Multivariate Analysis of Lower Cretaceous Monosulcate Pollen from Central-Western Argentina

ABSTRACT Monosulcate pollen grains are an important component of the Mesozoic palynofloras around the world. These type of grains are produced by several groups of plants and because of the morphological similarities among them, their taxonomic determination is difficult. Currently, there is scarce information about morphological differences of monosulcate pollen grains and this makes their botanical affinity in fossil specimens challenging. In this work, we analyzed statistically the morphological features of monosulcate fossil pollen grains recovered from six palynological assemblages from the Lower Cretaceous Lagarcito Formation of central-western Argentina. We used our own fossil dataset and data from the description of monosulcate pollen grains published from Cretaceous cycadalean in situ male cones from Patagonia. Furthermore, morphological variables from selected living species were added to support the analysis. According to our results, two main groups with clear morphological features were characterized among fossils. The use of an extant dataset allowed us to support the exine sculpture as an important variable when characterizing fossil pollen grains. Finally, our fossil dataset showed similarities with the in-situ pollen grains from the cycadalean cone species Androstrobus munku and Androstrobus patagonicus. This work is the first attempt to identify different morphological types among fossil monosulcate pollen grains recovered from the Lagarcito Formation of central-western Argentina. Our results will contribute to elucidate morphological variations in fossil pollen grains and may improve future taxonomic studies.


Introduction
Monosulcate pollen grains are known in the fossil record since the Late Paleozoic and have been widely reported in Mesozoic deposits around the world (e.g. Frederiksen 1980;Pole and Douglas 1999;Taylor et al. 2009). These kind of pollen grains were produced mainly by the Cycadales, Bennettitales and Ginkgoales (Taylor et al. 2009;Christenhusz et al. 2011) and they are related to several groups of plants. Due to their morphological similarities, the distinction between them has always been difficult (Frederiksen 1980). Cretaceous representatives of cycads and bennettitaleans from Argentina are only known from Patagonia (C uneo et al. 2010). Fossil male cones Androstrobus, attributed to the Cycadales were recorded and described from the Baquer o Group (Lower Cretaceous) by Archangelsky and Villar de Seoane (2004). The first record of Cycadales in Patagonia begins in the Triassic and extends up to the Oligocene (Passalia et al. 2010). In this region, the group was highly diverse and included several records represented by trunks, leaves, and pollen cones (e.g. Archangelsky and Villar de Seoane 2004;Villar de Seoane 2005;Del Fueyo et al. 2007;C uneo et al. 2010;Passalia et al. 2010;Mart ınez and Artabe 2014;Mart ınez et al. 2016Mart ınez et al. , 2018. Plant communities with cycads (zamiaceous forms), conifers, and occasional palms are known since the Early Cretaceous (Mart ınez et al. 2018). After the Eocene, there are no fossil records of Cycads in Patagonia (Wilf et al. 2016). Mart ınez et al. (2018) discussed their extinction, which could have been influenced by different abiotic and biotic factors occurring since the Cretaceous.
On the other hand, bennetitaleans from the Lower Cretaceous of Argentina, are represented only by five leaf genera, one genus of scale leaves, and another one of seed cones. All of them recorded from the Baquer o Group, and the Springhill and Kachaike formations but without any recovery of in-situ pollen grains (e.g. Men endez 1966; Archangelsky and Baldoni 1972;Baldoni 1974;Villar de Seoane 1999, 2003. Lastly, the Ginkgoales are documented in the Lower Cretaceous of Argentina with the occurrence of an important diverse association of foliar cuticles and female fructifications (Karkenia incurva Archangelsky), but this group was not recognized in the palynological record (Del Fueyo and Archangelsky 2001;Del Fueyo et al. 2007;Villar de Seoane and Archangelsky 2014). Macrofossil plant records producing monosulcate pollen grains were more abundant than the monosulcate pollen grains in Mesozoic sediments (Frederiksen 1980). However, monosulcate pollen grains are very well represented in the palynoflora of the Lagarcito Formation from central-western Argentina. Mego et al. (2019) concluded that the palynoflora of this unit was dominated in nearly all the assemblages by representatives of the gymnosperms (ranging from 28 to 75% of the total sum). Among them, monosulcate pollen grains reached up to 50% of the total composition in many assemblages. The monosulcate fossil pollen grains of Cycadopites-type are often indistinctly assigned to Cycadales, Bennettitales, and Ginkgoales due to the lack of systematic features to distinguish them (Cantrill and Poole 2002). Moreover, the botanical affinity of monosulcate pollen grains is difficult to determine when they are found dispersed in sediments (Archangelsky and Villar de Seoane 2004).
Morphological characters play an essential role in taxonomic identification, classification, and species delimitation (Herber 2002). Pollen morphological characters have been widely used in defining evolutionary trends (Lumaga et al. 2006;Gomankov 2009;Zare et al. 2014). A morphological analysis including quantitative variables was recently applied in fossil pollen grains of two species of Proteaceae from the Cretaceous-Paleogene boundary, resulting in a clear taxonomic differentiation (C ardenas et al. 2019). However, studies dealing with morphological analyses of fossil spores or pollen grains are still scarce (e.g. Osborn and Taylor 1995;Rull 2001;Duarte et al. 2014). Only a few works have attempted to clarify morphological differences among monosulcate pollen grains of Cycadales and Ginkgoales (Sahashi and Ueno 1986;Dehgan and Dehgan 1988;Tekleva and Krassilov 2009).
In this research, we focused on the orders Cycadales, Bennettitales, and Ginkgoales, mainly widespread during the Mesozoic, as the possible producers of monosulcate pollen grains recovered from the Lower Cretaceous Lagarcito Formation. The main goal of our study is to analyze the dataset of fossil monosulcate pollen grains (dispersed pollen grains from Lagarcito Formation and published data from in-situ Cycadalean pollen from fossil male cones from the Anfiteatro de Tico Formation, Baquero Group of Aptian age) using morphological features and modern statistical techniques.

Extant pollen grains
For the purposes of the present analysis, we collected our own material of Cycas revoluta Thunb (two samples obtained from different cones) (Plate 2 1-4), one representative of Zamia pungens L. f. ex Aiton (Plate 2 5-8), and two different samples from Ginkgo biloba L. (Plate 2 1-12) from herbarium specimens. Extant monosulcate pollen grains were obtained from the Instituto de Bot anica Darwinion and from the Museo Argentino de Ciencias Naturales 'Bernardino Rivadavia', Buenos Aires, Argentina. The material was processed following standard methods for pollen extraction of acetolysis (Faegri and Iversen 1989) after retreatment with KOH 10% (100 ) at the Laboratorio de Paleoecolog ıa del Cuaternario IANIGLA, CCT-CONICET, Mendoza. Extant species were selected due to their worldwide distribution and their conservative morphological features.
not considered in the analysis. We focused our study on the fossil dataset from the Lagarcito Formation and used extant specimens as a support dataset of our results. We added published data from in-situ cycadalean pollen of three Androstrobus species: Androstrobus munku, A. patagonicus and A. rayen described by Archangelsky and Villar de Seoane (2004) to our fossil dataset. Additionally, we examined an extant monosulcate pollen grain dataset (Cycas revoluta, 60, Zamia pungens, 30, Ginkgo biloba 60 specimens) and registered the same morphological variables that we previously observed on fossil specimens.
We performed a cluster and discriminant analyses to detect their morphological affinities. We coded a set of morphological variables observed in our fossil pollen grains ( Figure 2). We recorded information of the following variables: shape of the pollen grain, exine sculpture, exine thickness, colpus, margo, equatorial and polar diameters (Appendix 1). We followed the terminology proposed by Punt et al. (2007) and Archangelsky and Villar de Seoane (2004) for coding some morphological characters used in the Supplementary material 2. We used numeric categories based on the frequency interval to reduce the complexity of the analysis due to data variation of the equatorial and polar diameters and exine thickness (Supplementary material 1, Figure 1A-C). We constructed our dataset based on morphological features and their states (Supplementary material 2).
We performed a principal component analysis with seven morphological features to obtain an integrated response from the variables. This technique allowed us to explore the ordination of the fossil data and reduce the dimensionality of the original matrix.
We made a cluster analysis with the complete set of variables to divide the set of fossil data into two subsets according to their similarities for detecting morphological variations in the studied specimens. We used hierarchical average linkage method with a Euclidean distance measure. Cluster analysis consists of classifying a set of individuals based on the observed values of the variables (Becking 1957). Through the application from a similarity criterion, you can determine how close or similar they are with each other (Klastorin 1983). In this sense, we looked to distinguish similar morphologies of the analyzed specimens and explored them according to the obtained dendrogram. Additionally, we performed an analysis of variance (ANOVA) using the continues values (in micrometers) of polar and equatorial diameters to support the cluster differentiation. We tested the normality of these data using the Shapiro-Wilk's test and compared the mean values by Tukey's test (ὰ ¼ 0.05).
We conducted a discriminant analysis to observe the separation of groups. This analysis examines differences between the previously obtained groups from the cluster analysis to detect the variables that would exert the greatest influence on the grouping discriminant sub-spaces.
Finally, we added the extant monosulcate pollen grain dataset to our analysis. We considered the weight of each variable after a discriminant analysis and used the most relevant features to analyze a combined dataset of fossils and extant specimens for detecting morphological similarities.
We performed the data analyses and created the graphs using R version 4.0 (R Core Team 2019). The script is available in the Supplementary material 1.

Results
The principal component analysis showed that the two first components explained 70.5% of the total data variation. The first principal component explained 43.5% and the second one accounted for 27%. The morphological variables that mostly contributed to the ordination of the fossil features throughout the main components were polar diameter, shape, and equatorial diameter (Supplementary material 1, Table 1).
The cluster analysis allowed the identification of two main morphological types and the discriminant analysis indicated the shape and equatorial diameters as the most relevant variables (Supplementary material 1, Table 2). One of the groups was characterized by a morphological type with a predominant ellipsoidal to fusiform shape. The other group included a morphological type with a predominant circular shape (Figure 3). There were significant differences among the two groups for polar (p < 0.001) and equatorial diameters. One group was characterized by mean polar and equatorial diameters of 41.71 and 16.7 lm, respectively. On the other hand, the second group was characterized by mean polar and equatorial diameters of 27.68 and 12.61 lm, respectively (Supplementary material 1, Figure 2A and B).
The discriminant analysis using the morphological features for the three recognized extant species (Cycas revoluta, Zamia pungens and Ginkgo biloba) indicated the variables exine sculpture and exine thickness as the most relevant (Supplementary material 1, Table 3). The cluster analysis of both fossil and extant datasets did not achieve significant results to clearly support affinities for grouping fossil, and extant specimens. The discriminant analysis indicated the variable exine thickness as the most relevant in the separation of fossil monosulcate pollen grains (mostly with an exine thickness of 1 lm) and living specimens (mostly with an exine thickness between 2 and 3 lm) (Supplementary material 1, Table 4). Therefore, fossil and living specimens are grouped in two main groups without any clear affinities (Supplementary material 1, Figure 3). However, the second relevant variable resulted to be the exine sculpture. We analyzed this variable separately and we determined the percentages for each coded state in fossil and extant specimens. In the examined extant specimens, we found that the psilate and psilatescabrate exines correspond mostly to Cycadales, in this case Cycas revoluta and Zamia pungens. On the other hand, scabrate exine sculpture was mostly identified in Ginkgoales, represented by Ginkgo biloba. Regarding the fossil specimens, more than 98% presented exine sculpture which was psilate or psilate-scabrate ( Table 1). The addition of the three species of the genus Androstrobus showed similarities of our fossil pollen specimens to A. munku and A. patagonicus (Figure 4). This differentiation is supported by the discriminant analysis that indicated as relevant the variables shape, exine sculpture, and equatorial diameter (Supplementary material 1, Table 5).

Discussion
Pollen morphology within a single plant species has been regarded as a relatively constant trait, and it has been used as a taxonomically important character (Aguilar-Garc ıa et al. 2012). In this work, we analyzed the morphology of monosulcate fossil pollen grains from the Lower Cretaceous Lagarcito Formation. We examined a set of morphological variables clearly observed in fossil pollen grains, besides we added morphological variables from selected living species to support our analysis (Figure 2). According to the weight of the variables after the discriminant analysis, we identified two morphological types of fossil monosulcate pollen grains characterized by the shape and equatorial diameter (Figure 3). Pollen shape and size have been recognized as significant variables in determining botanical affinities (Smitha et al. 2018). Most of our fossil specimens presented ellipsoidal to Figure 3. Dendrogram showing the main monosulcate fossil pollen morphological types. One group was characterized by a predominant ellipsoidal to fusiform shape and equatorial diameter of more than 10 lm (red bars). The other group included a predominant circular shape and equatorial diameters of less than 10 lm (green bars).  fusiform or circular to sub-circular shapes (Plate 1). Equatorial diameter also indicated variations in the fossil pollen grains. One group presented less than 10 lm and the other one more than 10 lm up to 37 lm of equatorial diameter. Comparative pollen morphology of extant and fossil species was used to assign pollen grains to extinct gnetophytes for understanding the evolution of some traits (Tekleva and Krassilov 2009). For example, fossil pollen grains of Picea sp. were analyzed using comparative characters based on a large assemblage of modern reference pollen with relevant results (Lindbladh et al. 2002). Many authors (Grayum 1986;Dehgan and Dehgan 1988;Ueda and Tomita 1989;Rull 2001;Herber 2002;Kambhar et al. 2017;Ullah et al. 2018 among others) analyzed size and shape of pollen grains of extant or fossil specimens but separately. However, only a few such as Tekleva and Krassilov (2009) incorporated extant and fossil specimens together in a morphological analysis. Results on morphological characteristics of extant material of Cycadales and Ginkgoales analyzed herein, indicated similarities with the previous results obtained by Korszun and Klimko (2014) and Sahashi and Ueno (1986). The microsporangia and pollen morphology of Ginkgo biloba studied by Korszun and Klimko (2014) showed a similar variability regarding shape and exine sculpture to our examined extant specimens. Moreover, these authors observed that desiccated pollen grains resulted in a boat-shaped grain with a single longitudinal furrow while the fully swollen grains were almost spherical with a rounded or oval germinal aperture (Korszun and Klimko 2014) which is also coincident with some of our fossil pollen grains (Plate 1 9-12). Sahashi and Ueno (1986) investigated extant pollen grains of Ginkgo biloba and Cycas revoluta using scanning electron microscopy and concluded that pollen grains of both species presented an oblong shape with monosulcate apertures, and the swollen grains showed a spherical shape with a large and rounded germinal aperture. Our extant pollen grains of Ginkgo biloba (Plate 2 9-12) presented a similar shape to the specimens described by Korszun and Klimko (2014). Pollen grains of extant species of Cycadales are boat-shaped and bilaterally symmetrical. They range from narrow to wide elliptical or subcircular when viewed distally, and have a psilate exine surface (Dehgan and Dehgan 1988). Pollen grains of extant Cycas revoluta are morphologically similar to the extinct bennettitalean Cycadeoidea dacotensis. They have a small pollen grain size (25-40 lm), symmetrical with an elongate distal sulcus and a lamellate and thin exine (0.5 lm) (Millay and Taylor 1976). The size of extant angiosperm pollen grains resulted in a useful character in family or genera identification (Lindbladh et al. 2002;Zare et al. 2014;Khan et al. 2018;Ullah et al. 2018). Morphological features were relevant in the classification and identification of 35 species of extant gymnosperms, and pollen shape proved to be one of the most important variables in the differentiation (Khan et al. 2018).
Our analysis of fossil and living specimens together did not indicate shape as a relevant morphological variable. The variable exine thickness resulted as the most relevant in the separation of fossil and extant specimens. However, this differentiation was not relevant in our analysis (Supplementary material, Figure 1, Table 4) because the majority of fossil specimens had an exine thickness of 1 lm. In the extant specimens, exine thickness presented values of more than 2 lm. The second most important variable was the exine sculpture that was relevant to detect similarities between both groups. Our fossil specimens presented a high percentage of psilate and psilate to scabrate exines. Only a few specimens showed scabrate exine (Table 1). According to our extant specimens, a psilate exine was frequently observed in Cycas revoluta, psilate to scabrate in Zamia pungens, and scabrate exine in Ginkgo biloba.
The palynoflora of Lagarcito Formation was dominated in nearly all the assemblages by representatives of gymnosperms, and dispersed monosulcate fossil pollen grains of Cycadopites-type were recognized as the principal component (82 to 100%) of the palynological associations (Mego et al. 2019). Fossil evidence shows that this group had a broad distribution during the Cretaceous in Argentina (e.g. Archangelsky 1963Archangelsky , 1966Men endez 1965;Petriella 1969Petriella , 1978Baldoni 1977;Villar de Seoane 1997;Del Fueyo et al. 2007). Archangelsky and Villar de Seoane (2004) described in-situ cycadalean pollen of three Androstrobus species: Androstrobus munku, Androstrobus patagonicus, and Androstrobus rayen, from the Lower Cretaceous of Patagonia, Argentina, so far the single record of pollen bearing structures of cycads in Argentina. They recorded a variety of pollen cones found in the same bed, with different vegetative organs, leaves, and fronds that were referred to the cycads on the base of their cuticular structure. Androstrobus munku pollen grains are ellipsoidal to fusiform, up to 39 mm long and 28 mm wide, and present a sulcus slightly thickened which extends for most of the pollen grain length. The exine is smooth to slightly scabrate (1-1.5 mm). Androstrobus patagonicus pollen grains are monosulcate, up to 34 mm long and 28 mm wide, circular to subcircular in polar view, and with smooth exine (Archangelsky and Villar de Seoane 2004). When we analyzed our fossil dataset from the Lagarcito Formation together with the Androstrobus species, we found high morphological similarities especially considering the exine sculpture, shape, and equatorial diameter (Figure 4). Around 72.5% and 23.20% of our fossil specimens shared the above mentioned morphological characters with A. munku and A. patagonicus, respectively (Supplementary material 1, Table 5). These statistical results clearly link the dispersed material of monosulcate pollen grains of the Albian Lagarcito Formation with the material of the in-situ monosulcate pollen grains found in cycadalean cones of Androstrobus from the Aptian Anfiteatro de Tico (Archangelsky and Villar de Seoane 2004). These authors concluded that the Cycadales were not diverse during the Early Cretaceous, and had a wide geographical distribution, including Patagonia where they were abundantly represented. Thereby, most of monosulcate pollen grains recovered from the Lagarcito Formation from the Albian of central-western Argentina, could be assigned to the order Cycadales. Thus, in spite of the lack of macrofossils from the Lagarcito Formation, the present results from the statistical analysis support the presence of Cycadaleans in the San Luis Basin during the Cretaceous.

Conclusions
We presented a morphological analysis of fossil monosulcate pollen grains recovered from the Lower Cretaceous Lagarcito Formation. This is the first attempt to analyze the most significant morphological traits to discriminate monosulcate fossil pollen grains from South America. A high percentage of the analyzed monosulcate fossil pollen grains exhibit psilate to psilate-scabrate exines. This character appears to be present mostly in the extant Cycadales species in this work. Only a few pollen grains presented scabrate exines. The exine sculpture seems to be an important variable to characterize fossil monosulcate pollen grains. An important percentage of the fossil specimens from the Lagarcito Formation could be related to the genus Androstrobus due to their morphological similarity, especially to A. patagonicus and A. munku, all of them belonging to the Cycadales. In this way, the presence of this group of plants provides useful information for an accurate reconstruction of the vegetation during the Early Cretaceous of central-western Argentina. Our results are relevant in order to characterize morphologically monosulcate fossil pollen grains and represent a contribution to systematic studies.