Dental calculus has proven to contain a wealth of dietary information in the form of plant microfossils (Henry and Piperno, 2008; Hardy et al., 2009), proteins (Warinner et al., 2014a; Hendy et al., 2018), and other organic residues (Buckley et al., 2014). This dietary information can be preserved within the mineralised dental plaque over many millennia, providing a unique window into the food-related behaviours of past populations (Henry and Piperno, 2008; Tao et al., 2020; Jovanović et al., 2021) and extinct species (Hardy et al., 2012; Henry et al., 2014).

Until recently, only a few studies directly investigated the presence of plant microremains in the dental calculus of archaeological remains. The ability to extract phytoliths from the dental calculus of archaeological fauna to investigate diet was first noted by Armitage (1975), and later by Middleton and Rovner (1994), and Fox et al. (1996). Starches and phytoliths were extracted from human dental calculus by Cummings and Magennis (1997).

In more recent years, the study of dental calculus has increased exponentially, and the wealth of information contained within the mineralised matrix has largely been acknowledged. The use of dental calculus spans a wide variety of archaeological research areas, such as oral microbiome characterisation (including pathogens) through the analysis of DNA and proteins (Adler et al., 2013; Warinner et al., 2014b), microbotanical remains (Henry and Piperno, 2008; Hardy et al., 2009; Mickleburgh and Pagán-Jiménez, 2012), other organic residues and proteins from dietary compounds (Buckley et al., 2014; Hendy et al., 2018), and nicotine use (Eerkens et al., 2018). Especially the extraction of starch granules has become a rich source of dietary information, as starch granules have proven to preserve well within dental calculus over a variety of geographical and temporal ranges (Piperno and Dillehay, 2008; Henry et al., 2014; Tao et al., 2020; Jovanović et al., 2021).

Despite this, our knowledge of dental calculus and the incorporation pathways of the various markers is limited (Radini et al., 2017), as is our knowledge of information-loss caused by these pathways. Additionally, the methods we use to extract and analyse dental calculus, and make inferences on past diets represent another potential source of bias. Studies on both archaeological and modern individuals have explored these biases in more detail. Extraction methods were tested by Tromp and colleagues (2017), specifically regarding decalcification using HCl or EDTA. The authors found significantly more starches with the EDTA extraction method than the HCl extraction method; however, as noted by the authors, comparisons involving archaeological calculus are problematic due to variability between and within individuals. Studies conducted on modern humans (Leonard et al., 2015) and non-human primates (Power et al., 2015; Power et al., 2021) have explored how well microremains (phytoliths and starches) extracted from dental calculus represent the actual dietary intake. These studies are justifiably limited, despite meticulous documentation and observation, due to unknown variables and uncertainty involved in this kind of in vivo research. Dental calculus is a complex oral biofilm with a multifactorial aetiology and variable formation rates both within and between individuals (Haffajee et al., 2009; Jepsen et al., 2011), contributing to the stochasticity of starch representation being observed in numerous studies. Additionally, the concentration of oral α-amylase differs both between and within individuals (Froehlich et al., 1987; Nater et al., 2005), causing different rates of hydrolysis of the starch granules present in the oral cavity. Add to this the effects of the many different methods of starch processing (Hardy et al., 2018), as well as post-depositional processes that are still being explored (Mercader et al., 2018; García-Granero, 2020), and it becomes clear that using dental calculus to reconstruct diet is a highly unpredictable process.

In this exploratory study, we use an oral biofilm model to investigate the retention of starch granules within dental calculus in a controlled laboratory setting, allowing us full control over dietary input. Our main questions concern the representation of granules extracted from the calculus compared to the actual intake. How much of the original diet is incorporated into the calculus, and how much is recovered? Is there differential loss of information from specific dietary markers that affects the obtained dietary information, and how does this affect the representation of diet from extracted microremains?

We find that, despite the absence of α-amylase in the model, a limited proportion of the starch input is actually retained in the calculus. We also observed a shift in the size ratios of individual starch granules that are incorporated into the calculus, and that the number of incorporated starch granules increases as the size of the calculus deposit increases.

All samples yielded sufficient biofilm growth and starch incorporation to be included in the analysis (Figure 2), resulting in a total of 48 biofilm samples (two plates of 24), 45 of which were used for analysis (three samples were set aside for later analysis). Most control samples contained no starch granules, while some contained negligible quantities (see Supplementary Material).

Here, we have provided a method for exploring the incorporation of dietary starches into the mineral matrix of a dental calculus biofilm model. Our results show that a very low proportion of the starches exposed to the biofilm during growth are retained in the mineral matrix, and that the size of the starch granules may affect the likelihood of incorporation. The proportions of starch granules (of all sizes) present in the extracted samples were similar across all treatments (0.06% to 0.16%), despite large differences in absolute granule counts between wheat (mean = 25,404,000) and potato (mean = 3,016,000) solutions.

The absolute counts, however, differed more visibly between treatments and was proportional with the total count of granules in the treatment solutions. Wheat and mixed solutions had the highest absolute mean count of starch granules, and also had the highest absolute mean count of starch granules extracted from the dental calculus (Tables 2, 3). This suggests that the starches that are more frequently consumed will be present in higher quantities in the dental calculus, at least prior to inhumation and degradation in the burial environment. Despite the low proportion of granules recovered from the model calculus (0.06% to 0.16%), the absolute counts were still substantially greater than counts recovered from archaeological remains (Wesolowski et al., 2010; Tromp and Dudgeon, 2015; Tromp et al., 2017), which could in part be due to the lack of oral amylase activity in our model. Previous research conducted on dental calculus from contemporary humans and non-human primates suggest a high level of stochasticity involved in the retention of starch granules in dental calculus, and that starch granules extracted from dental calculus are underrepresented with regard to actual starch intake, which is consistent with our findings (illustrated by high standard deviations and low proportional incorporation). Leonard and colleagues (2015) found individual calculus samples to be a poor predictor of diet in a population, as many of the consumed plants were missing from some individual samples, but were present in others.

Power and colleagues (2015) presented similar findings in non-human primates, where phytoliths were more representative of individual diets than starch granules. The size bias is also consistent with the findings by Power and colleagues (2015), who found that plants producing starches 10–20 μm in size were over-represented; however, the representation of granules larger than 20 μm in their study is unclear.

We have also shown that the size of the starch granules influences the likelihood of incorporation into the calculus. Starch granules larger than 20 μm in maximum length were underrepresented in the calculus samples compared to the original starch solutions, an effect that was consistent across all three treatments. Medium granules (10–20 μm) were often over-represented (Table 4 and Figure 3). Large potato granules were most affected, potentially because of the greater size-range. They can reach up to 100 μm in maximum length, whereas wheat granules generally only reach up to 35 μm (Gismondi et al., 2019; Haslam, 2004; Seidemann, 1966, 174–176). Granule morphology may also play a role. Large wheat granules are lenticular and have a larger surface area compared to volume, whereas large potato granules are ovoid and have a larger volume compared to surface area (van de Velde et al., 2002; Jane et al., 1994; Reichert, 1913, 364–365; Seidemann, 1966, 174–176). Another potentially important factor from our results is the size of the calculus deposit. We found a strong positive correlation between size of biofilm deposit and retained starch granules (Figure 4A), meaning larger calculus deposits contain a higher quantity of granules; a result that contradicts findings from archaeological contexts (Wesolowski et al., 2010; Dudgeon and Tromp, 2014). When the concentration of starch granules per mg calculus is considered, the correlation is weaker, but still present (Figure 4B). While the larger deposits contain a higher absolute count, our findings also suggest that they contain a slightly higher concentration of starches. This may also explain the lower mean retention of starch granules in mixed treatments compared to wheat treatments. Wheat treatment samples (mean = 5.53 mg) were on average larger than mixed treatment samples (mean = 4.28 mg) (Table 1); and while mixed treatment solutions contained the highest mean overall granule counts, wheat treatment samples had the highest mean starch retention. Further research is needed to determine why this differs from previous archaeological findings.

The mechanism by which starch granules are incorporated into plaque and calculus remains largely unknown, and few studies have directly investigated potential mechanisms. We know that a proportion of the starch granules entering the mouth can become trapped in the plaque/calculus, and can be recovered from archaeological samples of considerable age (Buckley et al., 2014; Henry et al., 2014; Wu et al., 2021). Studies have also shown that not all starch granules come from a dietary source. Other pathways include cross-contamination from plant interactions in soil, such as palm phytoliths adhering to the skin of sweet potatoes (Tromp and Dudgeon, 2015), or accidental ingestion not related to food consumption (Radini et al., 2017; Radini et al., 2019).

When starch granules enter the mouth, whether through ingestion of food or accidental intake, they immediately encounter multiple obstacles. It is likely that the bulk of starch granules are swallowed along with the food, and are only briefly present in the oral cavity. Other granules that are broken off during mastication may be retained in the dentition through attachment to tooth surfaces (including plaque and dental calculus) and mucous membranes (Dodds and Edgar, 1988; Kashket et al., 1991). Bacteria also have the ability to adhere to starch granules (Topping et al., 2003), which would allow starches to attach to bacterial communities within the biofilm. These granules are then susceptible to mechanical removal by the tongue, salivary clearance, and hydrolysis by α-amylase (Kashket et al., 1996). The susceptibility of granules to hydrolysis depends on the crystallinity and size of the starch granule, as well as the mode of processing. Smaller and pre-processed (e.g., cooked) starch granules are more susceptible to enzymatic degradation, while dehydrated starches will have a reduced susceptibility (Björck et al., 1984; Franco et al., 1992; Lingstrom et al., 1994; Haslam, 2004; Henry et al., 2009). Cracks on the surface of the dental calculus, as well as unmineralised islands and channels may also be able to contain starch granules (Tan B. et al., 2004; Charlier et al., 2010; Power et al., 2014). Starch granules that are trapped in these pockets are (at least to some extent) protected from aforementioned clearance mechanisms, especially once a new layer of plaque has covered the surface of the plaque/calculus. The size bias against large granules (>20 μm) from both wheat and potato (Table 4) may give further credence to this incorporation pathway, as the smaller starch granules have an advantage over larger granules, and can be stored in larger quantities. This was also suggested by Power and colleagues (2014), who observed clusters of starches within dental calculus, rather than an even distribution across the surface of the dental calculus. Granules trapped in plaque/calculus may still be susceptible to hydrolysis, as α-amylase has the ability to bind to both tooth enamel and bacteria within a biofilm and retain a portion of its hydrolytic activity (Scannapieco et al., 1993; Nikitkova et al., 2013; Tan B. T. K. et al., 2004, Tan et al., 2004 B.). After the death of an individual, starches within dental calculus are susceptible to further degradation by post-depositional processes, depending on burial environment (pH, temperature, moisture content, microorganisms) (Franco et al., 1992; Haslam, 2004; Henry et al., 2009; García-Granero, 2020). Future study should explore how burial affects the recovery of starch from the biofilm model.

The absence of α-amylase in the model is a limitation of this study, as the total granule counts were not subject to hydrolysis. This would likely have reduced and affected the size ratios, as smaller starches may be more susceptible to hydrolysis (Franco et al., 1992; Haslam, 2004). The absence may also affect biofilm growth due to the lack of amylase-bacterium interactions (Nikitkova et al., 2013). Conversely, the model may benefit from the absence of α-amylase, because it can allow us to directly explore its effect on starch counts in future experiments, where α-amylase can be added to the model in concentrations similar to those found in the oral cavity (Scannapieco et al., 1993). We are able to show how absolute counts in the treatments cause a difference in incorporation. However, this was merely a side-effect of the difference in the number of granules in potato and wheat solutions of the same concentration (w/v). Further research should test multiple differing concentrations of the same starch type. The use of EDTA may also have affected counts. While previous studies have shown negligible morphological changes caused by exposure to EDTA (Tromp et al., 2017; Modi et al., 2020; Le Moyne and Crowther, 2021), these studies have not considered changes to separate size categories within starch types, and whether shifts in size ratios occur due to exposure to the pre-treatment chemicals. The total number of granules on a slide often exceeded a number that was feasible to count in a reasonable time period, so we calculated the total counts by extrapolating from three slide transects. Thus, we reasonably assume that the three transects are a good representation of the entire slide, and that the distribution of all granules on the slide is relatively homogeneous.

Finally, we only used native starches in the experimental procedure and the results will likely differ for processed starches (García-Granero, 2020). Based on the comparatively low counts obtained by Leonard and colleagues (2015, Supplement 2), processing and amylase may have a substantial effect on starch granule retention in the oral cavity.

While we are unable to sufficiently address the mechanism(s) of starch incorporation with the data obtained in this study, the dental calculus model presented here is uniquely suited to explore these questions and may improve interpretations of dietary practices in past populations. Further analyses using this model can address the call for more baseline testing of biases associated with dietary research conducted on dental calculus (Le Moyne and Crowther, 2021). Our high-throughput experimental setup allows us a higher degree of control over the factors that influence starch incorporation and retention, such as dietary intake, differential survivability of starches, and inter- and intra-individual variation in plaque accumulation and mineralisation. The latter is especially difficult to control in vivo as it is influenced by numerous factors including genetics, diet, salivary flow, and tooth position and morphology (Haffajee et al., 2009; Jepsen et al., 2011; Simón-Soro et al., 2013; Proctor et al., 2018; Fagernäs et al., 2021), as well as evolutionary differences (Fellows Yates et al., 2021). The set of limitations for our model differ from in vivo methods and, as such, we expect our model to complement the results and interpretations of existing and new in vivo studies. It can also facilitate training of students and researchers on methods of dental calculus analysis, such as starch and phytolith extraction and identification, where it can replace the use of finite archaeological resources.

This preliminary study shows that a very small proportion of the input starch granules are retained in a dental calculus model. This and previous studies have shown that calculus has a low capacity for retention of starch granules, an effect that is compounded by diagenetic effects in archaeological remains, resulting in low overall counts of extracted granules. The proportion of starches consumed will in many cases be reflected in the quantity of starches extracted from the dental calculus—i.e., the more starch granules entering the oral cavity, the more will be recovered from extraction—at least in modern calculus samples unaffected by diagenesis and hydrolysis. Whether or not this also applies to archaeological samples remains to be tested. Additionally, we have shown that the size of granules will influence the likelihood of incorporation, as large (>20 μm) starches have a decreased incorporation rate, medium (10–20 μm) starches an increased rate, and small (<10 μm) granules remained somewhat constant. The size of calculus deposit also seems to influence the capacity of granule incorporation; as the size of the deposit increases, so does the absolute count of incorporated granules.

While we have shown multiple factors that influence the likelihood of incorporation, the process still appears to be somewhat stochastic. Further research is needed to make sense of the contributing factors, and to explore the mechanisms of intra-oral starch incorporation and retention in dental calculus. The oral biofilm model described in this study provides a method to explore the incorporation and extraction of dietary compounds from dental calculus in a controlled laboratory setting. We do not expect our model to replace in vivo methods; instead, it can provide a complementary means to address the limitations of in vivo studies, and unearth the potential biases associated with dietary research conducted on archaeological dental calculus.