Forensic Science International: Genetics 17 (2015) 8–16

Short communication

Separation of uncompromised whole blood mixtures for single sourceSTR profiling using fluorescently-labeled human leukocyte antigen(HLA) probes and fluorescence activated cell sorting (FACS)

Lee Dean a, Ye Jin Kwon a, M. Katherine Philpott a, Cristina E. Stanciu a,Sarah J. Seashols-Williams a, Tracey Dawson Cruz a, Jamie Sturgill b,Christopher J. Ehrhardt a,*aDepartment of Forensic Science, Virginia Commonwealth University, 1015 Floyd Ave, Richmond, VA 23284, USAb School of Nursing, Virginia Commonwealth University, Medical College of Virginia, Richmond, VA 23284, USA

A R T I C L E I N F O

Article history:Received 23 June 2014Received in revised form 30 January 2015Accepted 10 March 2015

Keywords:Forensic mixturesHuman leukocyte antigenSTRFluorescence activated cell sortingMixture interpretation

A B S T R A C T

Analysis of biological mixtures is a significant problem for forensic laboratories, particularly when themixture contains only one cell type. Contributions from multiple individuals to biologic evidence cancomplicate DNA profile interpretation and often lead to a reduction in the probative value of DNAevidence or worse, its total loss. To address this, we have utilized an analytical technique that exploits theintrinsic immunological variation among individuals to physically separate cells from different sources ina mixture prior to DNA profiling. Specifically, we applied a fluorescently labeled antibody probe toselectively bind to one contributor in a mixture through allele-specific interactions with humanleukocyte antigen (HLA) proteins that are expressed on the surfaces of most nucleated cells. Once thecontributor’s cells were bound to the probe, they were isolated from the mixture using fluorescenceactivated cell sorting (FACS)—a high throughput technique for separating cell populations based on theiroptical properties—and then subjected to STR analysis.We tested this approach on two-person and four-person whole blood mixtures where one contributor

possessed an HLA allele (A*02) that was not shared by other contributors to the mixture. Results showedthat hybridization of the mixture with a fluorescently-labeled antibody probe complimentary to theA*02 allele’s protein product created a cell population with a distinct optical profile that could be easilydifferentiated from other cells in the mixture. After sorting the cells with FACS, genetic analysis showedthat the STR profile of this cell population was consistent with that of the contributor who possessed theA*02 allele. Minor peaks from the A*02 negative contributor(s) were observed but could be easilydistinguished from the profile generated from A*02 positive cells. Overall, this indicates that HLAantibody probes coupled to FACS may be an effective approach for generating STR profiles of individualcontributors from forensic mixtures.

ã 2015 Elsevier Ireland Ltd. All rights reserved.

Contents lists available at ScienceDirect

Forensic Science International: Genetics

journal homepage: www.else vie r .com/locate / fs ig

1. Introduction

DNA mixtures are a ubiquitous problem for forensiclaboratories. Although considerable effort has been made toestablish best practices for analyzing DNA profiles containingmultiple contributors [1], there remains no standardizedinterpretation procedure for caseworking units [2–6]. Overthe last several years, many laboratory methods have beenintroduced to help separate different components of a biological

* Corresponding author. Tel.: +1 804 828 8420.E-mail address: cehrhardt@vcu.edu (C.J. Ehrhardt).

http://dx.doi.org/10.1016/j.fsigen.2015.03.0031872-4973/ã 2015 Elsevier Ireland Ltd. All rights reserved.

mixture prior to PCR amplification and STR profiling. Theseinclude protocols for differential lysis [7], centrifugation [8–10],and flow cytometry [11] as well as application of microfluidicplatforms [12,13] and laser capture microdissection [14–16].Despite some isolated successes, most of these techniques aredesigned to analyze forensic mixtures containing two cell typeswith vastly different physical and chemical properties (e.g.,sperm and epithelial cells) and are incapable of separating cellsfrom the same tissue type deposited by different individuals. Yetsome of the most common types of forensic evidence mayinvolve mixtures of the same or similar cell type [4,17,18].Recently, the increasing sensitivity of STR profiling techniqueshas led to more of these mixtures being submitted as evidence

L. Dean et al. / Forensic Science International: Genetics 17 (2015) 8–16 9

to forensic caseworking units which often leads to interpretationbottlenecks, inconclusive case results, and growing case back-logs.

Cellular immunochemistry offers a promising avenue fordistinguishing similar cells from different sources in a forensicmixture before submitting the cells to STR typing. Of particularinterest are proteins within the human leukocyte antigen (HLA) classI complex, which are surface glycoproteins expressed on mostnucleated cells [19]. Their primary function is to serve as self-recognition markers for cells within the immune system. As suchthey play a significant role in several autoimmune disorders as wellas the body’s interaction with donor tissue during organ transplants[20,21]. For forensic analysis, the key attribute of HLA antigens is thewide range of molecular structures that result from geneticpolymorphisms within the HLA coding region [21]. The diversityof HLA alleles and their respective frequencies in a number of majorpopulation groups have been extensively documented in worldwidedatabases [22]. Prior to the development of forensic STR markers,molecular methods based on HLA profiling were commonly used toanalyze biological evidence for attribution purposes [23–25]. Unlikethese earlier methods, the techniques described herein harness thevariabilityofHLAsasexpressedonthecell surface, leavingcells intactfor downstream STR analysis.

In order to utilize HLA diversity for forensic mixture analysis,methods are needed that enable labeling of specific antigens on thecell surface, followed by detection and physical separation of labeledfrom unlabeled (or differently labeled) cells. Monoclonal antibodiescomplementary to specific HLA antigens and conjugated tofluorescent dyes are commercially available and routinely used tostudy antigen expression and immune responses within humantissue [26–28]. Cells with bound antibody probes can be detectedand physically separated from unlabeled cells with fluorescenceactivated cell sorting (FACS) which is widely used in clinical andresearch contexts to analyze cellular fluorescence and isolatesubpopulations of cells based on fluorescent or other opticalproperties of interest [29].

There has been limited exploration of FACS as a viable methodfor characterization and separation of cells in forensic mixtures,including separation of sperm cells from epithelial cells [11], and,more recently, separation of uncompromised mixtures of bloodand saliva [30]. Because each of these studies used traits specific toa particular cell type as a basis of separation, they cannot beapplied to forensic mixtures comprised of one cell type. Tocomplement this body of work, there is need for an approach thatseparates mixtures based upon contributor-specific traits prior toDNA extraction. Our objective with this work was to test HLA-antibody tagging coupled to FACS for this purpose, with theultimate goal of obtaining distinct STR profiles of individualcontributors to a blood mixture. In the process, we assessedhybridization and separation efficiency of our approach in light ofissues observed with the use of HLA antibodies in the clinicalcontext such as cross-reactivity.

While recognizing that “touch” or “trace” samples (i.e., epithelialcells transferred via contact) are the most commonly recoveredtypeof complex forensic mixture, for this initial study we tested wholeblood mixtures because we anticipated that white blood cells(WBCs) would allow for more straightforward proof-of-conceptthan epithelial cells. Working with WBCs in this novel application ofFACS allowed us to refer to and build upon existing studies andmethodologies already developed to examine antigen expressionon WBCs in clinical and research contexts. Additionally, liquid bloodmixtures may be encountered in forensic casework situations, suchas pooled blood that is collected before drying, or blood samplesfrom individuals chimeric due to medical procedure or otherbiological factor, e.g., fetal contribution to mother, monochorionicdizygotic twins [31,32].

2. Materials and methods

2.1. Samples—references and mixtures

Samples of anonymized whole blood were acquired from theTissue and Data Acquisition and Analysis Core (TDAAC) at VirginiaCommonwealth University School of Medicine under existinginstitutional review board (IRB) protocol #870, and were screenedforthepresenceorabsenceof theHLA-A*02alleleusingtheAllset+TM

Gold HLA A Low-Resolution SSP Kit (Life Technologies, Carlsbad, CA),following the manufacturer’s protocol. The A*02 allele was chosenfor thisstudybecause it is relativelycommonin each of the major U.S.population groups to varying degrees—with frequencies rangingfrom a low of approximately 20% in the African American populationto a high of approximately 30% in the Caucasian population—thoughnot so common that it is not useful in discriminating betweenindividuals [22]. This makes it an attractive candidate for caseworkapplication, potentially as part of a larger panel of antibodies, asfurther discussed in Section 4.

Following HLA-A typing, one whole blood sample that testedpositive for the A*02 allele was designated as the positivereference sample for these studies (aka Contributor 1), andthree whole blood samples that screened negative for theA*02 allele were designated negative reference samples (akaContributors 2–4). A two-person mixture was created by mixing1 mL of the positive reference sample with 1 mL of one of thenegative reference samples (Contributor 2). A four personmixture was created by mixing 500 mL of the positive referencesample with 500 mL each of the three negative referencesamples.

Flow cytometry analysis (Section 2.6) was performed on anadditional set of HLA-A*02 positive and HLA-A*02 negative celllines prepared in the same manner as the positive and negativereference samples (sections 2.2–2.4) to further examine non-specific probe binding and hybridization efficiency.

2.2. Red blood cell lysis

Red blood cell lysis was performed using ACK (Ammonium–

Chloride–Potassium) Lysing Buffer (Quality Biological Inc., Gai-thersburg, MD). After 10 mL of ACK buffer was added to each bloodsample, the samples were incubated at room temperature andcentrifuged for 5 min at 201 � g. All but approximately 50 mL of theresulting supernatant was discarded, PBS was added, and thesamples were again centrifuged. After the supernatant wasdiscarded, the entire process was repeated in order to ensurecomplete lysis of red blood cells.

2.3. Antibody staining

After red blood cell lysis, 1 mL of FACS buffer (PBS containing0.5% heat inactivated FBS. Gemini BioProducts, West Sacramento,CA) was added to each reference sample and mixture and vortexed.Nine hundred mL of each sample was stained with antibody probeaccording to the following protocol, with a small aliquot (100 mL)saved for use as isotype controls (see Section 2.4). Samples werecentrifuged at 201 � g for 5 min; supernatant was discarded andthe samples were vortexed again. Each sample was treated with1 mL of blocking buffer (MACS Miltenyi Biotec, San Diego, CA) toreduce non-specific binding of antibodies, and incubated for10 min on ice. Five mL of probe—FITC-conjugated anti-humanHLA-A*02 antibody (BioLegend, San Diego, CA)—were added toeach sample, followed by a 30 min incubation on ice. FACS bufferwas added and the samples were centrifuged again. Thesupernatant was discarded and the samples were resuspendedin 1 mL of FACS buffer.

10 L. Dean et al. / Forensic Science International: Genetics 17 (2015) 8–16

2.4. Isotype control

Flow cytometric analyses need to account for the level ofbackground fluorescence caused by any non-specific binding ofantibody probes, usually via the tail (Fc) region that is shared byantibodies of the same isotype, or class. Information aboutbackground fluorescence levels is utilized to optimize FACS gatesand ultimately improve the fidelity of cell sorting (see below,Section 2.6). To this end, we used anti-mouse IgG-FITC (Santa CruzBiotechnology, Dallas, TX), which shares the same isotype (andfluorophore) as our HLA probe, to stain 100 mL each of the positivereference, negative references, two-person mixture and four-person mixture in the same manner as we stained the sampleswith HLA probe, as described in Section 2.3. These “isotypecontrols” permitted us to measure the level of non-specific bindingoccurring in our samples.

2.5. Cell imaging

Because cross reactivity of HLA antibodies is known to be aproblem in flow cytometric analyses [33], we evaluated ourselected antibody probe for non-specific binding with fluorescencemicroscopy using the AMNIS ImageStream X Mark II (EMDMillipore, Billerica, MA), a specialized imaging flow cytometerthat is able to image individual cells in multiple different modesand fluorescence channels. Aliquots (�50 mL) of a duplicate stainedtwo-person mixture, containing one A*02 positive and oneA*02 negative contributor, were suspended in the instrumentand cell images were monitored and captured in real time in theBrightfield and FITC channels. This also allowed us to assessthe relative fluorescent signal from labeled cells to ensureadequate intensity for FACS.

2.6. Flow cytometry, fluorescence activated cell sorting and gating

Fluorescence intensity profiles of the stained positive andnegative reference samples and isotype control were acquired viaflow cytometry, and were used to establish gating criteria forfluorescence activated cell sorting (FACS) of the mixture samples.Flow cytometric analysis and FACS were performed on the BDFACSAriaTM Ilu (Becton Dickinson, Franklin Lakes, NJ) flowcytometer using a 488 nm laser for fluorophore excitation. Channelvoltages were set as follows: Forward Scatter (FSC, 124 V); SideScatter (SSC, 250 V); Alexa Fluor 488 (fluorescein isothiocyanate(FITC), 475 V). The flow rate for cell sorting was set to the ‘low’

preset on the instrument which corresponds to �0.008 mL min�1.For each sample, the entire volume of cell solution (�1 mL) wasprocessed with FACS. Data analysis was conducted using FACSDivav. 6.1.3 (Becton Dickinson). To prevent non-specific fluorescence ofcellular debris (due to autofluorescence or non-specific probebinding) from confounding our analysis, an initial scatter-gatingstep based on cells’ forward-scatter and side-scatter propertieswas conducted to isolate data from single, intact cells. Givenforensic samples may be variously aged and/or degraded, differentgating strategies may be required on such specimens (seeSection 4).

Since we ultimately sought to physically separate subpopu-lations of cells within the two- and four-person mixtures basedon their relative fluorescence intensity, we used the fluores-cence data from the isotype control and the stained referencesamples to develop optimized gating parameters to permitdifferential sorting of cells on this basis. The fluorescenceintensity thresholds of gates that we designed to capturelabeled (P2) and unlabeled (P1) cells can be observed in Fig. 2,and the actual cell counts that we obtained from the sort arediscussed in Section 3.2.

2.7. STR analysis

The positive and negative sorts of the two- and four-personmixtures, unsorted two- and four-person mixtures (represented byisotype control samples), positive reference sample, and all threenegative reference samples were submitted to STR analysis. DNAwas isolated using the QIAamp DNA Blood Mini Kit (QIAgen,Valencia, CA) following the manufacturer’s protocol. The Inves-tigatorTM Quantiplex kit (QIAgen), coupled with the ABI Prism7500 Real-Time PCR System (Life Technologies), was used toquantitate the DNA extract with a half volume reaction. DNA wasdiluted to 0.2 ng/mL and PCR amplification of STR loci wasperformed on a PerkinElmer 9600 Thermocycler using theAmpF‘STR COfiler1 kit (Life Technologies) for 28 cycles. Capillaryelectrophoresis was performed on the ABI 3130 Genetic Analyzer(Life Technologies) as described in the instruction manual, andresulting data was analyzed using GeneMapper ID1 v 4.1 Software(Life Technologies) according to the manufacturer’s recommen-dations. The analytical and stochastic thresholds used to interpretthe resulting data were set at 75 RFUs (relative fluorescent units)and 150 RFUs, respectively, as determined by previous validationstudies.

3. Results and discussion

3.1. Cell imaging

Examples of cellular images captured by the Imagestreamsystem are shown in Fig. 1. Each cell analyzed was imaged in boththe FITC (Ch02) and Brightfield (Ch04) channels. Uptake of theA*02 probe can be observed in the form of green fluorescence oncells imaged in the FITC channel; cells without bound HLA probeappear black. Comparison of captured images shows that stainedA*02 negative cells did not display fluorescence (Fig. 1a), whileA*02 positive cells did (Fig. 1b); this is indicative of specific probebinding. Consistent with these results, some cells from the twoperson mixture visibly fluoresced and others did not (Fig. 1c).Brightfield imaging, a form of light microscopy, is used to examinethe structural and morphological features of cells. Any differenceobserved in Brightfield imaging of the cells displayed in Fig. 1 issimply a function of the different types of white blood cells presentin the samples (e.g., granulocytes, lymphocytes).

3.2. Flow cytometric analysis of two- and four-person mixtures

Fluorescence data from cells within the reference samples andisotype controls that met the scatter gate parameters is displayedin the form of acquisition histograms (Fig. 2(a–c)). Fluorescenceintensity of cells detected in the FITC channel, shown on the x-axisin logarithmic scale, is plotted against the number events, or cells,analyzed at a particular intensity, shown on the y-axis. Cells fromthe isotype controls (data from 2-person mixture isotype control isdisplayed as Fig. 2a) and stained A*02 negative reference sample(Fig. 2b) should and did display minimal fluorescence, and cells inthe stained A*02 positive reference sample should and did displayan optical shift (i.e., increased fluorescence) consistent with probebinding (Fig. 2c). Hybridization efficiency was assessed byquantifying the proportion of cells in each optical region on thex-axis for the A*02 negative and positive samples. The proportionof cells in the A*02 negative sample that showed positivefluorescence varied between 6 and 8% across replicate experimentswith different cell lines (see Section 2.1). The percentage of cells inthe A*02 positive sample that exhibited no fluorescence was lessthan 0.1% across all experiments.

In both the two- and four-person mixture histograms (Fig. 2dand e, respectively), two distinguishable fluorescence peaks can be

Fig. 1. Imaging flow cytometry of white blood cells in FITC (Ch02) and Brightfield (Ch04) channels. (a) HLA-A*02 negative cells, (b) HLA-A*02 positive cells, and (c) mixture ofHLA-A*02 positive and HLA-A*02 negative cells.

L. Dean et al. / Forensic Science International: Genetics 17 (2015) 8–16 11

identified, consistent with antibody probe binding the cells of onecontributor to the mixture and not the other(s). FACS of the two-person mixture captured 1.3 � 106 events, or cells, that satisfied thepositive (P2) gate and 2.2 � 106 cells that satisfied the negative (P1)gate for STR analysis. For the four person mixture, 6.8 � 105 cellssatisfying the positive gate and 1.4 �108 cells satisfying thenegative gate were collected for STR analysis. Because each sortedcell population may contain small contributions from non-targetcontributors (e.g., the positive cell fraction may also contain someA*02 negative cells with non-specifically bound antibody and,conversely, cells in the negative gate may include A*02 positivecells that are not hybridized to an antibody probe), it is not possibleto estimate cell separation efficiency from flow cytometry dataalone. Instead, we assessed separation efficiency from the relativeallelic contributions to the STR profiles of sorted cell populations(see Section 3.3).

However, the fact that the proportion of cells in the negative gatewasconsistentlyhigher thanwouldbeexpectedbased onthe relativevolumes of each reference sample combined to create the cellmixtures may be a consequence of the intrinsic differences in WBCcount per volume of blood between contributors. Given “normal”WBC count encompasses such a broad range, it is quite plausible thatthe starting cell ratio was not 1:1 (or 1:3 for the four person mixture)[34]. The higher proportion of cells in the negative gate might also bepartially attributable to stochastic differences in the proportion ofcell debris or large clumps—which would have been initially gatedout based on FSC and SSC profiles—in each reference sample;histograms of all cell events, including cell debris and clumps,support this as a potential contributing factor (data not shown).Based on the STR results, as discussed in Section 3.3, we were able torule out the possibility that this phenomenon was due to significant

numbers of A*02 positive cells ending up in the negative sort, asmight happen if the probe was not binding efficiently (ourassessment of hybridization efficiency from the reference samples,as discussed above, is consistent with these results), or if thehybridization reactions were substantially under-saturated withantibody probe.

3.3. STR analysis of sorted two- and four-person mixtures

STR analysis of the sorted two- and four-person mixtures wasperformed to determine the efficiency of cell separation. Theprofiles acquired were compared to the STR profiles of the positivereference sample (Contributor 1) and negative reference samples(Contributors 2–4), and the results are provided in Tables 1 and 2.

Results showed that the A*02 positive contributor’s cells(Contributor 1) were isolated from both the two- and four-personmixtures with high fidelity (Fig. 3; Tables 1 and 2). There was aminimal contribution from the negative reference sample(s)(Contributors 2–4) in the STR profiles developed from the positivesorts: an average 3% contribution to the profile of the positive sortof the two person mixture, and an average 4% contribution to theprofile of the positive sort from the four person mixture.Nonetheless, the large difference in peak heights between themajor and minor components permitted the complete genotype ofthe major component to be inferred without the need for mixtureinterpretation (Fig. 3). Thus, for practical purposes, the positivesort of each mixture produced a single source profile, and thatprofile matched Contributor 1.

There was an even smaller contribution (i.e., only two minoralleles, and at a lower ratio) from the A*02 positive contributor tothe negative sort of the two-person mixture, which produced an

Table 1STR profiles developed from references and two person mixture separated by FACS. Minor peaks from negative contributor are observed in the A*02 positive sort, however,there is a clear major contributor whose STR profile matches that of the known HLA-A*02 positive reference sample (Contributor 1).

D3S1358 D16S539 Amel TH01 TPOX CSF1PO D7S820

Contributor 1 15,17 11,12 X,Y 7,8 8,11 7,11 8,10Contributor 2 17,17 10,12 X,X 6,9 8,11 12,14 8,132-person mixture before sorting 15,17 10,11,12 X,Y 6,7,8,9 8,11 7,11,12,14 8,10,13A*02+ (sorted) 15,17 11,12 X,Y (6),7,8,(9) 8,11 7,11,(12),(14) 8,10,(13)A*02� (sorted) (15),17 10,12 X,(Y) 6,9 8,11 12,14 8,13

(): minor peaks.

Fig. 2. Acquisition histograms representing FITC fluorescence intensity. (a) Isotype control; (b) HLA-A*02 negative reference; (c) HLA-A*02 positive reference; (d) mixture ofone HLA-A*02 positive reference sample and one HLA-A*02 negative reference sample; (e) mixture of one HLA-A*02 positive reference sample and three HLA-A*02 negativesamples. The region designated P1 (left side of histograms) indicates a fluorescence-based gate expected to include HLA-A*02 negative (NEG) cell populations. The regiondesignated P2 (right side of histograms) indicates a fluorescence-based gate expected to include HLA-A*02 positive (POS) cells. There is clear fluorescence-based separationbetween POS and NEG cell populations in the two- and four-person mixtures.

12 L. Dean et al. / Forensic Science International: Genetics 17 (2015) 8–16

Table 2STR profiles developed from references and four person mixture separated by FACS. Minor peaks from negative contributors are observed in the A*02 positive sort, however,there is a clear major contributor whose STR profile matches that of the known HLA-A*02 positive reference sample (Contributor 1).

D3S1358 D16S539 Amel TH01 TPOX CSF1PO D7S820

Contributor 1 15,17 11,12 X,Y 7,8 8,11 7,11 8,10Contributor 2 17,17 10,12 X,X 6,9 8,11 12,14 8,13Contributor 3 14,15 11,12 X,X 5,6 9,11 10,11 9,10Contributor 4 16,16 12,12 X,X 8,9.3 8,8 11,14 8,114-person mixturebefore sorting

14,15,16,17 10,11,12 X,Y 5,6,7,8,9,9.3 8,9,11 10,11,12,14 8,9,10,11,13

A*02+ (sorted) 15,(16),17 11,12 X,Y 7,8,(9.3) 8,(9),11 7,11,(14) 8,10A*02� (sorted) 14,15,16,17 10,11,12 X,X 5,6,8,9,9.3 8,9,11 10,11,12,14 8,9,10,11,13

(): minor peaks.

Fig. 3. Electropherograms at select STR loci for two- and four-person mixtures, pre- and post-sort. (a) Electropherograms of unsorted two-person mixture show anindistinguishable mixture of four alleles at each of two loci (TH01 and CSF1PO). After probe hybridization and cell sorting, electropherograms of DNA amplified from cellscollected in the positive gate show two major alleles per locus corresponding to Contributor 1, while electropherograms of the negative sort show two alleles corresponding toContributor 2. (b) Similar results are observed in the electropherograms of TH01 and D3S1358 for the four-person mixture experiment. These loci were selected for displaybecause of the presence of minor allelic contributions consistent with the non-target cell population in the positive sort (see Tables 1 and 2). Even in the presence of minoralleles, the results are unambiguous and easy to interpret.

L. Dean et al. / Forensic Science International: Genetics 17 (2015) 8–16 13

14 L. Dean et al. / Forensic Science International: Genetics 17 (2015) 8–16

STR profile from which a full genotype—consistent with Contribu-tor 2—could be easily inferred. The negative sort of the four-personmixture produced a profile consistent with a mixture of the A*02negative contributors (Contributors 2–4), without any extraneousalleles. These results are consistent with the data we obtained inthe hybridization efficiency experiments, where it was observedthat the percentage of A*02 positive cells that failed to fluoresce(and would hence erroneously end up in the negative sort) was lessthan 0.1% (Section 3.2). It is possible that the ratio of antibodyprobe to sample is slightly lower than is optimal; other researchershave used a significantly higher concentration [30]. We willexplore this possibility in future experiments.

The presence of minor contributions from the negativereference sample(s) in the positive sort might be attributed toone or more factors. We could expect some level of non-specificfluorescence given our finding, as described in Section 3.2, that6–8% of cells from A*02 negative samples from different cell linesexhibited positive fluorescence across replicate experiments. Non-specific fluorescence could be a result of cellular autofluorescenceor non-specific probe binding. The data supports the suggestionthat this non-specific fluorescence may be, to an extent, acontributor-specific phenomenon, given the variance in fluores-cence observed in different A*02 negative cell lines. Further, all butone of the A*02 negative alleles that appeared in the positive sort ofthe four-person mixture was consistent with coming fromContributor 4 (Table 2); the one allele obligatorily consistent withContributor 3 was present at a lower ratio than the other minoralleles (1.8%). A possible explanation is that Contributor 4 possessesan HLA allele that more easily cross-reacts with A*02 antibodiesthan do the HLA alleles of the other A*02 negative contributors. Tothe extent that cross-reactivity of HLA alleles is contributing to thelow level of mis-sorted cells observed, our continuing assessmentof additional HLA antibody probes (and combinations thereof) willinclude analysis of the scope of this phenomenon (see Section 4).

It is also possible that the observed minor contributions to thepositive sort could partially be a consequence of some number ofdead or damaged cells from A*02 negative contributor(s) (whichmay autofluoresce or bind probe non-specifically), or small clumpsof A*02 positive and negative cells (as seen in Fig. 1c, which wouldbe expected to display the labeled cell’s fluorescence and thus bediverted into the positive sort) that were not screened out duringthe initial scatter gating step (Section 2.6). Consistent with thecontributor-specific differences in minor allelic contributionsobserved as described above, we have hypothesized that referencesamples might differ in their proportion of clumps and cellulardebris (Section 3.2). Any contribution from dead or clumped cellsmay be addressed with minor methodological adjustments such aslowering the flow rate, taking additional steps to reduce cellularclumping (e.g., by adding EDTA or an enzyme digestion buffer [30]),adjusting the gating to more effectively remove cellular debris andclumped cells from the sort, and utilizing a viability dye to identifyand sort out dead cells (See Section 4). Minor allelic contributionsmight also result from cell free DNA present in samples adhering toand thus being sorted with cells, though the dissimilar levels ofminor contribution in the positive and negative sorts suggest thatthis was not a significant factor in our study. Such a phenomenonmay be tissue-specific (e.g., observed more with salivary epithelialcells than WBCs), and if so could help explain the higher levels ofnon-target contributions reported in other FACS studies [30].

4. Conclusions

These experimental results show that HLA-antibody probes canbe used to differentiate individual contributors in an uncompro-mised complex cell mixture containing only one cell type, and thatFACS is an efficient way to physically isolate antibody-labeled cells

prior to forensic DNA typing. Our methodology enabled us togenerate a complete STR profile for each contributor to a one-to-one mixture of cells, where traditional DNA analysis techniqueswould have produced an indistinguishable DNA mixture. In acasework scenario, this could result in an exponential increase inthe probative value of DNA evidence.

Of course, further research is needed before applying thistechnology to evidentiary samples in forensic casework. Ongoingefforts to deconvolve mixtures that approximate those encoun-tered by forensic analysts will ultimately demonstrate the utility ofthis technique for the forensic community. We tested ourmethodology on mixtures of liquid blood of fairly substantialvolume; while similar mixtures might occasionally be encounteredin casework (e.g., reference samples containing foreign tissue of adifferent genotype [31,32]), analysts are increasingly confrontedwith epithelial cell mixtures. “Touch” or “trace” samples, amongthe most prevalent form of biological evidence submitted toforensic laboratories, may bear relatively few cells, or anindividual’s contribution to an ample cell mixture may be small.Thus, antibody-tagging coupled with FACS must be tested onepithelial cell mixtures of varying quantity and contributor ratios,including controlled and mock case contact samples.

Additionally, analysts often encounter samples that are aged ordegraded, presumably containing cells with various degrees ofcellular damage. This damage is likely to change the fluorescenceprofile of cells such that it is difficult to detect specific antibodybinding to the cell surface although few studies to date haveexplicitly examined this [30]. Accordingly, we supplemented ouruncompromised blood studies with preliminary flow cytometryanalyses of dried blood samples from an A*02 positive donor and anA*02 negative donor. Approximately 100 mL of whole blood wasdried for 24 h on a glass slide and then re-eluted with FACS buffer.Each sample was then hybridized with A*02 antibody probe andcompared to fresh blood samples hybridized in the same way. Flowcytometry results showed significant loss of intact cells betweenfresh and dried samples (>50,000 intact cells recovered versus�1000, respectively). However, the average fluorescence intensity ofhybridized cell populations from dried A*02 positive samples washigher than dried cell populations from the A*02 negative sample.While this suggests that A*02 antibody probe had a strongerinteraction with A*02 positive cell targets in dried samples (anecessity for differentially labeling cells in a mixture by HLA type),we also observed that dried A*02 negative cells exhibited higherlevels of fluorescence than fresh samples, and this fluorescencedistribution showed some overlap with that of A*02 positive cells.The comparatively low levels of fluorescence in the unstained driedcontrol samples suggest that non-specific binding of the antibodyprobe to compromised cells may be a more significant factor thancellular autofluorescence, at least for minimally damaged cells.

Therefore, it is important that future studies test differentmethods for separating dead/damaged cells from mixtures (e.g.,optimizing the initial scatter gating step (Section 2.6), usingviability stains to isolate intact cells [35,36]), and further examinethe phenomena of cell loss and non-specific fluorescence incompromised samples and strategies to help overcome theseobstacles. Although we managed to recover intact cells from driedsamples, the relatively small percentage of cells recovered indicatesa need for methodologies that both reduce the stress on damagedcells during preparation and increase their recovery [30]. It is alsonecessary to explore ways to counter autofluorescence in compro-mised cells, such as utilizing broad spectrum autofluorescenceprofiling and probe conjugates that avoid autofluorescence overlap.We hypothesize that the non-specific fluorescence observed inintact cells recovered from dried samples is a result of probesbinding to permeable cell membranes, and thus it may be beneficialto test clinical techniques for repairing membranes that may allow

L. Dean et al. / Forensic Science International: Genetics 17 (2015) 8–16 15

damaged cells to be labeled with antibody probes and properlysorted along with uncompromised cells [37].

While testing on a broader range of samples will necessitatesome adjustments to our technique, there are many promisingaspects of this experimental approach for forensic applications.The antibody labeling and cell sorting workflow is performed inapproximately 90 min which is comparable to other cell separationtechniques (e.g., differential extraction). A preliminary costanalysis showed that adding antibody labeling and FACS to thefront end of the DNA testing methodology used here (Section 2.7)increased the per sample cost by approximately 10%. Aside fromthese operational considerations which are likely to change as thetechnique is developed, the most significant advantage of thisapproach is that fluorescence measurements and subsequent cellsorting are inherently non-destructive procedures, an essentialattribute for analyzing degraded samples or evidence with limitednumbers of target cells [30]. Any dead/damaged cells that areremoved from a mixture are nonetheless collected and can besubmitted to STR analysis, as is other non-cellular material offorensic significance, such as cell-free DNA. Indeed, the dichotomybetween damaged and intact cells may in some circumstancesprovide a useful basis for sorting individual contributors to amixture, as when fresh blood is collected from a surface expectedto also bear background quantities of aged cells.

Another potential advantage of FACS is that modern cell sortingflow cytometers are capable of detecting fluorescent light ofvarying wavelengths and sorting cells into several subpopulationsat once. This allows for, and future research should investigate, thesimultaneous application of multiple antibody probes directedtoward different HLA antigens and conjugated to differentfluorophores to resolve complex mixtures of three or moreindividuals. A single antibody probe is not capable of resolving amixture of more than two contributors, unless, after sorting, theunresolved portion of the mixture is composed of clear major andminor contributors (i.e., cellular contribution significantly differ-ent). Although the number of collection tubes in the FACSinstrument limits the number of cell populations that can besimultaneously sorted, sorted cells can potentially be re-sortedwith additional antibody–fluorophore conjugates.

Considering that the percentage of individuals in relevantpopulations possessing a specific HLA-antigen allele such asA*02 can range anywhere from <1% to �45% [22,38], employingmultiple antigen probes to a single evidence sample can drasticallyincrease the likelihood that individual contributors can be isolatedfrom the mixture. Our ongoing research in this area includesassessing what combination of HLA-antibody probes would bemost likely to fully separate an unknown cell mixture into itsindividual contributors, based on the different frequencies of HLAClass I alleles in the population.

There is always the possibility (indeed, it is more of aninevitability) that a forensic mixture will come along that cannotbe fully resolved into individual contributors, even with multipleantibody probes. However, the goal is to develop a system that sortsout as many contributors with high fidelity as possible, thusminimizing or eliminating the need for mixture interpretation.Further, removing even a single contributor from a complex mixturemay increase the interpretability of and/or the statistical significanceof interpretations that can be drawn from the remaining mixture,particularly by reducing PCR competition among DNA contributorsand stochastic amplification of lower level contributors.

Acknowledgements

The authors would like to thank Dr. Joseph Conrad for helpfuldiscussions and advice throughout this project. Services andproducts in support of the research project were generated by the

VCU Massey Cancer Center Flow Cytometry Shared Resource,supported, in part, with funding from NIH-NCI Cancer CenterSupport Grant P30 CA016059.

References

[1] B. Budowle, A.J. Onorato, T.F. Callaghan, A. Della Manna, A.M. Gross, R.A.Guerrieri, et al., Mixture interpretation: defining the relevant features forguidelines for the assessment of mixed DNA profiles in forensic casework, J.Forensic Sci. 54 (4) (2009) 810–821 Jul.

[2] P. Gill, C. Brenner, B. Brinkman, B. Budowle, A. Carracedo, M.A. Jobling, et al.,DNA Commission of the International Society of Forensic Genetics:recommendations on forensic analysis using Y-chromosome STRs, ForensicSci. Int. 124 (1) (2001) 5–10 Dec.

[3] Y.-K. Chung, W.K. Fung, Identifying contributors of two-person DNA mixturesby familial database search, Int. J. Legal Med. 127 (1) (2012) 25–33.

[4] T. Caragine, R. Mikulasovich, J. Tamariz, E. Bajda, J. Sebestyen, H. Baum, et al.,Validation of testing and interpretation protocols for low template DNA samplesusing AmpFlSTR (R) identifiler (R), Croat. Med. J. 50 (3) (2009) 250–267 Jun.

[5] C. Benschop, H. Haned, T. Sijen, Consensus and pool profiles to assist in theanalysis and interpretation of complex low template DNA mixtures, Int. J. LegalMed. 127 (1) (2011) 11–23.

[6] A.A. Mitchell, J. Tamariz, K. O’Connell, N. Ducasse, Z. Budimlija, M. Prinz, et al.,Validation of a DNA mixture statistics tool incorporating allelic drop-out anddrop-in, Forensic Sci. Int. Genet. 6 (6) (2012) 749–761.

[7] K. Yoshida, K. Sekiguchi, N. Mizuno, K. Kasai, I. Sakai, H. Sato, et al., Themodified method of 2-step differential extraction of sperm and vaginalepithelial-cell DNA from vaginal fluid mixed with semen, Forensic Sci. Int. 72(1) (1995) 25–33 Mar.

[8] T. Takatori, T. Sasaki, Isolation of spermatozoa in vaginal contents bycentrifugation in a colloidal silica gradient, Forensic Sci. Int.15 (1) (1980) 61–65.

[9] J. Chen, L. Kobilinsky, D. Wolosin, R. Shaler, H. Baum, A physical method forseparating spermatozoa from epithelial cells in sexual assault evidence, J.Forensic Sci. 43 (1) (1998) 114–118 Jan.

[10] O.E. Claassens, R. Menkveld, K.L. Harrison, Evaluation of three substitutes forPercoll in sperm isolation by density gradient centrifugation, Hum. Reprod. 13(11) (1998) 3139–3143 Nov.

[11] W.M.J. Schoell, M. Klintschar, R. Mirhashemi, B. Pertl, Separation of sperm andvaginal cells with flow cytometry for DNA typing after sexual assault, Obstet.Gynecol. 94 (4) (1999) 623–627 Oct.

[12] J.M. Bienvenue, N. Duncalf, D. Marchiarullo, J.P. Ferrance, J.P. Landers,Microchip-based cell lysis and DNA extraction from sperm cells forapplication to forensic analysis, J. Forensic Sci. 51 (2) (2006) 266–273.

[13] K.M. Horsman, S.L.R. Barker, J.P. Ferrance, K.A. Forrest, K.A. Koen, J.P. Landers,Separation of sperm and epithelial cells in a microfabricated device: potentialapplication to forensic analysis of sexual assault evidence, Anal. Chem. 77 (3)(2005) 742–749 Feb.

[14] D. Di Martino, G. Giuffre, N. Staiti, A. Simone, M. Le Donne, L. Saravo, Singlesperm cell isolation by laser microdissection, Forensic Sci. Int. 146 (2004)S151–S153 Dec.

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

[16] K. Elliott, D.S. Hill, C. Lambert, T.R. Burroughes, P. Gill, Use of lasermicrodissection greatly improves the recovery of DNA from sperm onmicroscope slides, Forensic Sci. Int. 137 (1) (2003) 28–36.

[17] J.J. Raymond, R.A.H. van Oorchot, P.R. Gunn, S.J. Walsh, C. Roux, DNA trace successrates relating to volume cime offences, Forensic Sci. Int. 2 (2009) 136–137.

[18] J. Butler, C.J. Word, M.D. Coble, DNA Mixture Interpretation: History,Challenges, Statistical Approaches, and Solutions, American Academy ofForensic Sciences, Seattle, WA, 2014 Feb 21.

[19] C.A. Janeway, P. Travers, M. Walport, J.D. Capra, The Immune System in Healthand Disease, 4th ed., Garland Publishing, New York, 1999.

[20] C.L. Murphey, T.G. Forsthuber, Trends in HLA antibody screening andidentification and their role in transplantation, Expert Rev. Clin. Immunol. 4(3) (2008) 391–399 May.

[21] N. Fernandez, J. Cooper, M. Sprinks, M. AbdElrahman, D. Fiszer, M. Kurpisz,et al., A critical review of the role of the major histocompatibility complex infertilization, preimplantation development and feto-maternal interactions,Hum. Reprod. Update 5 (3) (1999) 234–248 May–Jun.

[22] D. Middleton, L. Menchaca, H. Rood, R. Komerofsky, New allele frequencydatabase, Tissue Antigens 61 (5) (2003) 403–407. . May http://www.allelefrequencies.net.

[23] C.S. Harrington, V. Dunaiski, K.E. Williams, C. Fowler, HLA DQ-ALPHA typing offorensic specimens by amplification restriction fragment polymorphism(ARFP) analysis, Forensic Sci. Int. 51 (1) (1991) 147–157 Oct.

[24] A.M. Gross, R.A. Guerrieri, HLA DQA1 and polymarker validations for forensiccasework: standard specimens, reproducibility, and mixed specimens, J.Forensic Sci. 41 (6) (1996) 1022–1026 Nov.

[25] P.M. Schneider, C. Rittner, Experience with the PCR-based HLA-DQ-ALPHA DNAtyping system in routine forensic casework, Int. J. Legal Med. 105 (5) (1993)295–299 Mar.

[26] E.H. Weiss, B.G. Lillenfeld, G. Muller, E. Muller, N. Herbach, B. Kessler, et al.,HLA-E/human beta 2-microglobulin transgenic pigs: protection against

16 L. Dean et al. / Forensic Science International: Genetics 17 (2015) 8–16

xenogeneic human anti-pig natural killer cell cytotoxicity, Transplantation 87(1) (2009) 35–43 Jan.

[27] N.C. Kaneider, A. Kaser, H. Tilg, G. Ricevuti, C.J. Wiedermann, CD40 ligand-dependent maturation of human monocyte-derived dendritic cells by activatedplatelets, Int. J. Immunopathol. Pharmacol. 16 (3) (2003) 225–231 Sep–Dec.

[28] T. Horsburgh, S. Martin, A.J. Robson, The application of flow cytometry tohistocompatibility testing, Transpl. Immunol. 8 (1) (2000) 3–15 Mar.

[29] J.P. McCoy, Flow Cytometry in Clinical Diagnosis, 4th ed., American Society forClinical Pathology, 2007.

[30] T.J. Verdon, R.J. Mitchell, W. Chen, K. Xiao, R.A.H. van Oorchot, FACS separationof non-compromised forensically relevant biological mixtures, Forensic Sci.Int. Genet. 14 (2015) 194–200.

[31] K. Chen, R.H. Chmail, D. Vanderbilt, S. Wu, L. Randolph, Chimerism inmonochorionic dizygotic twins: case study and review, Am. J. Med. Genet. A161 (7) (2013) 1817–1824.

[32] R. George, P.M. Donald, S.K. Nagraj, J.J. Idiculla, R.J. Ismail, The impact ofchimerism in DNA-based forensic sex determination analysis, Malays. J. Med.Sci. 20 (1) (2012) 76–80.

[33] W. Levering, H. Wind, K. Sintnicolaas, H. Hooijkaas, J.W. Gratama, Flowcytometric HLA-B27 screening: cross-reactivity patterns of commercially

available anti-HLA-B27 monoclonal antibodies with other HLA-B antigens,Cytometry B Clin. Cytom. 54B (1) (2003) 28–38 Jul.

[34] C.A. Doan, L.G. Zerfas, The rhythmic range of the white blood cells in human,pathological leucopenic and leukocytic states, with a study of thirty-twohuman bone marrows, J. Exp. Med. 46 (3) (1927) 511–539.

[35] L. Zamai, R. Bareggi, E. Santavenere, M. Vitale, Subtraction of autofluorescentdead cells from the lymphocyte flow cytometric binding assay, Cytometry 14(1993) 951–954.

[36] S. Augier, T. Ciucci, C. Luci, G.F. Carle, C. Blin-Wakkach, A. Wakkach,Inflammatory Blood monocytes contribute to tumor development andrepresent a privleged target to improve host immunosurveillance, J.Immunol. 185 (2010) 7165–7173.

[37] F.A. Merchant, W.H. Holmes, M. Capelli-Schellpfeffer, R.C. Lee, M. Toner,Poloxamer 188 enhances functional recovery of lethally heat-shockedfibroblasts, J. Surg. Res. 74 (2) (1998) 131–140.

[38] K. Cao, J. Hollenbach, X. Shi, W. Shi, M. Chopek, M.A. Fernandez-Vina, Analysisof the frequencies of HLA-A, B, and C alleles and haplotypes in the five majorethnic groups of the United States reveals high levels of diversity in these lociand contrasting distribution paaterns in these populations, Hum. Immunol. 62(9) (2001) 1009–1030.