Droplet‐based optical trapping for cell separation in mock forensic samples

Optical tweezers have a wide range of uses for mechanical manipulation of objects in the microscopic range. This includes both living and static cells in a variety of biomedical and research applications. Single‐focus optical tweezers, formed by focusing a laser beam through a high numerical aperture immersion objective, create a significant force, which enables controlled transport of a variety of different cell types and morphologies in three dimensions. Optical tweezers have been previously reported to capture and separate spermatozoa from a reconstituted simulated postcoital sample. We report herein the development of a simplified, more efficient cell transfer protocol that can separate and isolate both spermatozoa as well as leukocytes, with similar efficiencies as those previously reported. The new cell transfer method was used to separate sperm cells from a reconstituted mixture of spermatozoa and vaginal epithelial cells, with complete STR profiles developed from 50 cells with little evidence of contribution from the female contributor to the mixture. This modified protocol was then used to separate 21 samples of enriched leukocytes, with trapped cells ranging from 5 to 22 cells. Complete STR profiles were developed from as few as 10 leukocytes. Thus, with minimal sample preparation and a short trapping time, this method has the potential to provide an alternative to traditional differential extraction methods for separation of sperm:nonsperm mixtures while also providing versatility for separation of cells with differing morphologies.

spermatozoa from a reconstituted simulated postcoital sample.We report herein the development of a simplified, more efficient cell transfer protocol that can separate and isolate both spermatozoa as well as leukocytes, with similar efficiencies as those previously reported.The new cell transfer method was used to separate sperm cells from a reconstituted mixture of spermatozoa and vaginal epithelial cells, with complete STR profiles developed from 50 cells with little evidence of contribution from the female contributor to the mixture.This modified protocol was then used to separate 21 samples of enriched leukocytes, with trapped cells ranging from 5 to 22 cells.Complete STR profiles were developed from as few as 10 leukocytes.Thus, with minimal sample preparation and a short trapping time, this method has the potential to provide an alternative to traditional differential extraction methods for separation of sperm:nonsperm mixtures while also providing versatility for separation of cells with differing morphologies.

K E Y W O R D S
DNA profiling, forensic biology, leukocytes, mixture profile, optical trapping, optical tweezer, spermatozoa

Highlights
• Optical trapping was used to separate cells from mock evidence samples collected from swabs.
• A new cell transfer method achieved full STR profiles for 6-10 leucocytes and 40-50 sperm cells.
• Optical trapping requires minimal sample preparation and only 20 min of trapping.
• The method has the potential to become a more efficient alternative to differential lysis.

| INTRODUC TI ON
Mixed samples are a common sample type encountered in forensic biological evidence, but the DNA analysis methods used often result in challenging profile interpretations.Many crime laboratories have validated integrated probabilistic genotyping software methods to deconvolute contributors from mixed profiles at the end of the laboratory workflow.However, with the development of a front-end separation method for mixed samples, different cell populations can be separated and analyzed individually, resulting in easily interpretable, single-source STR profiles.
With the optical tweezing approach, an optical trap is formed by tightly focusing a laser beam through an objective lens with a numerical aperture (NA) greater than 1.2 [13].The object of interest is trapped in the focus, or center, of the laser beam due to a gradient force that overcompensates the on-axis scattering force.Some of the critical components of an optical tweezer system already exist in forensic laboratories with the remaining components inexpensive and relatively simple to assemble; alternately, off-the-shelf trapping systems are reasonably priced.With either option, implementing optical trapping into a forensic laboratory is cheaper than some of the other explored cell-separation techniques and further, DNA analysts will be more familiar with the instrumentation used.
Optical tweezers were first introduced in biomedical research for fertilization, cell-cell interaction, embryology, microbiology, tissue engineering, single cell transfection, and other applications [14].
In one report, Reiner et al. [15] utilized optical tweezers to separate out single mitochondria from lysed human HL-60 cells, and performed multiple rounds of DNA amplification and mitochondrial DNA sequencing to detect the presence of heteroplasmic mtDNA within a single mitochondrion.
Previous work described application of the optical tweezer to forensic samples for the first time by optimizing a method and determining the required number of tweezed spermatozoa necessary to generate a full STR profile, and then trapping spermatozoa from an artificial mixture of vaginal epithelial cells and sperm cells (not extracted from a mixture swab).This resulted in a clean male profile [12].The previously used workflow involved ejecting the trapped cells onto a coverslip and placing this coverslip directly into a DNA lysis buffer.This workflow resulted in the consistent detection of DNA at the quantification stage along with expected STR profiles, which correlated with the number of cells tweezed.Overall, it was determined that approximately 50 trapped sperm cells were needed in order to obtain a full STR profile.
The purpose of this current work is threefold.First, we further optimized the droplet based extraction approach by eliminating the cover-slip support for cell deposition in the tweezer extraction step.Second, with this new methodology, we repeated our previous study by isolating ~50 spermatozoa from a sperm:vaginal epithelial mixture.Finally, we demonstrated the separation and quantification of cells from blood in order to demonstrate proof-of-concept that our tweezer-based protocol could be applied to other cell types and therefore other types of cases.

| Sample collection and preparation
Venous blood was collected from a single donor into a Vacutainer® containing EDTA (Beckton, Dickinson & Company), inverted for 15 s, and stored at 4°C until enrichment.Semen was collected by two volunteers and stored at −20°C in 500 μL aliquots.Vaginal swabs were collected by two volunteers, dried in swab boxes, and stored at room temperature for a minimum of 48 h up to 2 years.All samples used in this research were collected using informed consent and in accordance with an IRB approved human subjects sample registry and usage protocol (HM20002931 and HM20014405).
For leukocyte trapping from enriched whole blood, 1 mL of venous blood was added to 10 mL of ACK (Quality Biological Inc.), incubated at room temperature for 4 min, and centrifuged in a swinging bucket rotor at 300×g for 5 min at room temperature.The supernatant was removed and 5 mL of cold 1× phosphate-buffered saline (PBS) was added.The samples were centrifuged at 300×g for 5 min at 4°C, the supernatant was removed and 100 μL of cold PBS was added to each sample.Enriched whole blood samples were stored at 4°C for 24-72 h prior to trapping.
For simulated postcoital samples, a dried vaginal swab was placed in a microcentrifuge tube containing 300 μL ddH 2 O.Samples were incubated at room temperature for 5 min and periodically vortexed to allow the epithelial cells to elute off into solution.Neat semen was thawed on ice and diluted with bovine serum albumin (BSA) (4 mg/mL diluted with ddH 2 O) to create a 1:10 dilution.Mock sexual assault sample solutions were prepared by combining equal volumes of the vaginal cell solution and the 1:10 semen dilutions, resulting in approximate 5:1 ratios of vaginal epithelial to sperm cells.Dried mixture samples were created by immersing a sterile cotton swab in this solution and allowing it to dry.For trapping preparation, the entire cotton end of the dried swab was then cut into a tube containing 200 μL of BSA (4 mg/mL diluted with ddH 2 O) and allowed to incubate for 5 min with periodic vortexing.

| Cell separation
The optical tweezer is constructed from an AxioObserver D1 inverted microscope (Zeiss) fixed to an air-floated 3′ × 4′ vibration isolation table.For the cell isolation and capture process, a microscope slide was placed onto a motorized microscope stage (MLS203, Thorlabs) that was controlled with a joystick (MJC001, Thorlabs).
Each sample was pipetted as a droplet ranging from 300 to 400 nL onto a glass coverslip, adhered to the underside of a microscope slide with a 1-cm hole bored through the center to form a well.Hexadecane was pipetted into the well to secure the droplet into a regular shape and prevent evaporation.For trapping, an oil immersion 100× objective served as the optical tweezer focusing element for the 700 mW, 1064 nm continuous wave (CW) laser (CrystaLaser) to be focused through.The laser was attenuated with an OD1 neutral density filter, and aligned with appropriate optics into the back aperture of the microscope objective.The laser power at the trap focus was measured to be 25 mW.The infrared (IR) laser was used to trap the cells one at a time in its center and then each trapped cell was brought to the interface of the droplet and oil background.Once the cell reached the interface, it was released from the optical trap and remained there except for normal Brownian movement.Cell collection continued until the target number of cells were trapped.
Five to fifty cells were separated from each sample using the optical tweezer with cells placed in separate locations at the hexadecane:droplet interface.A glass capillary acting as a micropipette tip was used to extract the cells from the droplet via capillary action.The collection of cells in the capillary was then ejected using ≈650 hPa of pressure from a pump (Femtojet, Eppendorf) into a 1.5 mL microcentrifuge tube containing 300 μL of Buffer ATL (Qiagen™; Figure 1).At minimum, this process was performed on each sample in triplicate to ensure consistency and reproducibility of the separation method.

| DNA extraction and quantification
All DNA extractions were performed using the QIAamp DNA Investigator Kit (Qiagen™).Simulated postcoital and leukocyte samples were processed with the manufacturer's protocol "Isolation of Total DNA from Surface and Buccal Swabs" with the following modification: no carrier RNA was added at the lysis step, 20 μL of 1M DTT was added to the samples containing captured spermatozoa, and the elution volume was 30 μL.All DNA extracts were stored at −20°C.
DNA quantification was completed using the human-specific Quantifiler Trio DNA Quantification Kit (Applied Biosystem) and the ABI Prism 7500 Real-Time PCR System.The manufacturer's protocol was used with a modification for half-volume reactions.Data were analyzed using HID Real-Time PCR Analysis Software Version 1.2 (Applied Biosystems) with the threshold for the small autosomal, large autosomal, and Y targets set to 0.2 and the threshold for the IPC target set to 0.1.

| Sample concentration and STR analysis
All trapped cell DNA extracts were concentrated down to approximately 4-8 μL after quantification with the Savant DNA120 Speed-Vac concentrator (Thermo Fisher Scientific) at a low drying rate for approximately 20 min.No heat was applied.Amplification of the DNA extracts was performed using the Pow-erPlex® Fusion 5C System Kit (Promega).The manufacturer's protocol was used with a modification for half-volume reactions and run on the ABI ProFlex PCR System.The entire concentrated DNA extract was amplified for each sample analyzed.The following thermal cycler program was used: 96°C for 1 min followed by 30 cycles of 94°C for 10 s, 59°C for 1 min, and 72°C for 30 s, then 60°C for 45 min followed by a hold at 4°C.STR amplicons were separated using an ABI Prism 3130 genetic analyzer (Thermo Fisher).The PowerPlex® Fusion 5C System protocol was used with the following modifications: 0.3 μL of WEN ILS 500 and 9.7 μL of Hi-Di formamide were used for each sample.Subsequently, either 1 μL of allelic ladder or 1 μL of sample PCR product was placed into each well.Samples were injected at 3 kV for 5 s using 36 cm capillaries and POP-4 polymer.The samples were analyzed using GeneMapper™ ID-X Software Version 4.1x with an analytical threshold of 50 RFU.The generated DNA profiles were examined to check for completeness of profile, accuracy of profile (relative to the known donor profile), and for the presence of biological artifacts including allele drop-in, allele drop out, contamination, and DNA degradation.

| Analysis
Total DNA yield was calculated by multiplying the quantified concentration by the elution volume of 30 μL.A degradation index (DI) was calculated based on the ratio of small autosomal (SA) and large autosomal (LA) values as described by the manufacturer.Any sample with a DI value less than or equal to 1 indicated no degradation present [16].A simple linear regression was performed using the number of tweezed cells as the independent variable with a significance level set to 0.05 (α = 0.05) to determine the correlation between the number of tweezed cells and average peak height (APH), in addition to the number of tweezed cells and the number of complete STR loci detected.Linear regression was evaluated using JMP® Pro version 13.2.1 (SAS Institute Inc).
Other data analysis methods were performed in Microsoft Excel, and included total DNA yields, peak heights generated per cell type, quantity of cells tweezed, number of drop-in alleles, and degradation index (DI).

| Cell transfer optimization
Initially, trapped cells were isolated using the workflow developed by Auka et al. [12] with modifications that included ejecting the cells onto the 5-mm coverslip immersed in hexadecane.However, qPCR and CE data indicated that this modified workflow was not successful in developing interpretable STR profiles from cell types other than spermatozoa, which are more robust and thus less likely to lyse during the process.It is possible that the hypertonic conditions of the coverslip caused the leukocytes to lyse while drying, freeing the DNA to bind to the silica material of the coverslip.After testing several other transfer methods (Table S1), we found that pipetting the sample directly into a tube containing the lysis buffer, rather than onto a coverslip or into other solutions, increased the amount of quantifiable DNA present by 5-to 10-fold.Additionaly, eliminating the cover slip ejection step further simplifies the protocol and reduces the overall time required.Thereafter, we utilized the previously described workflow for isolating samples with the optical tweezer and followed this with direct injection of the tweezed cells into 300 μL of ATL lysis buffer (Figure 1).

| Testing of optimized cell transfer method with simulated postcoital samples
Once the optimized cell transfer method was developed, it was prudent to compare DNA yield and typing results in mixed liquid semen:vaginal fluid samples to the previously described method of ejecting cells onto a coverslip.Despite the simplification and reduction of steps in the method, the number of spermatozoa required to produce a full profile remained unchanged when the method of cell recovery was altered to direct injection of the cells into a microtube containing lysis buffer.As noted with the previous method, total DNA yield using the new cell transfer method correlated to the number of cells trapped (Figure 2A), and with the new method, 50 trapped sperm cells consistently resulted in full STR profiles (Figures 2B, 3).
F I G U R E 2 DNA yields (A) and STR profile success (B) for spermatozoa isolated from liquid semen/vaginal fluid mixtures using optical trapping.For all bars, n = 1 except for 22, 26, and 36 (n = 2) and 23 (n = 3).Error bars on observed yields represent the standard deviation from the mean for those samples with multiple replicate trapping events.Samples in excess of 23 cells showed >80% profile development, with samples containing 50 or more trapped cells consistently producing complete STR profiles.

F I G U R E 3
Representative electropherogram for 50 trapped sperm cells isolated from a liquid semen/vaginal fluid mixture using optical trapping.This dataset demonstrates the ability to obtain a full single-source STR profile no degradation and minimal carryover alleles from the female contributor.
Further, samples with more than 23 trapped sperm cells showed greater than 80% STR profile development (Figure 2B).Notably, in samples of 40, 50, 58, and 60 trapped sperm cells, the observed total DNA yield is higher than the theoretical yield.It is possible that more cells than expected could have been collected in these samples or that the higher DNA yield is the result of some free DNA in the droplet.The method of capillary action used to remove the cells has the possibility of collecting extraneous cells within the general area of the trapped cells as it is imprecise and has no fine method of reversal.Thus, if the capillary is left at the interface for too long, capillary action will continue and collect more liquid than intended.
Fortunately, this can be corrected by introducing back pressures of ≈10 hpa to slow capillary action and allow the operator to remove the capillary before extraneous cells can be collected.Additionally, of the 19 samples processed with the new cell transfer method, only six produced any carryover alleles from the female contributor in the mixture (mean < 1 per sample) and only six produced unexplained drop-in alleles (mean < 1 per sample).This observation is not unexpected due to the low copy number (LCN) nature of these samples and the results indicated a high level of sample purity capable of consistently producing clean, easily interpretable major contributor STR profiles (Table 1).Overall, these findings were consistent with our findings previously reported using the coverslip method for cell transfer [12].We note again that, in this instance, the cells were extracted from a swab rather than directly from a solution pre-mixture.

| Optimized workflow on leukocytes in enriched whole blood
After an optimized cell transfer method was established and tested using semen samples and mock postcoital samples, the same process was used to process trapped leukocytes from enriched blood samples.Samples ranged from 5 to 22 trapped leukocytes, with the theoretical DNA yield calculated based on the number of trapped cells multiplied by the amount of DNA present in a diploid cell (6 pg) [16].As noted with the postcoital mixture samples discussed above, our data demonstrate that the total DNA yields increased with increasing numbers of isolated cells and that the yields correlated with the calculated theoretical yields (Figure 4).Again, several samples produced total DNA greater than the theoretical yield, potentially due to addition of nearby cells during pipette uptake as observed with the semen samples.While this optimized cell transfer protocol does not allow for visualization of the cells after ejection (as with the previously described method [12]), the quantification data suggest a high rate of cell recovery when cells are injected directly into a tube containing the lysis buffer.Further, our data show little evidence of degradation from the trapped leukocyte samples (Figure 5 and Table S2), which supports other reports that the energy used for optical trapping is not sufficient to impact DNA integrity [17,18].STR typing results from the leukocyte samples indicated that a full STR profile could be generated using this cell transfer method with at least 10 trapped leukocytes (Figure 4B), with STR profile success positively correlating with the number of cells trapped (p = 0.0367).Two out of 6 samples with 10 trapped leukocytes generated full STR profiles (Figure 5), and one sample with six trapped leukocytes generated a 97% complete STR profile.The other samples with 10 trapped leukocytes generated STR profiles with 95.3%, 90.7%, 76.7%, and 83.7% of the expected alleles present (Table 2).
Out of the 21 trapped leukocyte samples, nine resulting in profiles containing only alleles that were consistent with the known blood donor, while 12 samples were observed to have between 1 and 7 unexpected drop-in alleles present (mean = 1.6 per sample, Table 2).
Many of the drop-in alleles observed could be attributed to operators of the optical tweezer, however several were unexplained and therefore are potentially due to stochastic effects [19,20].Overt contamination is less likely, as all extraction blanks and amplification controls passed as expected.

| CON CLUS IONS
The direct injection of trapped cells into lysis buffer as a modified cell transfer method provides for more efficient recovery of spermatozoa, and has been shown to be equally as viable with enriched leukocytes.The number of sperm cells needed for a full STR profile using our current trapping, cell transfer, and DNA analysis procedures remains unchanged from our previous method of approximately 50 cells.This could be explained using the equation derived by Lucy et al. [19] for haploid cells.Since one sperm cell TA B L E 1 STR profile data from spermatozoa isolated from liquid semen/vaginal cell mixture samples.copy number rescue methods were used to develop these profiles.

Female
A possible solution to lower the required number of trapped cells needed to produce full profiles is to use direct amplification methods rather than purification methods used to isolate the DNA from the sample after recovery [24].
In this report, we have also explored the use of optical trapping techniques on other cell types in order to demonstrate a more robust utility for the method.We have found that when this cell separation method is used on enriched leukocytes, full STR profiles can be produced with as few as 10 cells.While the enriched leukocytes used in this study were not from dried reconstituted blood, as would most likely be received as evidentiary material, this work represents an important first step towards that goal.Further, we have found that identification and trapping of unstained dried and reconstituted leukocytes in a field of erythrocytes is quite difficult and may require significant method modification.The 1064-nm trapping laser relies partially on a difference in refractive index be- Once trained, analysts will be capable of separating cells at a rate of two cells per minute with even more efficiency possible with further optimizations such as holographic trapping [25] and microfluidic control [26,27].

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflicts of interest.

O RCI D
Joseph E.

F
I G U R E 1 Schematic illustration of the optical tweezer cell separation and collection process (revised from Auka et al.).Optical tweezers are formed by aligning the beam from a 1064 nm CW laser into the entry port of an inverted microscope.The beam is expanded with a two lens telescope (L1, L2) and reflected off a dichroic mirror (DM) into the back aperture of a 100× NA 1.4 microscope objective.A CCD camera is mounted onto the exit port of the microscope and the experiment is illuminated from above (not shown) with a LED white light source.The zoomed in portion above the microscope illustrates the extraction protocol.(A) An aqueous droplet containing a mixture of cells (blue and red circles) is fixed onto a glass coverslip in an oil background (hexadecane).A micropipette tip with a tip diameter of approximately 5-10 microns is positioned nearby.(B) Optical tweezers trap a target cell and position it near the edge of the droplet.(C) This process is repeated as many times as necessary to separate the desired number of cells.(D) The pipette tip is positioned at the edge of the droplet and the capillary force draws the cells up into the tip.(E-G) The pipette tip is removed from the oil solution and placed into a centrifuge tube containing lysis buffer.Backing pressure is applied to eject the trapped cells into the tube.
of the genetic information of an individual, there must logically be some minimum number of cells required to be certain that all alleles are obtained from a sample.Using the reported equation, it would theoretically take approximately 12 cells at 99% certainty to get a full STR profile from an individual with 16 heterozygous STR loci.However, it is well established that modern DNA extraction processes result in loss of sample[20,21].If 77% of the sample is lost during the extraction process as has been previously reported, it would require approximately 50 haploid cells to obtain enough DNA to achieve a full STR profile [21, 22].van Oorschot et al. demonstrated between 20% and 76% loss from a sample following a Chelex-100 extraction and Idris and Goodwin demonstrated that the QIAamp DNA Investigator Kit (QIAGEN™) is overall less efficient than a Chelex-100 extraction [22, 23].These reports provide a clear explanation for our observations.Further, our results are consistent with the reported known minimum input DNA required for successful STR amplification (125 pg or 42 haploid cells) using modern commercial STR multiplex amplification kits [22, 23].It should be noted that besides vacuum centrifugation of the extracts, no additional low F I G U R E 4 DNA yields (A) and STR profile success (B) for enriched leukocytes isolated using optical trapping.For all bars, n = 1 except for 10 (n = 6), 13 (n = 2), 15 (n = 2), 18 (n = 2), and 22 (n = 2).Error bars on observed yields represent the standard deviation from the mean for those samples with multiple replicate trapping events.F I G U R E 5 Representative electropherogram for 10 leukocytes isolated using optical trapping.This dataset demonstrates the ability to obtain a full high quality, single-source STR profile from as few as 10 cells.
tween a cell and its environment.While spermatozoa have a sharp contrast between cell and background, leukocytes are harder to distinguish from their microscopic background.Upon drying and reconstitution, leukocytes lose their characteristic morphology and become misshapen and irregular.This lack of identifiable morphological characteristics may cause confusion between nucleicontaining leukocytes and cellular debris.To ascertain whether an object in view is a leukocyte or other material, nuclear staining may be necessary prior to trapping.The exact parameters of a staining methodology that works with the trapping laser will be explored in future research.Optical trapping has now been demonstrated to provide a simple, time-efficient method for separation of spermatozoa from mixed postcoital samples, extracted from a swab, and for isolation of leukocytes from enriched whole blood that leads to the development of complete, single-source STR profiles.The speed of trapping is dependent on relative abundance of target cells within the sample, and therefore as cell numbers are reduced, trapping times lengthen.The method of sample preparation described herein provides unique advantages over its closest competitors including lasercapture microdissection, micromanipulation, and flow cytometry, with simple reconstitution in water, PBS, or TE and subsequent trapping of sufficient cells within 20-30 min.Given that this procedure would also remove the need for a standard differential extraction, the time savings become quite substantial.With further work into other forensically relevant cell types as well as further optimization for contamination prevention, this method could easily be implemented into the front end of the forensic DNA analysis workflow.