1. Introduction
The Brassicaceae are a large family of flowering plants with a circumpolar distribution, comprising 338 genera and 3700 species, including several major agricultural crops [
1,
2]. The type genus,
Brassica, is by far the most important economically, and is currently grown in over 150 countries as a vegetable or oilseed crop [
3]. Six
Brassica species dominate current global production, viz.:
Brassica oleracea L.,
Brassica rapa L.,
Brassica napus L.,
Brassica carinata A. Braun,
Brassica nigra (L.) K. Koch, and
Brassica juncea (L.) Czern [
4]. Polyploidy has played a key role in the evolution of the cultivation of
Brassica species, with allopolyploid hybridization between the three diploid species (
B. rapa, AA,
B. nigra, BB, and
B. oleracea, CC) giving rise to three allotetraploids (
B. juncea, AABB,
B. napus, AACC, and
B. carinata, BBCC—[
5]). The domestication of these species is thought to have originated around the Mediterranean region of Europe [
6,
7,
8]. The family also contains several genera that feature as minor crops, including (among others):
Raphanus (
R. sativus L., the radish),
Armoracia (
A. rusticana G. Gaertn., B. Mey. et Scherb., horseradish),
Camelina (
C. sativa (L.) Crantz, gold of pleasure),
Nasturtium (
N. officinale R. Br., watercress) and
Sinapis (e.g.,
S. alba L., white mustard). These species have a long history of cultivation, particularly the Brassicas and
Sinapis [
9]. The Italian flora contains 70 genera with 311 species of Brassicaceae, including all of the cultivated species named above [
10].
Archeological information can provide useful insights into the timing, context, and nature of domestication events, but first requires appropriate methods for the reliable identification of old, often degraded plant materials. This problem can be addressed by reference to diagnostic phenotypic and molecular information sources. The fruits of Brassicaceae are widely used to perform species diagnosis within the family and can be a useful source of information from archaeological settings when present. Most of the taxa in the family contain siliques and silicle/silicula. The distinction of many of these species is often based on the length/width ratio of the siliques, or on whether they possess bivalved capsules that open from below [
11,
12]. However, some taxa possess fruit structures of a different kind, such as the lomentum type found in
Raphanus [
13]. Here, too, the features are often useful for the identification of taxa. The seeds are probably the most useful in an archaeological context for diagnostic purposes, and range in shape from spherical to flattened, feature diagnostic sculpturing of the testa, and are often the dispersal unit [
11].
European records of Brassicaceae associated with archaeological contexts date back to the Neolithic period, after which they spread geographically, with some records ambiguously listed as weeds s.l., but others with clear evidence of cultivation, primarily as oil crops or vegetables [
9,
14]. The most frequent and reliable finds for notable economic taxa have been seeds of
Brassica,
Sinapis, and
Camelina, although the certainty of identification can be compromised if the condition of the seeds is poor, especially if they are charred [
9]. However, when in good conditions, these features are diagnostic for several taxa [
15]. Given the variability in the condition of archaeological materials, there is a need to adopt multiple approaches for species diagnosis, especially when handling partly degraded materials.
Currently, in Italy, more than 700 sites have been investigated from an archaeobotanical viewpoint [
16,
17]. In the prehistorical period, Brassicaceae seeds were practically irrelevant in the sites of northern Italy [
18,
19], whereas in a synthesis on the seeds/fruits of food plants for the Roman period in northern Italy, only a few seeds of
Brassica sp. pl. and
Sinapis alba were found in 4 sites out of 70 (ca. 6%), and never in funerary-ritual contexts [
20,
21]. Based on the preliminary data of an analogous synthesis for the Middle Ages/Renaissance period [
22], remains of the taxa examined were found in 19 sites out of ca. 50 (about 40%), of which 16 were in the Emilia-Romagna region.
An archaeological site in the historic centre of Ferrara (Emilia-Romagna, Northern Italy—
Figure 1) has revealed deposits, datable to between the Middle Ages and the Renaissance, rich in botanical records, which are still being studied. This site contains an unusually large quantity of Brassicaceae seeds (in particular of
Brassica/Sinapis type), and was therefore deemed likely to present challenges for reliable species identification. Similarly, another coeval site in Lugo (Ravenna, Emilia-Romagna—
Figure 1) also returned a good number of
Brassica/Sinapis seeds and was also considered a suitable site to investigate the fidelity of species assignation.
In this study, the seeds were investigated by morphometric and genetic analyses. Furthermore, a historical and ethnobotanical interpretation of the taxa was performed using written sources and the exsiccata present in the oldest Italian herbaria.
2. Results
The Emilia-Romagna archaeological sites dating from the 6th to the 17th century AD included 19 contexts in which seeds were found, although in the majority of these (about 58%), the seeds were present at concentrations below 1 seed/L. Furthermore, in the vast majority of the contexts (about 84%),
Brassica and
Sinapis seeds/fruits accounted for <1% of all the seeds/fruits recovered. In 7 contexts (nos. 3, 7, 12a, 13, 15, 16a and 16b), the concentration of seeds/fruits varied between 1 and 13 seeds/L, although in all these cases, the Brassicaceae seeds represented ≤1% of all the seeds found. The sole context that showed a notable divergence from these trends was the brick tank in Ferrara (no. 16c—context I in Materials and methods), where the concentration of
Brassica/Sinapis exceeded 1700 seeds/L and accounted for more than 10% of all the seeds found. An intermediate condition was observed in the coeval context no. 15, where
Brassica/Sinapis were found in modest concentrations (8 seeds/L) and represented < 1% of all the seeds collected (
Table 1).
2.1. Archaeobotanical Analysis
The individual plant taxa varied in the distribution of the contexts in which they were found. The most widely found of the taxa identified to species rank were Rapistrum rugosum (L.) All. (11 of 19 contexts) and Brassica rapa (11 contexts), although the latter was far more abundant throughout. Most of the other species featured only in a minority of the contexts surveyed. For example, Sinapis alba seeds were identified with high certainty in several contexts, although few such seeds were found overall (present in 4 of the 19 contexts). The seeds of the genus Brassica were more abundant, with well-preserved representatives of 5 distinct species recorded in 11 of the 19 contexts (B. rapa, B. napus, B. juncea, B. nigra, and B. oleracea). Of these, the vast majority (ca. 95%) were discarded as B. rapa, but the residual Brassica seeds were far more restricted and subdivided between B. nigra (1.6%; 2 contexts), B. napus (1.4%; 5 contexts), B. oleracea (<1%; 1 context), and B. juncea (<1%; 1 context). Most of the other confirmed species were similarly restricted to 5 contexts or less, the sole exception being Myagrum perfoliatum L. (8 contexts).
The presence of less well-preserved or abnormal/intermediate samples meant that many of the samples could only be identified at higher taxonomic levels. For instance, around 2.5% of the seeds could only be provisionally identified as either unknown Brassica seeds (Brassica sp.) or as Brassica/Sinapis seeds, and species identity could not be determined for the seeds of the genera Lepidium (3 contexts), Rorippa (2), or Sisymbrium (1); the genus even remained unidentifiable for some seeds.
In addition to Brassica sp.pl. and Sinapis alba, some other taxa may also have had ethnobotanical value (e.g., Camelina sativa and Isatis tinctoria L.); others were probably simple instances of weeds s.l.
2.2. a-DNA Barcoding
As expected, the ancient seed samples invariably provided DNA yields that were too low and degraded to allow detection by agarose-gel electrophoresis or conventional spectral photometry. Similarly, all attempts using standard barcode primers and protocols used for conventional
rbcL or
matK barcoding failed to generate detectable amplicons. The application of the MT-PCR method did nevertheless succeed in producing visible amplicons for both barcoding genes. Success was modest for
rbcL. The first of the two regions targeted (
rbcL1) yielded only 8 products of the expected 218 bp, but none of the sequences generated from these amplicons matched either the
Brassica or the
Sinapis species expected from the seed morphology. However, 39 seeds produced strong amplification products for the second region (
rbcL2, 203 bp), and most of these yielded sequences of variable quality in both directions. BLASTn searches of the trimmed sequences against the NCBI database invariably identified the appropriate region of
rbcL. Subsequent searches of the BOLD Systems database revealed matches with sequence homologies above 95% for all but 8 of the samples (
Table 2). Most of the remaining samples matched the species that were clearly incongruent with the seed morphology and were therefore deemed contaminant amplicons. Most of these matched species or genera that were likely to have been in the vicinity of the excavation sites, such as
Citrus,
Juniperus,
Pinus,
Picea or
Cedrus. Thus, only nine samples were matched most closely to the three
Brassica species (
B. rapa,
B. napus and
B. oleracea) in the BOLD Systems database and the reference panel (
Table 2).
Greater success was achieved using primers that targeted the matK gene. Here, all six loci targeted within the matK gene produced sequences that corresponded to one of several members of the Brassicaceae. These are each described below, in turn.
MatK1. The different
matK primers for this amplicon varied in their capacity to generate viable products and in the ability of the resultant sequences to differentiate between species held on public databases. The conventional
matK1 primers performed the least well. These primers yielded only 12 sequences of modest quality, and none matched significantly with any reference barcode held on either the BOLD Systems or on the NCBI database. The performances of the
matk1a primers were substantially better, with modest quality sequences of 42–182 bp in length (after trimming) secured from 64 samples. Here, matches were recovered for 41 of the 65 samples. Most of these matched equally well to multiple species in terms of percentage identity, score quality, and e-value, but in all cases, the most strongly matched species were all members of the Brassicaceae family. A finer-level diagnosis than family was not possible for 7 samples, but of the remaining sequences, 16 were assigned to the tribe
Brassiceae, 10 to the genus
Brassica, one to another genus in the family, four to a single species of
Brassica, and three to another species within the family (
Supplementary Information Table S1). The
matk1b primers yielded fewer sequences (26), but these were generally longer (150–212 bp) than the
matk1a. Again, matches were secured for all 26 using NCBI BLASTn searches (
Supplementary Information Table S1). Of these, 14 matched two or more species of
Brassica (genus-level diagnosis) and six matched a single
Brassica species. The remaining six samples matched species, genera, or families outside the Brassicaceae and were therefore deemed contaminant sequences. The performance of
matk1c was slightly improved compared with the other
matK1 combinations. Trimmed sequences of between 118 and 181 bp were secured from 36 samples. Following BLASTn searches of the NCBI database, 26 species were matched to two or more species of
Brassica and one was identified to a single
Brassica species. The remaining samples were identified to the same level as the BOLD systems.
MatK4. These primers generated 19 sequences of generally good quality that varied in length between 189 and 279 bp, with 17 sequences falling in the range of 262–279 bp. Searches using these sequences produced matches with both databases (
Supplementary Information Table S1). Following sequence-homology searches, all except the shortest sequence matched two or more members of the genus
Brassica (genus-level diagnosis). The shortest sequence failed to match any species above the 95% similarity threshold.
Matk5. This combination of primers generated the most extensive set of sequences (90 samples), which varied in length from 75 to 225 bp. These sequences generated diagnoses at various taxonomic levels when searched against both public databases (
Table 3). The finest level of diagnosis was obtained when these sequences were subject to BLASTn searches on the NCBI database. Here, no sequences matched the reference barcodes of a single species. However, the number of hits on the NCBI database was high enough that the frequency of hits to one target frequently allowed the provisional identification of the sample. For example, 11 samples matched equally with 52 reference sequences of
B. oleracea, but also to four other sequences representing three other species. Given the scope for misidentification and the extent of hybridization in the genus, these 11 sample sequences were therefore designated as provisional
B. oleracea. Similar frequency distributions were also noted for 17 sample sequences matching
B. nigra and its allotetraploid sister species,
B. carinata. However, given that the latter is native to Eastern Africa [
40], these samples were provisionally identified as
B. nigra only. The diploid–allotetraploid sister pairing of
B. napus/
B. rapa could not be distinguished using this amplicon, but collectively produced another highly skewed distribution, with 11 sample sequences falling into this category (provisional
B. napus or
B. rapa). Thus, samples assigned to
Brassica species belonging to the cultivated U-triangle group [
5] accounted for 43% (39/90) of the sequences retrieved. Among the remainder, a further 7 samples matched multiple
Brassica species equally (genus level diagnosis). The residual 45 samples (50%) were identified at tribe or family level or with a single species of the family outside the type genus
Brassica, and were therefore deemed inaccurate diagnoses on the basis of incongruence with seed morphology (species-level diagnosis, Brassicaceae; see
Table 3).
2.3. Information from Ancient Herbaria
The search for samples of
Brassica and
Sinapis produced results only for 3 of the 7 assessed herbaria and made it possible to find 25 specimens (
Table 4). The species recognized with certainty were
B. napus,
B. nigra,
B. oleracea,
B. rapa and
S. alba; also in addition, some of the specimens were not clearly identifiable, and were therefore doubtfully attributed to the cited species, or simply to
Brassica sp.
Erbario Aldrovandi provided the best information, with 18 specimens belonging to 4 species (
B. nigra,
B. oleracea,
B. rapa and
S. alba), which were dated to the period 1551–1586 [
41,
42,
43,
44,
45]. Erbario ex Cibo B provided 6 specimens and 4 species (
B. napus,
B. nigra,
B. oleracea and
S. alba), datable at the period 1550–1553 [
46,
47]. Erbario Cesalpino provided only 1 specimen of
B. nigra, datable at the period 1555–1563 [
48].
Brassica oleracea was the most abundant species (6 specimens in Erbario Aldrovandi, 1 specimen in Erbario ex Cibo B);
B. nigra (
Figure 2) was the sole species present in all the herbaria.
B. napus was present only in Erbario ex Cibo B (in Erbario Aldrovandi its presence was doubtful) and
B. rapa was present only in Erbario Aldrovandi (
Figure 3).
Sinapis alba was a marginal presence, with only 3 specimens, of which 3 were in Erbario ex Cibo B and 1 was in Erbario Aldrovandi (
Figure 4).
3. Discussion
Our archaeobotanical analyses of the seed and fruit morphology were internally consistent and suggestive of the presence of several taxa belonging to the Brassicaceae. Notably, the presence of
Sinapis alba was found to be recognizable among the samples in which the external seed tegument was well preserved [
15]. Perhaps the most interesting aspect of these data was the presence of several species of
Brassica. The large quantity of seeds available for this genus [
50] in the two contexts examined allowed the unequivocal separation of the samples into morphological groups associated with species descriptions. On this basis, the dominant species among our sites seems to have been
B. rapa s.l., with representatives of
B. nigra and
B. napus also well represented. However, several studies have reported extensive intraspecific variability in the key diagnostic features of seed-surface architecture, seed size, and seed shape for species belonging to this genus [
51,
52,
53,
54]. Despite this known variability, none of the widely used diagnostic keys attempt to estimate the frequency of erroneous diagnosis [
15,
50,
55,
56,
57]. We therefore sought to support identifications based on seed morphology through the use of DNA sequencing as an independent mode of species diagnosis. DNA yields from ancient biological samples (aDNA) are typically low and highly degraded [
58]. However, several groups have successfully used the DNA extracted from ancient plant material to confirm species identity using short amplicons of the core DNA barcodes [
59,
60] or else elected to target supplementary or non-coding barcode markers [
61]. Such works typically use either large volumes of starting materials [
62] or next-generation sequencing platforms [
63]. Currently, we are unaware of any study using aDNA for the identification of individual
Brassica seed samples. Here, we were not able to secure high-quality barcode sequences in either
rbcL or
matK by direct PCR, but we were able to do so using a modified form of nested PCR [
64]. Several other works have reported similar success in recovering Sanger sequence data from materials containing low levels of template DNA, including DNA from animal feces [
65], adulterated food [
66], formalin-preserved specimens [
67], and dried museum specimens [
68].
In this study, we found that our ability to use aDNA-derived sequences to differentiate between members of
Brassica varied considerably between amplicons, with the best results deriving from the
matK5 primer set. Reference to the seed morphology allowed us to discount all barcode-based diagnoses that fell outside the genus
Brassica, most of which were relatively short sequence reads. It is perhaps notable that the finest level of taxonomic match was achieved after searches were performed using NCBI BLASTn searches. This may be at least partly attributable to the fact that the NCBI database contains a far higher proportion of data originating from next-generation sequencing than from Sanger sequencing, and the former is noted for its lower levels of technical errors [
69]. Considering only the search results that were deemed reliable (i.e., those within
Brassica) provided some support for the morphological identifications. Specifically, there was evidence of the significant presence of
B. nigra in both sites, as well as of
B. oleracea. The inability of the
matK5 sequences to distinguish between
B. napus and one of its progenitor species,
B. rapa [
5], precluded the confirmation of the distributions of these individual species, but was nevertheless congruent with the overall inferences of the presence made from the morphological analyses.
Both sources of plant diagnosis suggested a significant presence of
B. nigra. The efficacy of morphological diagnosis partly depends on the characteristics of the seeds of this species. In fact, in the genus
Brassica, originally, dark-seeded species (e.g.,
B. nigra and
B. napus) and yellow-seeded species (e.g.,
B. rapa and
B. juncea) are present [
70]: the first have significantly more lignin than the latter [
71], especially in the inner epidermis of the seed-coat (testa) [
72,
73], perhaps also because the seeds of
B. nigra must undergo a necessary dormancy period before germination [
72]. This feature surely makes the seeds of
B. nigra more resistant than others, and the careful choice of seeds made in view of the aDNA analyses probably caused a slight over-representation of this taxon.
The total number of specimens found in the Renaissance herbaria was not high if considered in the light of the economic importance of these species, and the presence of samples was limited to the three herbaria richest in
exsiccata (Erbario Aldrovandi, Erbario ex Cibo B and Erbario Cesalpino). The
B. oleracea was probably more easily available than the others due to its widespread cultivation as a vegetable; in fact, featured 7 specimens, plus one of uncertain attribution, comprising 1/3 of the total number of specimens found in the herbaria. In this regard, Mattioli [
74] referred that various cabbage “species” exist, and Durante [
75] noted that cabbages were cultivated and transplanted in all kitchen gardens and vineyards. By contrast,
B. napus and
B. rapa were only marginal presences among the herbarium samples, whereas
B. rapa was the most abundant species among the archaeobotanical remains. It is worth noting that during the 16th century,
B. napus was commonly known, as attested by Mattioli [
74]; therefore, it was probably commonly cultivated, and various “species” of it existed [
75];
B. rapa was also extremely common in Italy, especially in the north [
74,
75].
B. nigra, another medicinal plant [
76,
77], was present in all three herbaria, even if with only one sample each; finally,
Sinapis alba also had few records, in two herbaria only, despite its recognition as the common quality of mustard, cultivated in the kitchen gardens [
74,
78].
Ethnobotanical Considerations in the Possible Uses of Brassica and Sinapis Seeds
In the two pits, the significant abundance of seeds of the
Brassica sp.pl. and
Sinapis alba makes it possible to hypothesize a specific use of these taxa: we observed two contexts that evidently collected waste from food preparation [
24,
79].
It is known that, during the Middle Ages, various species of
Brassica were cultivated in Europe to obtain oil [
80,
81]. During this period, in fact, for economic, cultural, and religious reasons, the use of “minor“ vegetable oils in food preparation became quite widespread and nearly dominant in several territories, principally in those that were (at least partially) outside the range of
Olea europaea L. [
82]. In particular, the oil of rapeseed (
B. rapa/campestris var.
oleifera), was used up to the 20th century, particularly in northern Italy [
83], where the taxon was commonly cultivated, even in the mid-1500s [
74,
75]. On the other hand, the oil of swede (
B. napus) was frequently used as a condiment, as a fuel for oil lamps, to make soap, and in wool crafts [
75]. Thus, the numerous seeds of
B. rapa s.l. and
B. napus found in the archaeological contexts of Ferrara and Lugo probably represent the waste of gentle squeezing (since not many fragments are present) to obtain oil. It should be noted that they were also used as components in complex pharmaceutical preparations, such as the
theriaca: the seeds of
B. napus and
B. rapa were regarded as excellent counterpoisons and, therefore, inserted in the composition of numerous antidotes [
74].
On the other hand, various seeds of the genera
Brassica and
Sinapis are used as spices and ingredients in sauces, in particularly
B. nigra,
B. juncea, and
S. alba [
84,
85]. In the mid-15th century Michele Savonarola, ancestor of the more famous Girolamo, as a physician of Duke Borso d’Este in Ferrara, wrote a dietetics treaty, listing and commenting on the foods that were more or less commonly present on the tables of that epoch [
86]. Among various features, Savonarola notes that in Ferrara, “ogni contrada” (every neighborhood) had two or three sales counters for “senava” (mustard), which was consequently widely used in kitchens. Here, the term “senava” much probably identifies the mostarda, which in its current form was likely codified in the Middle Ages but was subsequently differentiated in diverse regional recipes [
87,
88]. Mostarda can be a sauce made of crushed mustard seeds only, or a very rich and complex product, with the addition of fruits (grapes or their derivatives, figs, apples, pears, quinces, blackberries, walnuts etc.) and various spices in seeds (anise, coriander, fennel, pepper etc. [
88]). It is interesting to observe that nearly all these potential ingredients were found among the botanical remains of the two pits [
38,
79]. Mustard is considered as one of the “universally” widespread elements of the cuisine of the late Western Middle Ages [
89,
90]; the contemporaneous presence of
B. nigra and
S. alba in both in Ferrara and Lugo suggests that this product could also have been produced in both contexts.
5. Conclusions
The multiproxy approach in this research was proven to be of interest. The two waste pits of Ferrara and Lugo offered a notable and unusual quantity of seeds of Brassica sp.pl. and Sinapis alba. Thus, in this case, optimal conditions were available in which to perform the traditional morphometric analyses on the seeds in the most effective way possible, showing non-negligible species diversity within the genus Brassica in both contexts.
In addition, despite the aforementioned limitations, the availability of such a large volume of remains made it possible to attempt research on the aDNA for a taxon normally not studied in this sense (in contrast to the frequency with which other important economic plant species, such as
Vitis vinifera L. or cereals, are studied [
111,
112,
113]). Results were obtained that could form the basis for new and more in-depth investigations in this field.
Our research on the ancient herbaria (which are datable to an epoch slightly later than the contexts studied) and other historical sources of the Middle Ages/Renaissance period allowed to understand how these species were frequent and widespread among the cultivated food plants. Furthermore, the seeds of these species were used to obtain oil and other seasonings, which, from the medieval period onwards, became typical elements of all European cooking.