Survey of Cellular Autofluorescence Variation in Saliva Deposits: Implications for Estimating Time Since Deposition

Abstract

Background/Objectives: The goal of this study was to characterize changes in autofluorescence of epithelial cells obtained from saliva stains that occur with time and investigate the potential for these changes to serve as time-since-deposition (TSD) signatures for this sample type. Methods: Saliva from 50 individuals was used to create 208 deposits that were aged between one day and nine months. Autofluorescence profiles of individual cells were obtained from each sample using imaging flow cytometry (IFC) and analyzed across nine different emission channels ranging between 435 nm and 800 nm. Results: Results showed strong evidence for linear increases in autofluorescence intensity when epithelial cells from a single donor deposit were measured over time (12 of 14 donors r ≥ 0.9). When autofluorescence profiles from all 50 donors were combined into a single time series, variation in autofluorescence intensity was observed between individual deposits with the same TSD. This inter-contributor variation decreased the overall strength of the linear relationship (r = 0.83) and yielded residual errors of ~8 days for samples that were actually 1 day old and ~82 days for samples that were over 180 days old using a linear regression model. Although this approach may not currently be amenable to estimating TSD to the day with high accuracy, clear, non-overlapping differences in autofluorescence intensity were still observed between certain time intervals, e.g., saliva deposits that were aged for 1 day compared to saliva deposits that were aged for more than 120 days. Conclusions: This suggests that cellular autofluorescence signatures have the potential to be probative when hypotheses for sample deposition involve disparate time intervals or as a screening tool for identifying which samples are most likely relevant to the crime in question based on their deposition time.

1. Introduction

The increasing sensitivity of DNA typing systems has led to a precipitous rise in the number of trace samples collected and processed in criminal cases. When evidence is collected from common spaces in particular, it can lead to an increased expenditure of limited resources on samples that lack probative value, e.g., finding the homeowner’s DNA on items at the scene of a burglary or recovering complex mixture profiles from a substrate that multiple people have handled. The prevalence of background levels of DNA deposited within common areas and/or substrates but at a completely different, unrelated time from the crime in question has long been recognized as a challenge for forensic DNA casework [1,2,3]. Resolving scenarios like these require information about the timing of deposition of biological material, or its “time since deposition” (TSD). Previous efforts have described various signature systems with the potential for estimating TSD in forensic biological samples including colorimetric changes [4,5], abundance of RNA biomarkers [6,7,8], and protein profiles [9,10,11]. Biochemical signatures inferred from Fourier-transform infrared spectroscopy (FTIR) and/or Raman spectroscopy profiles have also been reported [12,13,14]. While many of these studies have exclusively focused on blood stains, recent studies have also reported TSD signatures for other forensically relevant tissue types such as saliva, using approaches such as metabolic profiling of saliva stains [11,15] and characterizing the temporal changes in the bacterial communities that are co-deposited with saliva [16,17,18,19].

A promising alternative approach for estimating TSD involves characterizing autofluorescence emissions from individual cells within a biological deposit. The wavelengths at which a cell fluoresces and the intensity of emissions are largely a function of the types and relative abundance of compounds that are present [20]. Endogenous compounds like NAD/FAD, porphyrins, and keratins all fluoresce at a range of wavelengths [21,22,23] and are abundant in most cell types. Since biochemical components like these will change over time as a cell degrades, variation in cellular autofluorescence profiles has the potential to provide probative TSD estimates for biological evidence. Fluorescence-based detection of endogenous molecules such as NADH, tryptophan, and FAD has been used to develop TSD signatures for blood and menstrual blood samples [24,25,26]. Additionally, autofluorescence profiles of whole cells have been surveyed across a range of forensically relevant sample types and found to vary across both source tissue (e.g., epidermal, buccal, vaginal) and TSD of the sample [27,28,29,30].

An ongoing limitation with the most foundational TSD research has been the number of individuals used to build TSD signatures (usually less than 20). Consequently, the extent to which intrinsic biological variation between individuals may impact the accuracy of developed signature systems is largely unknown. For autofluorescence signatures specifically, inter-individual variation in cellular autofluorescence profiles has been observed in samples deposited under the same conditions [23,31] but has not been thoroughly characterized across a larger set of donors or across a broad range of TSDs. Therefore, the goal of this study was to survey cellular autofluorescence signatures as a function of TSD in a larger set of contributors to explicitly characterize the long-term variation in autofluorescence in cell populations representing (1) a biological deposit from a single individual analyzed at different TSDs (changes that occur strictly due to sample age) and (2) deposits from different individuals that have the same TSD (inter-individual variation). To accomplish this, autofluorescence profiles of both single cells and the entire saliva cell population were analyzed in a sample set comprising 50 individuals and 208 total contributor cell populations, aged between one day and approximately nine months. Autofluorescence profiles consisted of emissions between 435 nm and 800 nm and incorporated 84 different metrics capturing both intensity of fluorescence emissions and morphological attributes of individual cells. This constitutes one of the largest and most temporally extensive saliva TSD datasets reported to date.

2. Materials and Methods

2.1. Sample Collection

A total of 208 saliva samples were collected from 50 unique donors after approval by the Virginia Commonwealth University Institutional Review Board (approved protocol #HM20000454). Individuals were recruited without considering age, sex, or demographic.

Participants were asked to donate 1 mL of saliva into 15 mL conical tubes; the saliva was then spotted onto microscope slides in 100 µL aliquots and allowed to age for a period corresponding to a designated TSD between 1 day and 280 days. Samples were stored on non-porous benchtop surfaces under ambient conditions (temperature 21–23 °C and humidity ~40–45%). Samples were stored in a partial enclosure with no direct overhead light source but were not otherwise shielded from ambient light in the room. Each deposit was assigned a random TSD and the majority of donors (47 out of 50) are represented only once at a given TSD.

Dried saliva spots were then sampled with sterile cotton swabs (PN: 25-8062WC; Puritan; Guilford, ME, USA) pre-wetted with deionized water. Processing of all saliva samples followed previously published protocols [30] Elution of cells from the swabs began with an initial incubation in 1.5 mL of sterile, 1× phosphate-buffered saline (PN: 119-069-101; Quality Biological; Gaithersburg, MD, USA) for 20 min. Following incubation, all swabs were vortexed for one minute at 3000 rpm. The swabs were then discarded, and the remaining solution was filtered into microcentrifuge tubes with 100 μm filter paper (PN: 22363549; Fisherbrand; Hampton, NH, USA) to remove larger cell aggregations and/or non-biological particles. To complete elution, cell solutions were centrifuged at 21,130× g for two minutes and concentrated down to ~50–75 μL by removing the supernatant. The cell pellet was resuspended by vortexing prior to imaging flow cytometry (IFC) analysis.

For the first set of experiments involving single donor saliva stains analyzed over time, deposits were created with the following TSDs: one day, one week, two weeks, one month, two months, and three months. These time points were chosen to explicitly characterize changes in autofluorescence across TSD intervals that are often in question during an investigation, e.g., whether a sample is less than a week old or several months old. Three of these donor series were sampled at two additional time points corresponding to four and five months (C22, H42, M02). For the next experiment, saliva cell populations were analyzed by IFC at one day, two to seven days, between one and two weeks, between two and three weeks, and between one month and nine months in one-month increments.

2.2. Imaging Flow Cytometry

Samples were analyzed using a Cytek Amnis Flowsight (Cytek Biosciences; Fremont, CA, USA) imaging flow cytometer equipped with 405 nm, 488 nm, 642 nm, and 785 nm lasers, with voltages set to 100 mW, 60 mW, 100 mW, and 6.25 mW, respectively. Brightfield, side scatter, and multiple fluorescent images of individual cells were captured at 20× magnification. Channel 1 was used to capture brightfield (i.e., non-fluorescent) images, and channel 6 was used to capture side scatter emissions triggered by the 785 nm laser. Fluorescent images were captured in nine detector channels excited by different sets of lasers: channels 2–5 were excited by the 488 nm laser, and channels 7, 8 and 10–12 were excited by the 405 nm and 642 nm lasers. Emission spectra captured in each channel were as follows: channel 2 (505–560 nm; green); channel 3 (560–595 nm; yellow); channel 4 (595–642 nm; orange); channel 5 (642–745 nm; red); channel 7 (435–505 nm; violet); channel 8 (505–560 nm; green); channel 10 (595–642 nm; orange); channel 11 (642–745 nm; red); and channel 12 (745–800 nm; pink). The instrument configuration and channel order are shown in Figure 1. Samples were run until 10,000 events were acquired in the acquisition gate or for a maximum of 20 min.

2.3. Data Analysis

Following IFC, the resultant raw image files (.rif) were exported into Image Data Exploration and Analysis Software (IDEAS Version 6.0; EMD Millipore; Seattle, WA, USA) for the analysis of imaged cells. Individual cells were differentiated from debris or other non-cellular material by filtering based on size and gradient values; all cells included in this dataset have a gradient value greater than 50, an aspect ratio greater than 0.45, an area between 1450 μm² and 3000 μm² and a perimeter-to-area ratio <0.13. Fluorescence intensity of a cell population was defined as the average of all cells detected within these parameters.

The entire cell population dataset for this study is accessible through the following repository: DOI: https://doi.org/10.6084/m9.figshare.28892699. Pearson’s correlation coefficients (‘r’), linear regression statistics, Tukey’s HSD, and boxplots were calculated using RStudio: Integrated Development Environment for R (Version 2026.01.0+392; Boston, MA, USA).

3. Results

3.1. Autofluorescence Signatures in Single Contributor Cell Populations

The relationship between cellular autofluorescence and TSD was first tested in a series of single contributor saliva deposits that were analyzed across TSDs ranging between one day and approximately five months. Results showed that intensity of autofluorescence in certain detector channels has a strong linear correlation with TSD for each of the 14 contributors tested in certain detector channels. Specifically, out of the nine emission channels surveyed, autofluorescence intensity in channel 7 (excitation 405 nm; emission 435–505 nm) yielded correlation coefficient (r) values greater than 0.90 for 12 out of 14 contributors, with the remaining two contributor time series showing correlation coefficient values of 0.81 (donor T28) and 0.70 (donor C33) (Figure 2). Similar trends were observed with autofluorescence intensity measured in channel 8 (excitation 405 nm; emission 505–560 nm) (Figure S1, Supplemental Data). Linear increases in autofluorescence intensity with time were not observed in the other detector channels, i.e., autofluorescence intensity in channels 2, 3, 4, 5, 10, 11, and 12 (Table S1, Supplemental Data). Morphological attributes of saliva cell populations (i.e., non-fluorescent) analyzed in brightfield (channel 1) and side scatter (channel 6) also did not appear to vary with time.

Although overall linearity in channels 7 and 8 was consistent for each donor time series, variation was observed in the magnitude of autofluorescence across contributor cell populations for the same time points. For example, the average fluorescence intensity of epithelial cells obtained from donor M02’s saliva after one day was 1.9 × 10⁴ RFUs, whereas average autofluorescence intensity of epithelial cells obtained from H89′s saliva at one day was approximately two-fold higher at 4.5 × 10⁴ RFUs (Figure 2). Fluorescence intensity values at later time points also showed differences between donor cell populations at the same TSD, e.g., autofluorescence intensity ranges from 7.3 × 10⁴ to 2.0 × 10⁵ RFUs at TSD 90 for donors T28 and H89, respectively.

3.2. Comparison of Autofluorescence Variation with TSD and Across Contributor Cell Populations

To characterize the impact that inter-contributor differences in autofluorescence have on resolving the changes in autofluorescence that occur with time, a single larger time series was created by combining data from the previous experiment with cell population data from an additional 104 saliva stains representing 36 individuals and TSDs ranging between 1 day and 280 days. Results still showed evidence of a linear increase in cell population fluorescence intensity in channel 7 across the full time range (r = 0.83, Figure 3). The largest differences were observed between the freshest and oldest TSD samples; mean autofluorescence intensity in 1-day samples ranged between 1.2 × 10⁴ to 6.5 × 10⁴ RFUs, compared to 215-day or 280-day samples, which showed a range of autofluorescence intensity 10-fold higher, ~1.4 × 10⁵ to ~4.5 × 10⁵ RFUs.

Autofluorescence in channel 8 showed a similar trend between autofluorescence intensity and time (r = 0.82, Figure S3) with the largest differences between samples at either end of the time series. As with the individual donor time series, there were few discernible trends for other fluorescent channels or morphological attributes with TSD. As an example, within the brightfield channel, average cell area (μm²) had the highest level of variation between contributor cell populations with the same TSD and showed no clear linear relationship with increasing TSD (r = 0.27; Figure 4).

Although there was evidence of an overall linear increase in autofluorescence intensity in channels 7 and 8 with time, samples from adjacent time intervals often showed overlapping distribution of intensity values with no obvious transition with TSD. For example, the range of autofluorescence intensity values for samples with TSD of 215 days and samples with a TSD of 280 days was almost identical (Figure 3). Similarly, samples with TSDs of 120 days showed intensity values with a range completely overlapping samples with TSDs of 65 days and 90 days. Variation in autofluorescence intensity also occurred between samples with the same TSD but deposited by different individuals, i.e., inter-contributor variation. The largest levels of inter-contributor variation were observed in samples with the oldest TSDs, e.g., samples with a TSD of 215 days and 280 days both varied by as much as ~310,000 RFUs. For the earliest time points, inter-contributor variation was comparatively smaller, e.g., samples aged for one day were distributed between ~53,000 RFUs, indicating a much more homogenous cell population. Interestingly, some variation was also observed between replicate saliva stains from the same donor that were aged for the same period of time. As an example, three samples from donor R64 with a TSD of 1 day showed average autofluorescence levels that varied between 1.5 × 10⁴ and 4.2 × 10⁴ RFUs. Although intra-contributor variation was observed, the range of this variation (~47,000 RFUs) was not greater than the overall increase in autofluorescence seen over 280 days.

3.3. Estimating TSD for Unknown Samples Using Autofluorescence Profiles

Within the context of forensic casework and estimating TSD from unknown samples using autofluorescence, the magnitude of differences between certain cell populations with the same TSD contributes to prediction errors that also increase with TSD. To illustrate this, the linear model shown in Figure 3 was used to calculate the difference between the actual sample TSD and the predicted TSD from the regression equation if the sample were an unknown (i.e., the residual error) for each of the 208 contributor samples. The average residual errors for five different TSD intervals are shown in

Table 1. Residual errors for saliva TSD prediction.

	Time Since Deposition
	1 Day	10–20 Days	30–65 Days	80–165 Days	≥180 Days
Average Residual 1	8	21	23	35	82

¹ Values shown are number of days.

. For saliva samples with an actual TSD of 1 day, the average residual was ~8 days. The average residual error then increased with sample TSD to ~21 days (for samples with TSD of 10–20 days) and to ~82 days (samples with TSD ≥ 180 days).

While these results indicate that precise TSD estimates may not be feasible using autofluorescence intensity and standard linear modeling, the clear differences between cell populations from the earliest time points (1 day) and those from the oldest time points (180 to 280 days) suggest that it may be possible to resolve unknown samples if they have extremely disparate TSDs. To investigate this application, cell populations were grouped into discrete TSD intervals based on similarity/overlap in the range of observed average autofluorescence intensity values (Figure 5).

Results showed clear increases in the distribution of mean autofluorescence intensity across the entire TSD range (Figure 5). While there was considerable overlap among intermediate TSD intervals (i.e., all intervals between 8 days and 120 days), a distinct, non-overlapping range of autofluorescence intensities were still observed between early TSD intervals (1 day, 2–7 days) and later TSD intervals (120–165 days, 180–215 days, and 280 days). There was also an order-of-magnitude increase in autofluorescence intensity (i.e., ~2.7 × 10⁴ RFUs for 1 day, ~4.0 × 10⁴ RFUs for 2–7 days and ~1.9 × 10⁵ RFUs for 180–215 days (i.e., ~2.5 × 10⁵ RFUs for 280 days). Results of statistical tests further indicated that the most robust differences in fluorescence intensity were between samples that were less than 7 days old and samples more than 180 days old. Table S2 shows 95% confidence intervals and comparisons of mean autofluorescence of TSD groups with Tukey’s HSD analysis. Statistically significant differences were most frequently observed when the earliest time groups were compared against other non-adjacent time intervals, specifically, samples aged for 1 day versus those aged for 16–33 days, 40–100 days, 120–165 days, 180–215 days and 280 days (p < 0.001 for all comparisons); similarly, samples aged for 2–7 days showed statistically significant differences between those aged for 40–100 days, 120–165 days, 180–215 days and 280 days (p < 0.001 for all comparisons).

4. Discussion

Overall, the results from this proof-of-concept study show that the intensity of autofluorescence can increase with TSD for contributor cell populations recovered from a saliva stain. While autofluorescence changes were mostly linear when a single donor cell population was monitored over time, autofluorescence intensity clearly exhibits variation between contributor cell populations that is not a function of time and is likely to limit precise estimation of TSD (i.e., to the day). This donor-specific variation could be driven by intrinsic differences in the presence and/or abundance of endogenous fluorophores across cell populations. For example, nicotinamide adenine dinucleotide (NAD) is a metabolic coenzyme prevalent in saliva and/or buccal epithelium; NAD has a broad excitation and emission spectrum consistent with the observed autofluorescence in channel 7 and 8 [22,32]. Previous studies have also demonstrated that NAD abundance can vary between individuals, including buccal epithelium [31,33]. Another possible source of biological variation between donor cell populations is differences in the relative abundance of lipofuscins and oxidized lipids. Both of these molecules have excitation and emission spectra that overlap with channels 7 and 8 [20] and have been shown to vary across saliva samples from different individuals as a function of chronological age or other donor-specific factors [34,35].

Nevertheless, the observed variation in autofluorescence profiles between extremely recent deposits (less than 7 days old) and much older deposits (over 180 days old) could still provide probative context when there are multiple hypotheses for deposition defined by large differences in TSD, e.g., the prosecution’s hypothesis is that a person of interest deposited saliva one day prior, which is consistent with the timing of a crime; the defense’s hypothesis is that the person deposited saliva several months earlier, which was unrelated to the crime. Based on this dataset, if a forensic saliva stain collected at a crime scene exhibits an average fluorescence intensity in channel 7 that is less than 3.2 × 10⁴ RFUs (the median intensity observed for samples aged 1–7 days) this would be consistent with a sample that was deposited recently rather than deposited several months prior.

This type of autofluorescence signature could also potentially be used for non-probative applications such as triaging evidence samples and identifying which are most likely to contain biological material deposited within the time frame of the crime. Since many crime scenes can contain copious amounts of older “background” DNA that predates the crime at issue, being able to confidently screen samples based on TSD could be very useful, particularly given the reality of limited resources. Conversely, in cases where recent contamination by laboratory staff is a potential concern, being able to effectively screen out fresh (e.g., day-old) samples may be beneficial.

5. Conclusions

These results contribute to the growing body of research on TSD signatures for saliva deposits by explicitly characterizing the magnitude of inter-donor variation in cellular autofluorescence across TSD up to nine months. Our primary finding that donor-to-donor differences can significantly impact temporal changes in autofluorescence suggests that phenotypic variability of cell populations may be an ongoing challenge for any TSD signature system based on biochemical properties of saliva. However, even if precise TSD estimates are not currently possible with this approach, the observed differences in autofluorescence intensity across disparate TSDs have potential applications for differentiating very fresh deposits (i.e., samples less than a week old) from aged deposits (i.e., samples that are ~six months or older) or triaging samples on the front end of the DNA profiling workflow.

As with most foundational studies, expanding the dataset to include additional contributor cell populations with TSDs not represented in the current time series is likely to improve the resolution between time groups. Alternative quantitative approaches, including those that incorporate machine learning, are a promising avenue of future research that may minimize the impact of inter-donor variation and increase the precision of TSD estimates (e.g., methods described in [19,28]). However, the effectiveness of these approaches will likely be dependent on sample size and number of donors sampled [36,37].

Similarly, it will be important to include environmental conditions such as heat, UV light exposure and moisture into the reference dataset since these are likely to accelerate the rate at which autofluorescence changes with time, as with has been observed with other types of TSD signatures [38,39]. These environmental effects can then be incorporated into predictive quantitative models for TSD as has been demonstrated with types of forensic analyses (e.g., post-mortem intervals [40]).