Tree Forensics: Modern DNA barcoding and traditional anatomy identify roots threatening an ancient necropolis

Ancient burial caves represent some of the most important sources of information on human history. In a world heritage site in Israel, such caves are under threat due to tree root growth penetrating from the ceilings and causing a risk of cave collapse. To facilitate historical conservation, while avoiding cutting of the mixed forest above the caves, we identified the tree species responsible for the damage, facilitating their specific cutting. Accurate identification of tree roots balances the needs to protect human structures on one side, while conserving the majority of vegetation on the other. Our approach is applicable to the management of ancient sites, as well


| INTRODUC TI ON
This investigation was conducted as part of conservation activities in the necropolis of Beit She'arim UNESCO World Heritage site in Galilee, Israel. According to Josephus Flavius, Beit She'arim ("Besara") was founded at the end of the 1st century BCE, as the administrative center of the estates of Queen Berenice in the Jezreel Valley (Avigad, 1971). By the 2nd century AD, the Sanhedrin (supreme Jewish council) moved to Beit She'arim and the town was mentioned in rabbinical literature as an important center of Jewish learning. Rabbi Judah the Prince (Yehudah HaNasi), head of the Sanhedrin and compiler of the Mishna, lived and was buried there. This led many Jews from all over the country and from the Jewish diaspora, from nearby Phoenicia to far-away Yemen, to come there in order to be buried in the vicinity of his grave. As a result, there are dozens of burial caves in this area (Avigad, 1971;Mazar, 1957;Rosenfeld, 1995;Weiss, 1992). The Jewish town was destroyed in 351 AD and the earthquake of 363 AD affected the area as well. The place was later occupied by a Byzantine village until the 7th century AD (Mazar, 1957;Meyers, 1999).
The soft chalk rock, into which the tombs were hewn, undergoes accelerated erosion by roots and water. The roots penetrate into the caves' cavities at a depth of about 20 meters. In recent years, roots have protruded into the burial caves through the soft limestone ceiling.
As root penetration into the necropolis intensified, maintenance of the ceiling structure was compromised, and the risk of collapse emerged.
The need arose to identify the plants responsible for this damage to the caves in order to find ways to prevent worsening of the situation.
Identifying organisms in the field has been the cornerstone of ecological research ever since its inception. For over 200 years, scientists have been collecting biological samples systematically in the field, assigning them to known taxa. In recent years, careful morphological and anatomical practices traditionally used for identification are being complemented by established genetic methods, taking advantage of the DNA sequence as a fundamental identifier. Sufficient variation between taxa in specific DNA regions laid the foundation to the barcoding approach. Since 2003, DNA barcoding has become a standard method for the reliable, relatively simple, identification of organisms (Hajibabaei et al., 2007;Moritz & Cicero, 2004). In fact, standard procedures for DNA barcoding have become common tools in genomics, with implications to molecular phylogenetics and population genetics (Hajibabaei et al., 2007).
In addition to the well-established use of DNA barcodes for the identification of microbiota such as bacteria and fungi, a need for the barcoding of land plants has also emerged (C. P. W. Group, 2009;Kress et al., 2005). The applicability of several plastid DNA regions has been tested (C. P. W. Group, 2009;Lahaye et al., 2008), as well as the whole-chloroplast genome sequence (Li et al., 2015), and the use of the nuclear internal transcribed spacer region (ITS) of the 18S-26S rDNA has proved most effective (Kress et al., 2005;Li et al., 2015). The ITS was the most commonly sequenced locus used for plant phylogenetics, showing high levels of interspecific divergence (Kress et al., 2005).
In a comparative study of seven candidate DNA barcodes, ITS2 showed the highest suitability for DNA barcoding applications, with 93% success rate in species identification among 4,800 plant species (Chen et al., 2010). Plant DNA barcoding is of particular interest when studying plant parts which are hard to identify, such as roots and ancient pollen. DNA barcoding has been used in archaeobotany to reconstruct past plant use, agriculture, diet, and vegetation (Schlumbaum et al., 2008). In this respect, the ITS markers are advantageous, since the high copy number of rRNA increases the probability of detection in small amounts and ancient DNA samples. Seeds found in Neolithic site in Italy were identified as Olea europea and Cornus mas, shedding light on the diet of its inhabitants (Gismondi et al., 2012). The utilization of DNA barcoding in the field of plant ecology is advancing fast. For example, over 1,000 Mesoamerican orchid species have been identified in a single study (Lahaye et al., 2008). Other biodiversity inventories have been compiled in additional projects (Valentini et al., 2009). DNA barcoding is also essential where only the remains of organisms are found, much like in modern-day forensics (Valentini et al., 2009). As in many other research fields in ecology and biology, plant DNA barcoding techniques were applied in Europe, North America, and Japan, i.e., to plant species of the temperate region. At present, woody plant species of the Mediterranean region are under-represented in DNA databases.
In spite of the recent advances mentioned above, the most obvious, classical way to identify plants is by using structural characteristics. Numerous atlases and books exist (Carlquist, 1988;Schweingruber et al., 2011;Crivellaro & Schweingruber, 2013;Fahn et al., 1986), some of which are specific for Mediterranean woody species (Crivellaro & Schweingruber, 2013;Fahn et al., 1986) that can be used in order to identify woody plant species. However, while studies on Mediterranean wood anatomy are prevalent (Crivellaro & Schweingruber, 2013;Fahn et al., 1986;Mrak & Gricar, 2016), very few deal with root structure (Rewald et al., 2012), and none deal with roots of Mediterranean species. That is not surprising, as root identification presents quite a few challenges (Maeght et al., 2013;Rutherford, 1980). While woody roots possess some distinct structural characters, the very thin roots are difficult to identify, although such identification is still possible (Rewald et al., 2012). Also, root morphology and anatomy vary with soil type and growth conditions. Moreover, root anatomy often differs from stem anatomy, and separate identification keys must be obtained. Nonetheless, root morphology and anatomy can be successfully used for identification (Uma & Muthukumar, 2014;Battipaglia et al., 2011). Interestingly, some identifying characters used include root odor, crystals, and mycorrhiza.
In this study, we combined ITS2 DNA barcoding with traditional morpho-anatomical identification methodologies to facilitate two distinct, yet intimately linked, purposes: (a) To identify roots protruding into ancient burial caves in order to pinpoint the species involved and settle conflicting needs for vegetation conservation and antiquities preservation, and (b) to add Mediterranean woody plant species to the existing barcoding database, thereby facilitating future research in the region.

| The research problem
In recent years, staff of the Israeli Nature Parks Authority which manages the Beit Shea'rim Park that includes the ancient burial caves noticed an ongoing process of plat root penetration through cracks in the cave ceilings ( Figure 1). In some places, parts of the caves' ceilings and walls were on the verge of collapse. To ensure the conservation of the heritage site antiquities, an experts meeting was conducted at the site, on 19 December 2017. An immediate need to remove the trees growing in the mixed forest on top of the necropolis was agreed upon. One of the suggested solutions was to cut down the trees and bushes, wait until the root dried and then fill the gaps with cement preserves. Yet, this need conflicted with Israel's nature conservation legislation, prohibiting the removal of mature trees anywhere without a special permit. The main challenge we were faced with was identifying which of all the plants are responsible for damaging the ancient caves. It was, therefore, decided that a selective tree removal (at the species level) will be performed, once the tree species, to which the roots belong to, are identified. A 'tree forensics' project was initiated, using both traditional morphoanatomical and modern DNA barcoding techniques.

| Site description
The Beit She'arim UNESCO World Heritage site is located in the Galilee region of Israel, 20 km West of Nazareth (32° 42′ 12.8124" N 35° 7′ 44.5656" E, elevation 120 masl). Climate at the site is thermo-Mediterranean, with an annual precipitation of ~530 mm, and average minimum, mean, and maximum temperatures of 16°C, 21°C, and 25°C, respectively. The soil is shallow rendzina, overlying a chalk-  (Avigad, 1971;Mazar, 1957). Aaronshon, a noted botanist, visited the area in 1904 and described it as a rich grazing ground covered by grasses with a handful of oaks. Many shepherds were bringing their sheep to this lush grazing ground from the whole region of Jezreel Valley (Aaronsohn, 1940;Patishi, 1983). At the time of the archaeological excavations, root penetration to the ancient caves was not mentioned at all. In pictures taken at the time, the area was clearly devoid of large trees (Avigad, 1971;Mazar, 1957). At the end of Avigad's book, he acknowledged the Gardening Department personnel and its director Mr. Yanai for planting decorative trees and shrubs in the area.

| Site recent history
The woody vegetation that currently surrounds the burial caves in-  (Avigad, 1971;Mazar, 1957). Aaronsohn, a noted botanist, visited the area in 1904 and described it as a rich grazing ground covered by grasses with a handful of oaks. Many shepherds were bringing their sheep to this lush grazing ground from the whole region of Jezreel Valley (Aaronsohn, 1940;Patishi, (1983). At F I G U R E 1 Penetration of plant roots through the cave limestone ceilings of the Beit She'arim necropolis, Galilee, Israel. Photos taken by the authors in January 2018. (a) Entrance to one of the burial caves, showing the Mediterranean maquis ~5 m above the cave and on the hillslope, with Cupressus, Quercus, Pistacia, and other trees; (b) the view inside the cave with sarcophagi, roots can be seen penetrating from the ceiling; (c) roots of Pistacia palaestina growing through the limestone on top of the entrance of another cave; (d) view of a cave ceiling affected by roots; and (e) close-up of a cave ceiling, showing several root penetration points the time of the archaeological excavations, root penetration to the ancient caves was not mentioned at all (Avigad, 1971;Mazar, 1957).
In an aerial photograph ( Figure 2) taken in 1944 of the site, it can be clearly seen that the area was devoid of large trees. At the end of Avigad's book (Avigad, 1971), he acknowledged the Gardening Department personnel and its director Mr. Yanai for planting decorative trees and shrubs in the area. The trees are now an integral part of the park landscape and highly valued by the staff and visitors.

| Sample preparation and morpho-anatomical root identification
Roots (5-7 cm long; 5-15 mm in diameter) were sampled and immediately placed in 50-ml plastic tubes and brought to the laboratory.
The fresh samples were photographed for morphological identification and transferred to 70% alcohol for storage till sectioning.
Hand cross-sections were made with razor blade. The sections were stained either with TBO (O'Brien et al., 1964) or with Safranin-Alcian blue (Shtein et al., 2020). The sections were analyzed under a stereo microscope (Olympus SZ2-ILST, Japan) and photographed using the microscope camera (Olympus LC20, Japan).
The preparation of species identification key was the most time-consuming part of the morpho-anatomical identification. At first we made the mistake of sampling the roots in the field, in proximity to the tree, and subsequently we realized that all the oak roots we sampled were in fact Pistacia roots. Thus, caution must be made when using field-collected samples for root identification keys. Our current key is at the genus level, but in the future a more extensive key could be made. Once the identification key has been prepared, the identification was straightforward. As we used free hand sections and a simple double-staining procedure, the identification took about 15 min per sample. However, caution must be used, as root structure might sometimes vary with habitat and the key should be adjusted accordingly (Maeght et al., 2013). Additionally, as a broad range of characters is used for identification, a good knowledge of plant anatomy is required, as compared to a less challenging identification using molecular markers. Nonetheless, in a given habitat with a known range of species, morpho-anatomical identification is possible and could give good results.

| Site of reference plant material
To provide verifiable DNA barcodes, mature leaves of 19 Mediterranean woody plant species were collected from identified specimens in the Tel-Aviv University Botanical Garden on 21 March 2018. A sample of 5-10 leaves was collected from an individual plant of each species. Leaves were dried at 60°C for 48 hr prior to preparation for DNA extraction. In addition, tree roots were collected from potted plants at the Tel-Aviv University Botanical Garden for the preparation of a morpho-anatomical identification key.

| DNA extraction, amplification, and sequencing
Among multiple barcoding approaches, we chose the ITS2 marker, i.e., part of the nuclear internal transcribed spacer region (ITS) of the 18S-26S rDNA, which has proved most effective for plant phylogenetics, showing high levels of interspecific divergence (Chen et al., 2010;Kress et al., 2005;Li et al., 2015). We applied this approach to identify roots at the species level. DNA-based identification of plant individuals is also possible, however, requires a deeper investigation of multiple loci, which was beyond the scope of this project.
DNA extraction was based on Wang, Qi & Cutler (1993) method modified to woody plant species. Briefly, dry samples were grounded using a mixer mill (MM 400; Retsch GmbH, Haan, Germany) for six minutes at 20 frequency/second using two 5-mm stainless steel milling balls (Retsch). NaOH (400 µl 0.5 N) + 1. Full ITS sequences were subjected to Hidden Markov Model (Keller et al., 2009) to remove the conserved 5.8S and 28S sequences and retain the correct position of the ITS2 within the full ITS sequence (Koetschan et al., 2012). Species identification was evaluated with BLAST1 method (Howard et al., 2008) on three different parts of the ITS sequence: full ITS sequence, primary structure of the ITS2, and secondary structure prediction of the ITS2 using tools from the ITS2 database (Koetschan et al., 2012;Schultz et al., 2005). Blast results were typically with 96%-97% similarity, and always > 90% similarity. All verified ITS2 sequences were provided in Table S1.

| DNA identification of the trees responsible for the problem
Comparison of ITS2 DNA sequences from 19 woody plant species of the Mediterranean maquis against the NCBI database and the ITS2 database (http://its2.bioap ps.bioze ntrum.uni-wuerz burg.de/) yielded mixed results (Table 1). Only four and ten species were accurately identified by the NCBI and ITS2 databases, respectively.
Additional eleven and seven species were misidentified as related species of the same genus (i.e., identified correctly at the genus level, but not at the species level) by the respective databases. For example, Quercus calliprinos and Q. boissieri were both identified by the NCBI database as Q. poilanei (Table 1). This result demonstrates the importance of local projects, and specifically with endemic species, whose sequences are rarely studied. Notably, both Quercus calliprinos and Q. boissieri belong to rather distant clades from Quercus poilanei within the phylogeny of Quercus (Hipp et al., 2020). Rhamnus, Crataegus, Clematis, and Tamarix presented a similar case. Last, four and two species were not identified at all. The ITS2 database performance was consistently better than that of the NCBI. In the example of Quercus calliprinos, the misidentification of the ITS2 database as Q. coccifera was very close phylogenetically, as these two species diverged from each other only during the Pliocene, ~3 mya (Hipp et al., 2020). Therefore, we developed novel ITS2 DNA barcodes for 10 species which were missing from both databases. These sequences, along with the sequences of the other nine species, were uploaded to the ITS2 database, adhering to the database guidelines. Among the identified tree species, the two Pistacia species allowed for investigation of DNA polymorphisms among roots of different individuals. Within an ITS fragment sequence of 425-435 bp, we identified six single nucleotide polymorphisms (SNPs) in Pistacia atlantica, which were mostly related with one specific samples. Interestingly, in Pistacia palaestina, there was only one SNP in this fragment. The final sequences that were uploaded to the database were those representing the majority of samples from each species.
Next, root samples were taken in situ at Beit She'arim ( Figure 1).
Roots were sampled at three-five locations in each of three burial caves in the necropolis: The Sidonian cave, the Museum cave, and the cave of Rabbi Judah the Prince (Yehudah HaNasi). Each root sample was split into two fragments, one for DNA identification, and the other for morpho-anatomical identification.

| Morphological identification of the trees responsible for the problem
For the morpho-anatomical identification, the efficient approach was to distinguish several unique traits for each genus, rather than making a large key for each species. Root surface color, periderm structure and xylem vessel distribution were among the most useful traits, showing large and notable differences between the species, with fibers, ray structure, and secretory cavities used as auxiliary  Table 2). Three other roots were identified as the gymnosperms (two Cupressus sempervirens, and a Pinus halepensis), and other two as perennial herbs (Capparis spinosa and an unidentified member of the Cucurbitaceae). Note that Cupressus sempervirens was existing in the databases, and did not require novel sequencing, unlike some of the species in Table 1

| D ISCUSS I ON
The integrity of the bedrock structure was affected by both the di- transport, e.g., using H 2 18 O, could potentially improve the resolution of our study to the individual tree level (Maeght et al., 2013;Schenk & Jackson, 2005), yet require far more resources. Our choice of using the (ITS)2 marker system proved well in our case, in spite of its known problems, such as polymorphisms within species (observed here for Pistacia atlantica, but not for P. palaestina) and contamination by mycorrhizal DNA. It is possible that the spatial homogeneity of the samples contributed to a reduced level of such polymorphism that would otherwise complicate the identification.
In addition, a visual inspection showed that mycorrhizal colonization of our roots was lower than that of tree roots in wetter biomes, e.g., in a temperate forest (Rog et al., 2020). Interestingly, ITS markers were also those which enabled the identification of ancient cedar wood in Jerusalem's temple mount (Rogers & Kaya, 2006 (Schenk & Jackson, 2005). In mid-latitudes specifically, the occurrence of roots deeper than 2 m depends on soil structure (Schenk & Jackson, 2005). In places where shallow soil is underlain by weathered bedrock, roots tend to grow into the rock, in turn extracting water and nutrients from it (Schwinning, 2010). In our site, all three predisposing conditions exist: a shallow Rendzina soil covers a soft chalk-limestone bedrock, and rainfall is restricted to October-April, with a ~6 months zero precipitation dry season. Furthermore, woodland plant species, such as in our case, tend to allocate significantly more biomass to roots, than species of the temperate, boreal, or tropical forest (Poorter et al., 2012). Interestingly, unlike in natural caves (Adams et al., 2019;Howarth et al., 2007;Jackson et al., 1999), part of the roots in our study did not seem to extend into the space of the caves (Figure 1). One hypothesis is that the drier conditions in our site, compared with the previously reported settings, prohibited water, and nutrient uptake from the cave itself, in contrast to the rock environment (Schwinning, 2010).
It has been suggested that rooting depth might be a more important plant trait than root biomass, shaping plant communities by hydraulic lift (Maeght et al., 2013), which has already been reported in a Mediterranean pine species (Penuelas & Filella, 2003). Our findings of Pistacia, Cupressus, Cercis, and Pinus roots at 5 m depth below surface shed new light on the eco-physiology of these important taxa.

| CON CLUS IONS
The efforts by botanists in combining novel and traditional methods for identification of subterranean roots, disconnected from the plants themselves, enabled the devising of management procedures that served both the antiquities preservation and landscape maintenance. The DNA barcoding applied here to the local Mediterranean vegetation identified certain lacunae in the database, which have been filled through analysis of verified plants from the botanical garden. Additionally, as a broad range of characters is used for the morpho-anatomical identification, a good knowledge of plant anatomy is required, as compared to a less challenging identification using molecular markers. Nonetheless, in a given habitat with a known range of species, morpho-anatomical identification is possible and could give good results. the Tel Aviv University Botanical Garden for providing tree samples.

AUTH O R CO NTR I B UTI O N S
The study was initiated by IC, IS, and DBY. AE, IS, and TK designed the research. GJ and IR conducted the DNA barcoding study. IS conducted the morpho-anatomical study. IC and DBY were in charge of the archaeological conservation and information. TK wrote the manuscript with contributions from all authors.