WK5: DNA EVIDENCE

Contents lists available at ScienceDirect

Forensic Science International: Genetics

journal homepage: www.elsevier.com/locate/fsigen

Simplification of complex DNA profiles using front end cell separation and probabilistic modeling

Nancy A. Stokesa, Cristina E. Stanciua, Emily R. Brocatoa, Christopher J. Ehrhardta,⁎, Susan A. Greenspoonb

a Department of Forensic Science, Virginia Commonwealth University, 1015 Floyd Avenue, Richmond, VA, 23284, United States b Virginia Department of Forensic Science, 700 N. 5th St, Richmond, VA, 23219, United States

A R T I C L E I N F O

Keywords: DNA mixtures Flow cytometry Cell separation Probabilistic modeling TrueAllele

A B S T R A C T

Forensic samples comprised of cell populations from multiple contributors often yield DNA profiles that can be extremely challenging to interpret. This frequently results in decreased statistical strength of an individual’s association to the mixture and the loss of probative data. The purpose of this study was to test a front-end cell separation workflow on complex mixtures containing as many as five contributors. Our approach involved se- lectively labelling certain cell populations in dried whole blood mixture samples with fluorescently labeled antibody probe targeting the HLA-A*02 allele, separating the mixture using Fluorescence Activated Cell Sorting (FACS) into two fractions that are enriched in A*02 positive and A*02 negative cells, and then generating DNA profiles for each fraction. We then tested whether antibody labelling and cell sorting effectively reduced the complexity of the original cell mixture by analyzing STR profiles quantitatively using the probabilistic modeling software, TrueAllele® Casework. Results showed that antibody labelling and FACS separation of target popula- tions yielded simplified STR profiles that could be more easily interpreted using conventional procedures. Additionally, TrueAllele® analysis of STR profiles from sorted cell fractions increased statistical strength for the association of most of the original contributors interpreted from the original mixtures.

1. Introduction

One of the biggest challenges with DNA evidence is the presence of cell populations from multiple contributors which can result in de- creased statistical strength of STR profile interpretation and, poten- tially, loss of evidence. Many methods have been developed to separate contributor cell populations prior to DNA profiling including micro- fluidic manipulations [1], laser capture microdissection [2], and flow cytometry based techniques such as fluorescence activated cell sorting (FACS) [3,4]. However, one limitation of these approaches is that they have largely been demonstrated on mixtures containing only two con- tributors and/or have been applied to fresh or uncompromised mixture samples. Although probabilistic genotyping systems can perform ana- lyses on mixtures that contain three or more contributors which are superior to human analysis [5,6], limits remain as to the number of contributors that can be successfully disentangled [7]. This is particu- larly in true for mock casework samples that display stochastic im- balances that impact low level contributors, and create allelic and locus drop-out [8]. Therefore, there is still considerable need for front-end techniques that can reduce the complexity of mixtures with three or

more individuals prior to DNA analysis and facilitate the generation of single or near single source STR profiles.

The purpose of this study was to test a workflow for resolving complex biological mixtures that combines front-end cell separation with probabilistic genotyping of the simplified sorted cell fractions. A similar approach has been previously demonstrated with laser capture microdissection as the front end separation approach for enhanced in- terpretation of buccal cell mixtures containing two contributors in equal ratios [9]. We have built upon this work by processing two-, three-, four- and five-contributor mixtures where only one cell type, blood, is present. Front-end separation was accomplished using anti- body probe labelling and Fluorescence Activated Cell Sorting (FACS), a high-throughput, non-destructive cell separation technique previously described for forensic applications [3,4,10,11]. The abundance of an- tigen targets on white blood cells and average DNA yield make this a useful sample system for investigating this workflow. Additionally, complex blood mixtures may be encountered in forensic casework fol- lowing homicides with multiple victims, mass disasters, or terrorism incidents.

We employed fluorescently labeled antibody probes targeting the

https://doi.org/10.1016/j.fsigen.2018.07.004 Received 22 February 2018; Received in revised form 26 June 2018; Accepted 2 July 2018

⁎ Corresponding author. E-mail address: [email protected] (C.J. Ehrhardt).

Forensic Science International: Genetics 36 (2018) 205–212

Available online 17 July 2018 1872-4973/ © 2018 Elsevier B.V. All rights reserved.

T

A*02 allele of the Human Leukocyte Antigen (HLA) Complex to selec- tively label individual contributor cell populations in a mixture that were recovered from dried whole blood stains. Cell populations were then physically sorted into two fractions, A*02 positive and A*02 ne- gative (referred to as ‘P2’ and ‘P3’, respectively), each of which con- tained a simplified subset of contributors from the original mixture. The unsorted and sorted fractions were subjected to STR profile analysis and both human and software interpretations using the TrueAllele®

Casework System (‘TA’) for probabilistic modeling. Probabilistic inter- pretations were compared to traditional analyst assessments using standard caseworking protocols.

2. Materials and methods

2.1. Blood sample preparation

Human whole blood samples (n=9) were obtained from the Tissue and Data and Acquisition and Analysis Core Facility at Virginia Commonwealth University pursuant to Institutional Review Board protocol #870. Blood samples were screened for the HLA-A*02 allele as previously described [3]; four were HLA-A*02 positive (sample IDs 93, 96, 103, 106) and five were HLA-A*02 negative (sample IDs 94, 95, 104, 105, 107). Multiple contributor blood mixture samples of two to five donors were prepared in the ratios (volume:volume) shown in Table 1. Next, 500 μl of each whole blood mixture was dried in a petri dish and incubated at room temperature for approximately 16 h. After the incubation, cells were eluted from the surface by pipetting 1ml of 1x Phosphate Buffered Saline solution into the petri dish and transfer- ring the cell solution into a 1.5 ml microcentrifuge tube. Samples were then subjected to red blood cell lysis using the Ammonium-Chloride- Potassium (ACK) lysis buffer (Thermofisher Scientific, Waltham, MA). A 50 μl aliquot of each lysed mixture was retained for the unsorted samples and the remainder of each mixture was labeled with FITC- conjugated anti-human HLA-A*02 antibody (BioLegend, San Diego, CA). As part of our initial optimization experiments, we tested three different concentrations of antibody probe: 5 μg, 2 μg, and 0.5 μg (per 30,000 cells). No appreciable differences in the proportion of hy- bridized cells were observed between 5 μg and 2 μg samples (Figure S1). Five micrograms was used for all hybridization experiments. Mixtures were then processed using FACS to produce the sorted samples as de- scribed in [3]. Untreated blood for each of the nine contributors was used for donor reference samples.

2.2. Fluorescence activated cell sorting (FACS)

Fluorescence activated cell sorting (FACS) was performed using a BD FACS Aria II (Becton Dickenson, Franklin Lakes, NJ) in the Flow Cytometry Core Laboratory on the Medical College of Virginia campus of VCU. FACS separation of antibody-labeled white blood cells was accomplished using a 488 nm laser and gating criteria for discrimina- tion of HLA-A*02-labeled and HLA-A*02-unlabeled cells into the P2

and P3 fractions, respectively.

2.3. DNA extraction

DNA extraction was performed using the DNA IQ™ system which was previously validated for low level samples [12]. All DNA pur- ification reagents were provided in the DNA IQ™ kit (Promega, Ma- dison, WI). Briefly, samples were placed in 1.5 ml microcentrifuge tubes and cell lysis was performed in 160 μl of a Proteinase K buffer (TNE, 2.5% Sarkosyl), 20 μl of 0.39M Dithiothreitol (DTT), and 20 μl of 20mg/ml Proteinase K). Samples were incubated at 56 °C for 2 h, then substrate material was removed to a spin basket in the sample tube and centrifuged at 10,000 x g for 5min to remove excess liquid. DNA pre- parations of the blood mixture and reference samples were also per- formed using the Biomek®NXP Automation Workstation (Beckman Coulter, Inc., Indianapolis, IN) following the same process but auto- mated. The purified DNA was stored at 4 °C.

2.4. DNA quantification

DNA was quantified by real-time PCR (qPCR) using the Plexor® HY System (Promega) in a MX3005P™ Quantitative PCR instrument (Stratagene, Santa Clara, CA) equipped with Plexor® HY Analysis soft- ware, as detailed in [13]. The Plexor® HY System (Promega, Madison WI) simultaneously quantifies human and male DNA and amplifies an internal positive control that may indicate sample inhibition.

2.5. STR amplification and analysis

STR amplification of extracted DNA was performed using the PowerPlex® Fusion System (Promega, Madison, WI) in a GeneAmp 9700 thermal cycler (Applied Biosystems, Carlsbad, CA), as per manu- facturer’s protocol. The 25 μl reactions allowed for the addition of 15 μl template; the maximum amount used was 0.5 ng DNA in a STR am- plification, though most samples had much less than this in the PCR. Separation of PCR products was accomplished by capillary electro- phoresis (CE) in a 3500xl Genetic Analyzer followed by STR data analysis using the GeneMapper®ID-X v1.4 software program (Applied Biosystems, Carlsbad, CA) or data analysis using TrueAllele® Casework probabilistic modeling system (Cybergenetics, Pittsburgh, PA).

As part of our initial method development we also tested whether direct amplification and STR profiling of the sorted cell populations with the Powerplex Fusion system compared with results obtained from DNA IQ™ extraction. Direct amplification was performed according to the manufacturer’s protocol with the following modification: 15 μl PunchSolution™ Reagent was added to a PCR tube containing the pel- leted cell sample or reagent blank, mixed by pipetting, capped, and incubated at 70 °C for 30min. The entire sample was then subjected to PCR amplification. Results indicated no clear differences in the number of alleles detected across either method (comparison tables shown in Table S1). All results reported in this study were obtained using DNA IQ™ method for extraction of DNA from unsorted mixture samples, contributor reference samples, and sorted cell fraction P2 and P3.

Qualitative (analyst) assessment of STR profiles followed Virginia Department of Forensic Science (VDFS) procedures for calling alleles, examination of controls and identification of artifacts in samples. For mixture samples, allele assignment to contributors was based on com- parison to known donor reference profiles. Alleles were noted as either unique to a donor, shared with at least one other donor, or non-donor (not attributable to any of the contributors of the sample). In a case- work setting, qualitative approaches alone would not utilize all of the data present within an STR profile, underscoring the need for quanti- tative interpretation protocols such as TA. Thus, we used both quali- tative and quantitative analyses of mixtures for this study. Quantitative assessment of selected STR profiles was performed using TrueAllele®

Casework software [5,8]. This probabilistic modeling system uses all of

Table 1 Contributors and ratios for each mixture.

Number of Contributors

Mixture Ratios (vol:vol)

Contributors in Mixture 1

2 1:1 93(+):94(-) 2 1:1 95(−):96(+) 3 1:1:1 105(−):106(+):107(−) 3 1:1:2 105(−):106(+):107(−) 4 1:2:2:3 103(+):104(−):106(+):107(−) 5 1:1:1:1:1 103(+):104(−):105(−):106(+):107(−)

1 Contributors are listed in the same order as the mixture ratios. “+” or “-” indicates whether donor cell populations exhibited interactions with the HLA- A*02 antibody.

N.A. Stokes et al. Forensic Science International: Genetics 36 (2018) 205–212

206

the peak height and position data from an electropherogram to develop most likely explanations for the profile by use of Markov chain Monte Carlo (MCMC) sampling of the data. The TrueAllele® Casework (TA) mixture deconvolution process is performed in the absence of any re- ference profiles unless a reference is “assumed”. No references were assumed for this study. There is no drop-in or drop-out factor calculated or needed for the TA analysis process. Instead, the allele data, in the form of peaks, is modeled de novo for each electropherogram. Every possible allele pair combination is tested and the probability assessed to explain that mixture profile. After the mixture deconvolution process is complete, then comparisons, in the form of likelihood ratios, are per- formed for all reference profiles of interest. Moreover, the TA process requires a minimum of two reproducible independent TA analyses of the STR data, thus if a value brackets zero, small positive log(LR) for one run and small negative log(LR) for the other, it will also be inter- preted as inconclusive.

The hypothesis utilized in this study for all mixtures was as follows: the LR hypothesis (Hp) is that a person contributed their DNA to the mixture, along with N-1 unknown contributors. The alternative (Hd) is that the mixture contains N unknown contributors. Qualitative and quantitative assessments of blood samples were compared for con- currence of results.

3. Results and discussion

3.1. Blood mixture samples

Blood from five different contributors was used to prepare mixture samples derived from two, three, four or five of those donors in speci- fied ratios (Table 1). White blood cells from each of these mixture samples were labeled with HLA-A*02 antibody and sorted by FACS to the P2 or P3 fractions corresponding to cell populations that bound to the antibody probe and cell populations that did not bind to the probe, i.e., A*02 positive and A*02 negative phenotypes, respectively. The fluorescence histograms and sorting gates for the two contributor mixtures are shown in Fig. 1, while the three, four, and five contributor fluorescence histograms and sorting gates are shown in Fig. 2.

STR profiles were generated from each sorted cell population and compared to the reference profiles of the contributors for that mixture sample. For three, four, and five contributor mixtures, alleles unique to each contributor are color-coded in the genotype table for ease of vi- sualization. Each donor was assigned a color: donor 103 = gold, 104 = purple, 105 = red, 106 = green, 107 = blue. Within each subsample profile an allele unique to a donor was marked with that donor’s color (Tables 4,6,8,10), otherwise uncolored boxes indicate that allele was shared by more than one donor in the mixture. All mixture samples and sorted cell populations were qualitatively analyzed in this manner.

3.2. Blood mixture samples – two contributor mixtures

Two separate, two contributor mixtures containing an A*02 positive

and an A*02 negative contributor were created in 1:1 ratios. The fluorescence histogram of the first cell mixture (contributors 93 and 94) after antibody hybridization shows two distinct peaks consistent with the presence of an A*02 positive and an A*02 negative contributor (Fig. 1 left panel). DNA profiling of the unsorted mixture (Table 2) showed full STR profiles for both donors. After sorting, the P2 sorted fraction (A*02 positive) showed a complete, single source profile for donor 94. There were no alleles from 93 detected in this fraction (Table 2). The P3 sorted fraction (A*02 negative) showed a full profile for the negative donor, 94, with only five minor alleles consistent with 93 detected. The peak height ratio of major to minor contributor ranged between 8:1 to 10:1.

A second mixture composed of donors 95 and 96 showed similar results (Table 3). Although the fluorescence histogram showed two distinct peaks consistent with an A*02 positive and an A*02 negative contributor, cell populations exhibited more apparent overlap with less distinct differences in peak fluorescent intensity compared to the pre- vious mixture histogram (Fig. 1, right). Complete STR profiles for both

Fig. 1. Fluorescence histograms and sorting gates for 93+ 94 and 95:96 two contributor mixtures. HLA-A*02-labeled cells were sorted into the P2 fraction, and HLA-A*02-unlabeled cells were sorted into the P3 fraction.

Fig. 2. Fluorescence histograms and sorting gates for the three, four, and five contributor mixtures. HLA-A*02-labeled cells were sorted into the P2 fraction, and HLA-A*02-unlabeled cells were sorted into the P3 fraction.

Table 2 STR profiles from two-person mixture (Donors 93, 94).

94 Reference (−)

93 Reference (+)

93+94 UnSorted 1

P2 Sorted 1 P3 Sorted 1

D8S1179 13,15 10,13 10,13,15 10,13 13,15 D21S11 28,32.2 28,31 28,31,32.2 28,31 28,32.2 D7S820 12 10,11 10,11,12 10,11 12 CSF1PO 10,11 10,12 10,11,12 10,12 10,11 D3S1358 16,17 16,17 16,17 16,17 16,17 TH01 6,7 7,8 6,7,8 7,8 6,7 D13S317 11,12 8,12 8,11,12 8,12 11,12 D16S539 11,12 9,13 9,11,12,13 9,13 (9),11,12 D2S1338 20,25 19,24 19,20,24,25 19,24 (19),20,25 D19S433 13,15 12.2,15.2 12.2,13,15,15.2 12.2,15.2 13,15 VWA 17 15,16 15,16,17 15,16 17 TPOX 8,10 8,9 8,9,10 8,9 8, (9),10 D18S51 15,17 12,17 12,15,17 12,17 15,17 AMEL XY XY XY XY XY D5S818 11,13 8,12 8,11,12,13 8,12 11, (12),13 FGA 23 21 21,23 21 (21),23

1 Minor peaks are shown in parentheses.

N.A. Stokes et al. Forensic Science International: Genetics 36 (2018) 205–212

207

donors were detected in the unsorted mixture. The P2 sorted fraction contained a full profile for the positive donor, 96, with seven minor alleles detected from the negative donor, 95. The peak height ratio of major to minor contributor ranged in this mixture between 8:1 to 10:1. The P3 sorted fraction gave a complete, single source profile for the negative donor. No alleles from the positive donor were detected in this fraction (Table 3).

Results from the two-person mixtures indicate that antibody hy- bridization can be used to selectively label and sort contributor cell

populations in a dried blood sample. Easily interpretable STR profiles consistent with each contributor were obtained from the two sorted cell fractions with only minor contributions from the non-target con- tributor. The decreased separation between A*02 positive and A*02 negative populations compared to earlier studies (i.e., (3)) may be due to increases in autofluorescence after drying as suggested previously [4] or due to increases in non-specific probe interactions due to degrada- tion of cell targets after drying. Alternatively, differences in the effi- ciency of antibody hybridization may be donor-specific depending the presence of cross-reactive HLA alleles, i.e., non-A*02 antigens binding to A*02 antibody probe [14].

3.3. Blood mixture samples – three contributor mixtures

Next, three donor samples (105, 106, and 107) were used to create two separate, three contributor blood mixtures in ratios of 1:1:2 and 1:1:1. Donor 106 was HLA-A*02 positive whereas donors 105 and 107 were HLA-A*02 negative (Table 1). Therefore, in both mixtures the P2 fraction should have been enriched in donor 106 whereas the P3 frac- tion should have been enriched in cells from donors 105 and 107. For the 1:1:2 mixture, STR profiles from the unsorted mixture yielded full profiles for all three contributors and profiles from the P2 fraction primarily contained alleles consistent with donor 106 and with only two alleles from donor 107 and two alleles from donor 105 detected (Table 4). The STR profile from the P3 fraction was enriched for donors 105 and 107, with six alleles from donor 106 detected (Table 4).

Quantitative assessment by TA confirmed the qualitative results for the three contributor mixture sample (1:1:2). TA log(LR) values for the unsorted subsample were within 100-fold of each other, ranging from 9.4714 to 11.4591 (Table 5), which is equivalent to likelihood ratios of 2.9 billion and 287 billion, respectively. This indicates that it is 2.9 billion to 287 billion times more probable to observe the obtained DNA results if the person of interest contributed their DNA to the mixture,

Table 3 STR profiles from two-person mixture (Donors 95, 96).

95 Reference (A*02−)

96 Reference (A*02+)

95+96 UnSorted 1

P2 Sorted 1

P3 Sorted 1

D8S1179 14,16 14,15 14,15,16 14,15 14,16 D21S11 28 28,29 28,29 28,29 28 D7S820 11 8,10 8,10,11 8,10 11 CSF1PO 8,10 10,12 8,10,12 10,12 8,10 D3S1358 15,16 15,16 15,16 15,16 15,16 TH01 8,9 6,7 6,7,8,9 6,7,(8) 8,9 D13S317 10,13 11,12 10,11,12,13 11,12,(13) 10,13 D16S539 11 8,10 8,10,11 8,10,(11) 11 D2S1338 16,20 22,25 16,20,22,25 (16),

(20),22,25 16,20

D19S433 12,14 13.2,18.2 12,13.2,14,18.2 13.2, (14),18.2

12,14

VWA 15,18 15,17 15,17,18 15,17 15,18 TPOX 8,11 8,11 8,11 8,11 8,11 D18S51 15,17 16,23 15,16,17,23 16,23 15,17 AMEL XX XY XY XY XX D5S818 11,12 11,12 11,12 11,12 11,12 FGA 22,24 22,23 22,23,24 22,23,(24) 22,24

1 Minor peaks are shown in parentheses.

Table 4 Genotype table for the three contributor (1:1:2) blood mixture.(For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)

Red= 105, Green= 106, Blue= 107.

N.A. Stokes et al. Forensic Science International: Genetics 36 (2018) 205–212

208

along with N-1 unknown contributors, than if the mixture contains N unknown contributors. TA analysis of the sorted cell populations pro- vided quantitative support that donor 106 was enriched in the sorted P2 fraction and donors 105 and 107 were enriched in the sorted P3 frac- tion. Specifically, the sorted P2 fraction subsample yielded a log(LR) of 13.0249 for contributor 106, an enrichment of almost 100-fold greater

from the unsorted subsample. Concurrently the log(LR) values for contributors 105 and 107 in the sorted P2 subsample were -16.5399 and -11.9631, suggesting that they were excluded from the P2 cell population (Table 5). TA analysis of the sorted P3 subsample yielded a negative log(LR) value for donor 106, which indicated that this con- tributor was excluded from the P3 subsample. Donor 105 had a log(LR) value of 11.5894, an almost 100-fold increase from the unsorted log (LR) of 9.7462. Donor 107 displayed a log(LR) value of 20.5656 in the sorted P3 subsample, an increase of more than 11 orders of magnitude from the unsorted subsample (Table 5).

Although DNA was not recovered from the P3 fraction of the 1:1:1 mixture, STR profiling results from the unsorted mixture and the P2 fraction suggested efficient separation of the A*02 positive contributor from the mixture. Specifically, full STR profiles for each of the three contributors were detected in the unsorted fraction whereas a full profile for only donor 106 was detected in the P2 fraction (Table 6). Only one allele from a non-target contributor (107) was detected in this fraction. TA analysis indicated that there was only statistical support for

Table 5 TrueAllele® Casework analysis for the three contributor (1:1:2) blood mixture sample.(For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)

1Donor colors correspond to genotype charts in Table 4.

Table 6 Genotype table for the three contributor (1:1:1) blood mixture.(For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)

Red= 105, Green= 106, Blue= 107.

Table 7 TrueAllele® Casework analysis for the three contributor (1:1:1) blood mixture sample.(For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)

1Donor colors correspond to genotype charts in Table 6. 2DNA was not detected in the P3 fraction for this mixture.

N.A. Stokes et al. Forensic Science International: Genetics 36 (2018) 205–212

209

donor 106 in the P2 fraction (5.1996 106 compared to -1.7661 for 107 and -15.338 for 105, Table 7).

3.4. Blood mixture samples – four contributor mixture

The 1:2:2:3 four contributor blood mixture sample was prepared with donors 103, 104, 106, and 107. STR profiles from the unsorted subsample yielded full profiles for all four contributors (Table 8). Based on respective HLA phenotypes, P2 should have been enriched in donors 103 and 106 and P3 should have been enriched in donors 104 and 107. The actual STR profile from the P2 fraction was enriched for donor 103, as seen by the frequency of gold colored alleles in the genotype chart (Table 8, center). The P2 fraction shows few alleles from donor 106, compared to what we would expect given the dominance of their alleles in the P2 fraction of the three contributor mixtures. The STR profile resulting from the P3 fraction shows alleles consistent with donors 104 and 107, representing all alleles for both of those contributors. All al- leles consistent with 106, except for one allele at D2S1338 are also present, however a qualitative analysis does not utilize much if any allele peak height information and mixture weight assessments and thus

does not always provide a complete picture of the data. Probabilistic modeling showed evidence of all four contributors in

the unsorted mixture with LR values of ∼5.7, 6.3, 9.3, and 10.2 for donors 103, 107, 104, and 106 respectively (Table 9). After sorting, the P2 fraction showed significant enrichment for donor 103 (25.8097) and the P3 fraction showed significant enrichment for donors 104 and 107 (13.8653, 10.6459 respectively). There was only limited statistical as- sociation of donor 106 in either sorted cell fractions (3.902 in P2 and 4.5862 in P3). Overall, TA analysis provided quantitative support for one of the A*02 positive contributors and both A*02 negative con- tributors in the corresponding sorted cell fractions. Although a few unique alleles for donor 106 were detected in the unsorted mixture profile as well as the sorted P2 and P3 fractions (Table 8), the lower statistical support for 106 from TA analysis suggests proportionally fewer cells were sorted into either P2 or P3 fractions. This may be due to incorrect partitioning of donor 106 cells into the P3 fraction from inefficient antibody hybridization. We note that poor detection of al- leles from donor 106 was observed in multiple cell mixtures for this study (1:2:2:3 and 1:1:1:1:1 shown below). Direct comparison of hy- bridized cell populations from the donors 103 and 106 (both A*02

Table 8 Genotype table for the four contributor (1:2:2:3) blood mixture.(For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)

Yellow=103; Purple= 104; Green=106; Blue=107.

Table 9 TrueAllele® Casework analysis for four contributor (1:2:2:3) blood mixture sample.(For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)

1Donor colors correspond to genotype charts in Table 8.

N.A. Stokes et al. Forensic Science International: Genetics 36 (2018) 205–212

210

positive) after drying indicates that donor 103 cells have stronger in- teraction with the probe as evidenced by higher proportion of cells above 10 [3] RFU (Figure S2). Additionally, mixtures in which donor 106 is the only A*02 positive contributor exhibit lower fluorescence intensities in the P2 subpopulation compared to mixtures where donor 103 is present (e.g., P2 populations in top two histograms versus bottom two histograms in Fig. 2), further suggesting that this is a contributor-specific trend.

3.5. Blood mixture samples – five contributor mixture

The five contributor 1:1:1:1:1 blood mixture sample was composed of donors 103, 104, 105, 106, and 107. All alleles consistent with the five donors were observed in the unsorted subsample (Table 10). In a forensics laboratory this would be a very challenging mixture, not only due to the number of contributors but also because all were present in equal measure. In many, if not most forensic laboratories, the mixture would be deemed uninterpretable due to its complexity and potential

information would be lost. Donors 103 and 106 were HLA-A*02 positive and are expected to

sort into the P2 fraction whereas donors 104, 105 and 107 were HLA- A*02 negative and are expected to sort into the P3 fraction. Qualitatively, the STR profile generated from the sorted P2 fraction was enriched for donor 103, with 10 unique alleles detected from this contributor compared to three unique alleles detected from con- tributors 105 and 106 (Table 10). The STR profile generated from the sorted P3 fraction showed the highest number of unique alleles for donors 104 (= 8), 105 (= 10), and 107 (= 9), with only three alleles detected that were uniquely attributable to donors 103 or 106 (Table 10). Qualitative assessment determined that the sorted P2 sub- sample yielded all alleles for donor 103, and the sorted P3 subsample generated all alleles for donors 104 and 107 and nearly all for 105. The limited number of unique alleles from donor 106 detected in the P2 fraction are consistent with results from the 1:1:2 mixture and could indicate less efficient antibody hybridization to this contributor cell population and subsequent sorting into the A*02 negative fraction

Table 10 Genotype table for the five contributor (1:1:1:1:1) blood mixture.(For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)

Yellow=103; Purple= 104; Red=105; Green=106; Blue=107.

Table 11 TrueAllele® Casework analysis for five contributor (1:1:1:1:1) blood mixture sample.(For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)

1Donor colors correspond to genotype charts in Table 10.

N.A. Stokes et al. Forensic Science International: Genetics 36 (2018) 205–212

211

(discussed further below). Additionally, physical adhesion or clumping of target cells to non-target cells has been observed in previous flow cytometry studies [3] and could have contributed to non-specific sorting, and allelic drop-out in this experiment.

Quantitative assessment of the five contributor sample was per- formed using TA (Table 11). The unsorted subsample included all five donors, with log(LR) values ranging between 3.4128-8.4417. TA ana- lysis of the sorted P2 subsample yielded log(LR) of 28.0754 for donor 103, a value that is comparable to a single source sample [15], and negative values for the other four contributors. The sorted P3 sub- sample yielded log(LR) of 19.1138 for donor 104, an increase of nearly 12 orders of magnitude. While donor 105 produced equivalent values for unsorted and sorted fractions, the log(LR) for donor 107 increased by 6 (a million times more likely increase). Donors 103 and 106 showed LR values of -3.7186 and -12.6020 respectively suggesting that they were excluded from the sorted P3 subsample. The TA results provided quantitative confirmation that donor 103 was enriched in the sorted P2 fraction, donors 104 and 107 were enriched in the sorted P3 fraction while donor 105 stayed the same, and that donor 106 was not detected after the cell sorting process.

4. Conclusions

The data presented here suggests that antibody probes combined with FACS can be used for the front-end separation of contributor cell populations in two-person dried blood mixture samples to generate single source STR profiles. Further, for mixtures containing three, four, or five individuals, binary sorts based on the presence or absence of an HLA allele can be combined with probabilistic modeling procedures to enhance the interpretation of complex mixture samples. Results from mixtures containing three or more individuals may potentially be fur- ther improved by combining different antibody probes in the initial hybridization steps to enhance discrimination of cell populations during FACS and/or sorting cell populations into more than two fractions (i.e., non-binary sort) depending on the nature of the fluorescence histogram and the initial resolution of contributor cell populations with the mix- ture sample. Alternative antibody probes may be particularly useful for labelling contributor cell populations that exhibit decreased separation efficiency with a given probe (e.g., donor 106 with A*02 probe). With this, it may be necessary to systematically investigate binding effi- ciencies of specific antibody probes against dried cell populations containing a range of subtypes of the target allele (e.g., subtypes of A*02 described in ([16])) as well as cell populations with non-target antigens within the same cross-reactive group as the target allele [14,17].

Future studies can also make effort to collect and profile cells from the discarded fraction of the FACS instrument that result from sorting errors or events falling outside the initial gating parameters for target cells. For some complex mixtures retaining this fraction may help detect certain contributor cell populations that are incorrectly sorted and also would be a generally advantageous practice for degraded and/or low template samples. Although these results suggest that antibody based cell labelling and FACS separation may be used on dried/compromised samples, we acknowledge that as the extent of sample degradation in- creases (i.e., dried for> 24 h), decomposition of antigen targets and/or autofluorescence may present more significant obstacles. Applying this workflow to the full range of sample types and conditions encountered in forensic casework may require more robust probe targets or alter- native autofluorescence-based signatures that can be detected in aged samples [18].

Grant information

This project was funded by the National Institute of Justice Award

numbers 2013-DN-BX-K033 and 2015-DN-BX-K024. Flow cytometry services in support of the project were provided by VCU Massey Cancer Center, supported in part with funding from NIH-NCI-P30CA016059. The sponsoring agencies were not involved in the study design; col- lection, analysis and interpretation of data, or the decision to submit the article for publication.

Acknowledgements

The authors gratefully acknowledge Julie Farnsworth for providing technical assistance for this project.

Appendix A. Supplementary data

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.fsigen.2018.07.004.

References

[1] K.M. Horsman, J.M. Bienvenue, K.R. Blasier, J.P. Landers, Forensic DNA analysis on microfluidic devices: a review, J. Forensic Sci. 52 (July(4)) (2007) 784–799.

[2] S. Seidl, R. Burgemeister, R. Hausmann, P. Betz, T. Lederer, Contact-free isolation of sperm and epithelial cells by laser microdissection and pressure catapulting, Forensic Sci. Med. Pathol. 1 (2005) 153–158.

[3] L. Dean, Y.J. Kwon, M.K. Philpott, C.E. Stanciu, S.J. Seashols-Williams, T. Dawson Cruz, et al., Separation of uncompromised whole blood mixtures for single source STR profiling using fluorescently-labeled human leukocyte antigen (HLA) probes and fluorescence activated cell sorting (FACS), Forensic Sci. Int. Genet. 17 (July) (2015) 8–16.

[4] T.J. Verdon, R.J. Mitchell, W. Chen, K. Xiao, R.A. van Oorschot, FACS separation of non-compromised forensically relevant biological mixtures, Forensic Sci. Int. Genet. 14 (January) (2015) 194–200.

[5] M.W. Perlin, M.M. Legler, C.E. Spencer, J.L. Smith, W.P. Allan, J.L. Belrose, et al., Validating TrueAllele® DNA mixture interpretation, J. Forensic Sci. 56 (Nov(6)) (2011) 1430–1447.

[6] J.A. Bright, D. Taylor, C. McGovern, S. Cooper, L. Russell, D. Abarno, et al., Developmental validation of STRmixTM, expert software for the interpretation of forensic DNA profiles, Forensic Sci. Int. Genet. 23 (July) (2016) 226–239.

[7] H.M. Coyle, Sources of computational error in probabilistic genotyping software used for DNA mixture interpretation, Int. Res. J. Comput. Sci. 2 (May(5)) (2015) 12–16.

[8] S.A. Greenspoon, L. Schiermeier-Wood, B.C. Jenkins, Establishing the limits of trueallele® casework: a validation study, J. Forensic Sci. 60 (September(5)) (2015) 1263–1276.

[9] J. Ballantyne, E.K. Hanson, M.W. Perlin, DNA mixture genotyping by probabilistic computer interpretation of binomially-sampled laser captured cell populations: combining quantitative data for greater identification information, Sci. Justice 53 (June(2)) (2013) 103–114.

[10] W.M. Schoell, M. Klintschar, R. Mirhashemi, D. Strunk, A. Giuliani, G. Bogensberger, et al., Separation of sperm and vaginal cells based on ploidy, MHC class I-, CD45-, and cytokeratin expression for enhancement of DNA typing after sexual assault, Cytometry 36 (August(4)) (1999) 319–323.

[11] M. Katherine Philpott, C.E. Stanciu, Y.J. Kwon, E.E. Bustamante, S.A. Greenspoon, C.J. Ehrhardt, Analysis of cellular autofluorescence in touch samples by flow cy- tometry: implications for front end separation of trace mixture evidence, Anal. Bioanal. Chem. 409 (July(17)) (2017) 4167–4179.

[12] Forensic Biology Procedures Manual Extraction of DNA, (2017) Apr 28; Available from: http://www.dfs.virginia.gov/wp-content/uploads/2017/06/210-D2004-FB- PM-Extraction-of-DNA.pdf.

[13] Forensic Biology Section Procedures Manual: Fluorescent Detection PCR-Based STR DNA Protocols: Powerplex 16 Bio System, (2009), pp. 1–263.

[14] G.E. Rodey, T.C. Fuller, Public epitopes and the antigenic structure of the HLA molecules, Crit. Rev. Immunol. 7 (3) (1987) 229–267.

[15] M.W. Perlin, K. Dormer, J. Hornyak, L. Schiermeier-Wood, S. Greenspoon, TrueAllele casework on Virginia DNA mixture evidence: computer and manual interpretation in 72 reported criminal cases, PLoS One 9 (March(3)) (2014) e92837.

[16] S. Song, M. Han, H. Zhang, Y. Wang, H. Jiang, Full screening and accurate sub- typing of HLA-A*02 alleles through group-specific amplification and mono-allelic sequencing, Cell Mol. Immunol. 10 (November(6)) (2013) 490–496.

[17] J.A. Wade, C.K. Hurley, S.K. Takemoto, J. Thompson, S.M. Davies, T.C. Fuller, et al., HLA mismatching within or outside of cross-reactive groups (CREGs) is associated with similar outcomes after unrelated hematopoietic stem cell transplantation, Blood 109 (May(9)) (2007) 4064–4070.

[18] E.R. Brocato, M.K. Philpott, C.C. Connon, C.J. Ehrhardt, Rapid differentiation of epithelial cell types in aged biological samples using autofluorescence and mor- phological signatures, PLoS One 13 (May(5)) (2018) e0197701.

N.A. Stokes et al. Forensic Science International: Genetics 36 (2018) 205–212

212