera belonging to three subfamilies: Peganoideae Engl., Tetradiclidoideae Engl., and Zygophylloideae (Tackholm 1974;El Hadidi 1972), but recently Boulos (2000) recorded fi ve genera.Systematics of Zygophyllaceae may be diffi cult because only leave or fruit morphological characters are traditionally in use by taxonomists.Th ese characters are variable even within genera, or diff er with one another in their distribution patterns among genera and may not refl ect natural groups.
Th e aim of present study was to confi rm the phenetic relationships between genera within the family in Egypt based on a large number of characters (61) with the use of numerical taxonomy and to check whether these results correspond to the systematics of the genera proposed by Engler (1896a;1931), El Hadidi (1975) and Sheahan & Chase (1996;2000)

Plant material
Th e present study is largely based on herbarium material kept in the following herbaria: CAI, CAIM, and SHG.A fresh material of most of the taxa from several localities in Egypt and Saudi Arabia was also studied.In the analyses, species constituted the OTU (Operational Taxonomic Unit) see Appendix 1.In order to broadly sample the variation, the OTU's consist of a number of collections (either herbarium specimens or fresh material or both) from diff erent localities in Egypt.For some species from Egypt few specimens were available, in that case specimens from Saudi Arabia were used (Table 1).

Morphological characters observations
Table 2 shows the characters and character states scored for plant, pollen, and seed morphology, averaged for each OTU.In total 61 characters were taken into consideration, comprising 13 quantitative and 48 qualitative characters.Twelve of the qualitative characters were scored as binary and the remaining were scored as multi-state characters.Eleven characters from the 61 characters were treated as invariable characters and the rest are variable characters.

Vegetative parts, fl ower and fruit characters
Th e measurements for all specimens of a taxon were averaged into one OTU score for each of the characters.OTU scores for quantitative characters were arithmetic means based on at least 3 specimens (whenever possible).Because herbarium specimens cannot be considered to be a random sample of the species, we followed Wieringa (1999: 62-65) by calculating the mean of the minimum and maximum measurement.For some of the OTU' s observations for some of the characters were not available, and these omissions were coded as missing data (-999).Th e complete data matrix is available on request at the Botany Department, Faculty of Science, Sohag University, Egypt.

Pollen grain and seed characters
Th e data of pollen morphology were mainly obtained from Erdtman (1952); Agababyan (1964) ;El Hadidi (1966); Yunus & Nair (1988); and Perveen & Qaiser (2006).Dried mature seeds were first examined by light microscope (Olympus type BH-2), and 5-10 seeds for each taxon were selected to cover the range of variation when available.Seeds were mounted on stubs with double adhesive tape.Th e stubs were sputter-coated with gold/palladium for 3 min.in a EMITECH K550.Aft er coating, the specimens were examined with a Jeol-6300 scanning electron microscope, using accelerating voltages at 15-20 KV.

Data analysis
Two types of analyses were performed with NTSYS-pc 2.02k soft ware (Applied Biostatistics Inc., Setauket, New York, USA).Firstly, I performed a cluster analysis using average taxonomic distance and UPGMA clustering (procedures SIMINT, SAHN, and TREE).To reduce the eff ects of diff erent scales of measurement for diff erent characters, the values for each character were standardized with procedure STAND, according to the formula: yI,STD = (yi-AVGyi)/ STDyi), Where the default value in NTSYS-pc (STAND) for yi = the value to be standardized, AVGyi= the average of all values for the character, and STDyi = the standard deviation.Th e cophenetic correlation coeffi cient between the distance matrix and the tree matrix was calculated to examine the goodness of fi t of the cluster analysis to the distance matrix (procedures COPH and MXCOMP).Secondly, a principal coordinates analysis (PCO) was performed, using the product-moment correlation as a coeffi cient.Th e procedure SIMINT was used to calculate the distance matrix based on STAND data, the procedures EIGEN, PROJ, and MXPLOT to perform the PCO.A PCO was preferred rather than a PCA (Principal Components Analysis), because a PCO performs better on data sets with missing data (Rohlf 1972).

Cluster analysis
Figure 1 shows the UPGMA phenogram comprising all OTU's in the present study.Th e cophenetic correlation of distance matrix and tree matrix was 0.88, indicating a good fi t of the phenogram to the distance matrix (see Rohlf 1993).

Discussion
Systematics must largely rely on morphological characters to defi ne taxa.Problems in classifi cation arise when the taxa display a large amount of variability, due to phenotypic plasticity (van den Berg & Groendijk-Wilders 1999).Groups within the Zygophyllaceae have become specialized for similar arid habitats, which has probably led to much parallel and convergent evolution.Some characters which turn out to be among the most homoplasious and distinguish taxa, such as the presence of staminal appendages, nectariferous disc and endosperm.Th ese characters may be of value only among subgroups within the family and for distinguishing genera, and not for phylogenetic purposes in the whole family (Sheahan & Chase 1996).Several authors have tried to provide a natural system to divide the Zygophyllaceae into subfamilies, tribes and sections (Engler 1896a;1931;      El Hadidi 1975;Sheahan & Chase 1996;2000;Beier et al. 2003, see Table 1).Some of these studies were based on a small number of morphological characters or few numbers of the taxa.In the present study, a large number of characters were scored and numerical methods (UPGMA and PCO) were applied to study the relationships between taxa and to estimate the level of variation between them.UPGMA gives insight into the degree of similarity among the OTUs and whether they form groups/ clusters, and indicates the level of variation between species.On the other hand, PCO refl ects which characters are important on the axes, and indicates the signifi cant characters on the basis of the highest factor loading (Table 3).Th us it becomes clear which characters help diff erentiate between the groups and can be useful to distinguish taxa.Generally, our results demonstrated similarity between the UPGMA clustering and PCO ordination.In general, six taxonomic groups can be distinguished

Fagonia group
Th e taxonomy of Fagonia is very diffi cult mainly due to a high degree of plasticity and thereby adaptations to climatic conditions (Zohary 1972;Danin 1996).Ozenda & Quézel (1956) grouped the North African Fagonia species into four natural groups: (1) the F. kahirina-cretica-fl amandii group, (2) the F. arabica-bruguieri group, (3) the F. glutinosa-latifolia group, and (4) the F. microphylla-group.However , El Hadidi (1966) classifi ed Fagonia species into three natural groups based on pollen and trichome morphology.Also , El Hadidi (1972& 1974) classifi ed Fagonia-species into complexes.Beier et al. (2004) studied the phylogenetic relationships within genus Fagonia based on trnL & ITS DNA sequences and they did not support the natural groups of Fagonia, and showed that all species from the Old World, except Fagonia cretica form a weakly supported clade.Abdel Khalik & Hassan (2012) investigated seed and trichome morphology of Fagonia in Egypt and indicated that the seed and trichome morphology are useful in distinguishing the species and not supporting the natural groups.Th e results of both cluster and principal coordinated analysis presented here confi rmed that all species of Fagonia form a well-distinguished group, characterized by several characteristics: obconical fruit shape, smooth seed architecture, non-winged mericarp edge, ovoid seed shape, fi lament length (3-7 mm), seed size (1.8-4.2 x 1.3-3.7 mm), subprolate to prolate pollen shape, 1-3-foliolate leaf and glandular hair shapes.Within this group, we can show that Fagonia arabica, F. bruguieri, F. sinaica, F. latifolia, and F. scabra form a subgroup; F. cretica forms a subgroup and another one includes the remaining species of Fagonia.Th ese results are incongruent with those of Ozenda & Quézel (1956), El Hadidi (1966;1972;1974) and partially agree with both of Beier et al. (2004) and Abdel Khalik & Hassan (2012).

Zygophyllum and Peganum group
Van Zyl (2000) presented morphological analysis of the genus Zygophyllum in South Africa, and supported the divi-sion of Zygophyllum into the two subgenera Agrophyllum and Zygophyllum, which was contended by previous authors (Endlicher 1841;Van Huyssteen 1937).Van Huyssteen (1937) treated Z. coccinum, Z. album and Z. aegyptium as members of section Mediterranea Engl., and circumscribed Z. simplex and Z. decumbens in the section Bipartia Huysst.Moreover, she placed Z. dumosum in a separate section Alata Huysst.based on characters of fruit, stamens, and leaves characters.Sheahan & Chase (2000) analysed both rbcL and trnL-F sequences from 36 members of Zygophyllaceae and indicated that Zygophyllum is a polyphyletic genus, and show that, both Z. simplex and Z. decumbens were joined together in the same clade and the clade is a sister to the clade composed of Zygophyllum album, Z. coccinum, and Tetraena.
In some publications Peganum was placed in a separate family (Takhtajan 1969;El Hadidi 1975).Furthermore, Sheahan & Chase (1996) divided Zygophyllaceae into fi ve subfamilies, whereas Peganum, Malacocarpus and Nitraria were separated from Zygophyllaceae and close to members of Sapindales, and also they recommended of Peganaceae and Nitrariaceae.However, Takhtajan (1980) and El Hadidi & Fayed (1995) returned it to Zygophyllaceae.Th e results of both cluster and principal coordinated analysis confi rmed that the group of all species of Zygophyllum and Peganum harmala is a well-distinguished group, characterized by: short fl ower pedicle length (1-7 mm), schizocarpic fruits, 10 stamens, style length (1-2 mm), perennial to sub shrub plants, 2-foliolates, obtuse leaf apex, tricolpate pollen type, 5 locules per fruit.Within this group, we can show three subgroups: Zygophyllum simplex and Z. decumbens subgroup; Z. album, Z. aegyptium and Z. dumosum subgroup, and another subgroup includes Z. coccineum and Peganum harmala.Our results identify three branches that are congruent with the Zygophyllum sections proposed by Van Huyssteen (1937), and the results from previous studies of phylogenetic relationships by Sheahan & Chase (2000) and Beier et al. (2003).Moreover, the results indicate that Peganum harmala is not such an isolated species as has been thought (Takhtajan 1969;El Hadidi 1975;Sheahan & Chase 1996).Some characters that have been put advance as unique for it, such as tricolporate pollen grains, non-winged fruit, 5 styles, present of nectiferous disc, and endospermic seed, are also found in the other members of Zygophyllum, and these results are congruent with those of Takhtajan (1980) and El Hadidi & Fayed (1995).
Tetradiclis group Engler (1896a;1931) delimited Tetradiclis under subfamily Tetradiclidoideae with in Zygophyllaceae.However , El Hadidi (1975); Takhtajan (1980Takhtajan ( , 1983)); Sheahan & Chase (1996;2000), andBolous (2000) removed it from the family Zygophyllaceae and put it under a family Peganaceae.Moreover, Takhtajan (1996) treats it under a separate monotypic family Tetradiclidaceae.Morphologically, Tetradiclis tenella is quite distinct and easily identifi able.It is distinguished from other Zygophyllaceae by the exstipulate pinnatisect leaves, , fl ower small (1 mm) and tetramerous, 4 loculed fruit.It shares with genera of Zygophyllaceae the absence of fi lament appendages, the annual habit, the loculicidal capsule, fl attened leaf blade, and the obtuse leaf apex.Th e results of both cluster and principal coordinates analyses distinct the group from the other species by herbaceous stem, exstipulate leaves, leaves opposite below and alternate above, tetramerous fl ower, 4 loculed fruit.
Tribulus group Engler (1931) put Tribulus in tribe Tribuleae within the subfamily Zygophylloideae.However, El-Hadidi (1975;1977) proposed a new family Tribulaceae based on Engler's tribe Tribulae . El Hadidi (1978)  According to Hosni (1978), the most reliable and constant characters for Tribulus species, and hence of major systematic value, are those of the mature carpels and the size of the fl ower, while other characters such as the habit of the plant, hairiness, size of leaf, morphology of the style and stigma are of minor systematic value.Furthermore, Mohamed (2006) studied seed morphology, seed proteins and Iso-enzymes in Tribulus, and divided this genus into two main groups: (i) T. macropterus and T. mollis, and (ii) T. terrestris, T. bimucronatus, T. pentandrus, and T. kaiseri.Th e results of both cluster and principal coordinated analysis confi rmed that the group of all species of Tribulus is a well-distinguished group, characterized by: multi-foliolate leaf, opposite leaflet, fl at, 5-10 stamens, fi lament without appendages, oblate spheroidal pollen shape, pantoporate, coarsely-reticulate pollen schizocarpic fruit, 5 indehiscent mericarps.Within this group, we can show that T. bimucronatus and T. kaiseri form two subgroups: 1) T. macropterus, T. mollis, and T. terrestris and 2) another one includes T. pentandrus.Th ese results disagree with El Hadidi (1978) and partially agree with Mohamed (2006).
Seetzenia group Engler (1931) and El Hadidi (1975) placed Seetzenia with Fagonia and Zygophyllum in the subfamily Zygophylloideae within the family Zygophyllaceae.However, Sheahan & Chase (1996;2000) proposed a new subfamily Seetzenioideae.According to the cluster and principal coordinates analysis, Seetzenia lanata (Seetzenioideae) shows the largest distance from all other groups, and is distinct from the others by trifoliate leaves, apetalous fl owers, 5 stamens, 5 styles, tricolpate pollen type, oblong seed shape, crustaceous seed architecture.Th is result disagree with those of Engler (1931) and El Hadidi (1975) and agrees with Sheahan & Chase (1996;2000) for separating it as a subfamily.

Conclusion
UPGMA and PCO analysis can be used to study the morphological variation within the family to determine the discontinuities among genera and subfamilies.My results indicate that Zygophyllaceae are heterogeneous, including Peganum harmala which has been proposed to form a separate family.Th ere are many splits between subfamilies Zygophylloideae, Tribuloideae, Tetradiclidoideae and Seetzenioideae which seem to be distinct groups.However, there is also some degree of similarity among certain taxa of the subfamilies Balanitoideae, Peganoideae and Zygophylloideae.I consider Zygophyllum the most heterogeneous of the Zygophylloideae because I found that the taxa from this genus interspersed with taxa from Peganoideae (Peganum harmala), and this is congruent with the results of Sheahan & Chase (2000).Although this study has contributed new conclusions to literature, it is limited to the known genera in Egypt.A comprehensive study covering all genera would be necessary to make a more thorough classifi cation and it would be very useful for the further studies to use molecular data.

Figure 2 .
Figure 2. Scatterplot of the 29 OUTs plotted against the fi rst principal coordinate by the second principal coordinate.

Figure 3 .
Figure 3. Scatterplot of the 29 OUTs plotted against the fi rst principal coordinate by the third principal coordinate.

Figure 4 .
Figure 4. Scatterplot of the 29 OUTs plotted against the second principal coordinate by the third principal coordinate.

Table 2 .
are shown.Characters and character states used in morphometric analysis of the Zygophyllace

Table 3 .
Morphological characters showing highest factor loading on the fi rst three Principal coordinate axes.Th e shaded numbers mean characters with high factor loading > 0.6.