Impact of Contamination Upon Reliability of DQA Typing: Lessons from the O.J. Simpson Case

The O. J. Simpson case subjected forensic DNA testing to the most intense scrutiny to date. Discovery materials were extensive and derived from three separate laboratories (the LAPD-SID laboratory, Cellmark Diagnostics, and the California Department of Justice DNA Laboratory). This enabled an unprecedented comparison of forensic laboratories, and an objective analysis of actual data relevant to many of the criticisms and concerns expressed by critics of forensic DNA testing. For this particular case, the DNA testing involving the Polymerase Chain Reaction (PCR) was most severely criticized. Many of the concerns regarding PCR-based forensic testing were previously suggested in the National Research Council (NRC) report (1). Although the NRC arguments were founded primarily on theoretical grounds, testimony and discovery data presented during the Simpson case confirmed their validity. The specific concerns regarding forensic DNA testing authenticated by evidence presented in the Simpson trial are discussed in detail below. The controversy surrounding most of these specific issues has been previously reviewed in a number of publications (1,2,3,4). Therefore, my discussion here will be limited to specific evidence introduced during the Simpson trial directly related to concerns regarding DNA testing.

The manipulation of evidence prior to arrival at the DNA laboratory can introduce a substantial risk of cross-contamination. The PCR technique is extraordinarily sensitive, with the ablility to detect only a few molecules of human DNA. However, the method is not discriminating as to whose DNA is detected and it is easy for human DNA other than that of the evidence item to contaminate the sample. The most subversive form of contamination involves the inadvertent transfer of DNA from one item to another during evidence handling. This cross-contamination of one evidence item with another could potentially result in false matching. There are no specific laboratory controls that currently detect cross-contamination. Furthermore, once DNA is transferred, it will be consistently typed as a match regardless of how many genetic markers or different laboratories are used to confirm the analysis. The likelihood of cross-contamination occurring is increased when investigative personnel are careless and/or unfamiliar with the PCR technology and the evidence consists of very low concentrations of excessively degraded DNA.

During testimony, the criminalists involved with the Simpson case admitted that they received no training specific to the collection of evidence that ultimately would be tested by DNA or PCR-based methods. The evidence was not preserved correctly and there was excessive degradation of the DNA. Procedures followed indicated a lack of aseptic technique. Gloves were only changed infrequently, when they appeared dirty. Although substrate controls were interspersed between evidence items by the LAPD, there was evidence that they were not always manipulated in parallel with the evidence at every stage of analysis. Specifically, items were sorted and mailed to the other laboratories without the corresponding control items. No one that testified seemed sure of the significance of handling these controls in a manner that might detect cross contamination. The order of evidence collection (i.e. Rockingham, then Bundy, then back to Rockingham) was such that cross contamination resulting in false matching could possibly occur.

The manipulation of degraded, low concentration DNA in the proximity of high concentration, high molecular weight DNA can introduce substantial risk of cross-contamination. On the morning of June 15, 1994, an LAPD criminalist opened a vial of Mr. Simpson's blood in order to prepare exemplar blood stains. Some of this blood soaked through a chemwipe onto the criminalist's glove. Although the glove was changed, the area was not bleached or cleaned before immediately proceeding to analyze first the Rockingham glove and then the Bundy crime scene blood drops. The manipulation of the reference blood containing high concentrations of high molecular weight DNA in the proximity of evidence with extremely small concentrations of extremely degraded DNA introduced the possibility of cross-contamination from Mr. Simpson's reference blood to the evidence items.

On the next morning, the LAPD criminalist extracted DNA from evidence collected at Rockingham (where Mr. Simpson acknowledged he cut himself) and then extracted DNA from reference specimens of Nicole Brown Simpson and Ronald Goldman. When these reference specimens were typed by the LAPD and later by Cellmark Diagnostics and the California Department of Justice DNA Laboratory, faint additional alleles were observed. The results are consistent with the interpretation of cross contamination from O.J. Simpson's blood into Nicole Brown Simpson's reference blood as illustrated in Figure 1. The contaminant alleles were recorded as the polymarker GC B allele by Cellmark Diagnostics and as the DQ alpha 1.2 allele by the California Department of Justice. The only other possible explanation would be a pattern of cross-hybridization in three different labs and in two different genetic systems that coincidentally shows these alleles. These observations lend credence to the plausibility of cross-contamination and were never disputed by the prosecution.

Controls as currently performed are inadequate in number and design, and are not designed to always reveal contamination. The DQ alpha dot blot typing system, when performed according to the recommended protocol, is very specific. Typed results of a reference sample from a given individual should only contain one or two visible alleles. The appearance of additional alleles, even of weak intensity, is an indication that something is potentially wrong with the typing. The detection of a third allele can be the result of procedural errors: failure to maintain the proper analysis temperature or salt concentration; using too much DNA so that background cross hybridization becomes evident; and, most importantly, contamination of the sample with extraneous human DNA or PCR amplicons. Contamination by extraneous DNA can sometimes (but not always) be confirmed by the appearance of alleles on extraction or "no DNA" (water) controls. Evidence of low level contamination can also become evident by the random appearance of weak alleles throughout a typing run. Low level contaminant alleles might be faint, especially in reference samples, since they must compete with the sample alleles for amplification; however, the same low level contaminant alleles in a sample with low DNA concentration might be scored as a genuine typing result, causing a typing error.

Evaluation of evidence for the Simpson case included an extensive review of 1069 total strips typed by the LAPD laboratory from May 1993 through August 1994 (509 exempla, 129 extract blanks, 42 amplification blanks). All instances of unexpected additional alleles on known exempla or negative control were recorded as contaminants and/or PCR artifacts (Figure 2). These were then confirmed as contaminants if they were found as third allele on known exempla several times within a run, or found on negative substrate or on amplification blanks within the same run, or if the contaminants were alleles not explainable by cross-hybridization and/or stringency wash problems (i.e. 2,3,4 alleles). Then, the contaminants were analyzed by allele (Table 1) and the source of contamination by control (Figure 3).

Results of the evaluation of the LAPD validation and control specimens typed between 5/20/93 and 8/25/94 documented a pattern of accumulating contamination reaching major proportions. For example, after observing no contamination during the first month DQ alpha was run, there was 66/437 (15%) contamination in 1993, and an increase to 99/243 (41%) in 1994. Contamination was observed for every month from June of 1993 through August of 1994 and ranged from a low of 5% to as great as 60% of the exempla and/or control strips. Every DQ alpha allele was observed as a contaminant, but the incidence of each fluctuated from month to month. The LAPD laboratory experienced the most severe incidents of contamination in January of 1994 and again from April of 1994 through at least May of 1994. Due to the minimal number of control items evaluated in June of 1994, it is difficult to estimate the extent of the problem during the time period when evidence for the Simpson case was handled. However, contamination of extraction controls in June of 1994 confirm that the problem existed at this time. The predominant contaminant alleles observed from April through August of 1994 include 1.1, 1.2, 1.3 and 4. These are the precise alleles critical to evaluation of the evidence in the Simpson case. Although it is not possible to precisely pinpoint the source of contamination, the fact that from May through August of 1994 only 9% of the no DNA (water) controls were contaminated compared with 33% of the extraction controls indicates sample handling or DNA extraction as the most likely source of contamination.

Therefore, firm evidence exists to show that the LAPD laboratory experienced a chronic persistent contamination problem during this time. The observed contamination also provided for an analysis of how efficiently the controls detect low level contaminants. The negative amplification and extraction blank controls did not always reveal contamination. Contamination was revealed only as third alleles on exemplars for 28% of the contaminated runs. The contaminant alleles at the LAPD laboratory appeared and disappeared at random from run to run and month to month. These observations indicate that the number of controls run by forensic labs may be inadequate to insure that contaminants are not present.

Although this type of contamination would be expected to only cause random errors, it is an indication of how careful the laboratory handles specimens. A laboratory with frequent contamination incidents is likely to also have cross-contamination for which there are no controls. It is unlikely that specimen handling would be performed with greater care than DNA extraction or PCR typing. The observed contamination rate in a laboratory might therefore be expected to be an underestimate of the true contamination and error rate.

DNA typing errors can occur at or around the time of evidence testing as a result of low-level (i.e. less than C dot) contamination. The significance of observed contamination upon the reliability of forensic DNA typing is controversial. It is frequently interpreted as inconsequential by proponents of the testing, while opponents consider it to invalidate the results. Proponents argue that if the contaminant allele is less intense than the C dot, it could not be confused with or overwhelm the true alleles found in evidence items, even those with low DNA concentrations. Opponents argue that the DQ-alpha system is not a quantitative method. Dot intensity can fluctuate so that a signal less than C on one strip may show up as greater than C on other strips in the run, especially at low DNA concentrations. Furthermore, any observation of contaminant DNA indicates that alleles in addition to those noticed to be contaminants might also be present. This is supported by the random appearance and disappearance of contaminants as observed in the LAPD data in the Simpson.

Review of the LAPD laboratory typings for known exempla and proficiency specimens provides proof that low-level (less than C dot) contaminants can cause errors in typing. There were five such errors discovered in the LAPD laboratory. On May 25, 1994, a hair shaft from an exemplar DQ alpha 2,3 hair clearly typed as a 1.2,2. This is an example of a contaminant overwhelming a true allele in a specimen with very little DNA. The negative controls for this run-date were clean. However, a contaminant 1.2 allele was observed for the shaft of another hair in this run, and a contaminant 1 allele was observed on a cloth control run on the preceding day. It should be emphasized that the controls for the specific run were inadequate to detect the contaminant. The pattern of contamination only became evident by looking at the controls for the run of the preceding day. The contaminant observed is DQ alpha 1.2, a critical allele in the O. J. Simpson case, and found at the LAPD laboratory only three weeks prior to testing the Simpson evidence.

In another incident, a reference blood specimen from a proficiency test was incorrectly typed in two out of three runs. The DQ alpha 1.2,4 blood clearly showed a 1.3,4 typing in the first run. For this date, weak 1.2,4 (no C) DNA was evident on a cloth control. Notice that the 1.3 contaminant allele was not observed on the controls. In a repeat run the blood again showed a 1.3,4 type. This time the cloth control had weak 1.2,1.3 (no C) DNA. Once again, the controls had levels of contaminants less than C, yet showed up as typable on the blood reference, and the 1.3 was identified as a contaminant on only one of two controls. In the final run, the controls were clean and the blood correctly typed as a 1.2,4.

The remaining two errors involved a differential extraction of a mock sexual assault standard. Despite observing sperm microscopically, the sperm alleles were not detected on two separate occasions. From this, the possibility of missing alleles arises when there are very small amounts of DNA present as a mixture.

The interpretation of mixtures of DNAs is problematic. Because of the extreme sensitivity of PCR, mixtures of different human DNAs are likely to occur with specimens found at crime scenes. Unlike interpretation of typings where the specimen is known to have come from a single individual, with mixtures it is not possible to determine if additional alleles are actually due to a mixture of DNA from more than one contributor or due to contaminants and/or artifacts such as cross-hybridization. Interpretation of forensic results consistent with the presence of a DNA mixture can therefore be rendered subjective and ambiguous. This was accentuated in the Simpson case where the California Department of Justice DNA Laboratory interpreted a weak DQ alpha 1.3 as consistent with a mixture including Ronald Goldman for a Bronco item, yet ignored DQ alpha 1.3 alleles of similar dot intensity observed on Bundy blood drops and positive controls.

It is not possible to determine the DNA concentration attributable to each contributor to a given mixture, presenting yet another problem for interpretation of results. Consequently, the DNA of the major contributor might be present at sufficient amounts to be detected on the C dot, yet the DNA concentration of the minor contributor might be at such low levels that preferential amplification might occur, resulting in missed alleles from the minor contributor's type. The California Department of Justice DNA Laboratory's DQ alpha typing for the Bronco steering wheel item 16/(29), for example, was a 1.1,1.2 with weaker 4 allele. When discussing this result in its report the DOJ DNA Laboratory states that "With regard to bloodstain #DNA-16/(29), neither victim's full type is evident in the results for the minor component (i.e., the "4" allele). The minor contributor could be type 4,4. However, considering that this is the minor component of a mixture with a low starting DNA (total 0.4 ng), a clear exclusion of Goldman (1.3,4) as the contributor to the minor component is precluded." In other words, even when a suspect's allele is not present, he may still not be excluded as a possible contributor. Under these circumstances, alternative interpretations are possible, rendering conclusions subjective and open to bias.

In a case with multiple victims and an unknown number of potential assailants such as the Simpson case, many of the genetic markers can be rendered uninformative. For example, simply mixing the blood of Nicole Brown Simpson (DQ alpha 1.1,1.1) with that of Ronald Goldman (DQ alpha 1.3,4) would result in a mixture of alleles consistent with the inclusion of O. J Simpson (DQ alpha 1.1,1.2) even when his blood is not present. This same mixture would also falsely include individuals with DQ alpha types 1.1,1.1; 1.1,1.3; 1.1,4; 1.2,4; 1.3,4; or 4,4 as potential third contributors. Similarly for all of the two allele polymarker genes (LDLR, GYPA, and D7S8), mixing the types of the two victims results in 100% of the population being consistent with inclusion as a third contributor in the mixture. It is not clear how to deal with the statistical interpretation of such mixtures in order to fairly account for all individuals potentially falsely included and how to weight the uninformative genetic markers.

Laboratory error rates should be defined through routine blind proficiency testing. Much of the evidence in the Simpson case consisted of extremely degraded blood mixtures containing very low concentrations of DNA. In order to prove how reliably such evidence can be typed, there should be proficiency testing done on similar specimens. However, none of the proficiency tests performed by any of the labs in the Simpson case are of this sort. Furthermore, the proficiency tests are few in number and the analyst is aware of the test. This sort of open proficiency test frequently is the most frequently used type. The LAPD laboratory has a good record on its proficiency testing in spite of its contamination problems. When inspecting the raw data for these tests, obtaining a correct type required repeat testing until a consistent pattern emerged. The Simpson case underscored the need for blind proficiency testing in order to more accurately determine error rates on a continuous basis for each lab and each analyst. It is clear that different forensic labs can vary considerably in their experience level and quality. The error rate during the analysis of a particular case will vary within in a lab, dependent upon the case load volume, periodic contamination events, staff turnover, and analyst experience. An estimate of error during analysis of each case is critical to the defendant and his defense, and is relevant and necessary to a jury's assessment of the weight to be given such evidence.

Many of the concerns discussed above could be addressed by incorporating the following suggestions into the protocol for crime scene collection. First, police detectives and criminalists should receive specific training for DNA evidence collection. This training should emphasize aseptic technique in order to minimize cross-contamination risks, as well as methods of preserving DNA evidence and collection strategy. By incorporating the following additional controls during evidence collection, most of the other concerns discussed above would be addressed.

Second, substrate controls introduced at the time of evidence collection should be collected for all cases. Preliminary experiments are required to assess whether carrier DNA is necessary to stabilize low concentrations of DNA on these control swatches. The blank swatches should be manipulated so that there is a control interspersed between all evidence collected. In situations where the evidence is not conducive to collection of a substrate control (i.e. too bloody to find an adequate area), a blank manipulation swatch should be collected. These crime scene negative controls could then be processed in parallel with the evidence at every stage of manipulation and/or analysis. These controls will determine the background levels of DNA at the crime scene as well as to incorporate additional negative controls to better detect cross-contamination. Extraction blanks and negative amplification controls should continue to be performed in order to detect contamination specific to the particular PCR laboratory.

Finally, exempla should be coded by the investigator and not revealed to the laboratory. Blind control exempla and specimens that mimic evidence (including blood mixtures, low concentrations, degraded specimens, etc.) could be manufactured by an agency such as the College of American Pathologists (CAP). These could then be purchased by agencies responsible for crime scene collection. At the time of evidence collection for each case, the criminalist could include a coded exemplar control and a control item similar to the evidence, which would act as blind proficiency controls for that particular case. Cumulative analysis of the results for these controls would then provide data to determine error rates for each laboratory and each analyst.

Attempting to perform reliable PCR based testing for forensic investigations is indeed a monumental task. The likelihood of obtaining an uncontaminated PCR result from crime scene evidence is much lower than from a clinical or research environment. Even if every possible precaution is taken to prevent contamination, the crime scene evidence is likely to produce DNA mixtures, further complicating the interpretation of the results of the typing. Such concepts are easily grasped by a jury, as was clear in the O. J. Simpson case. Accordingly, successful introduction of PCR-based forensic DNA evidence will require proof that low concentration DNA mixtures can be reliably typed, supported by routine blind proficiency testing. In addition, a careful strategy of contamination control measures must be thoroughly documented, starting with evidence collection at the crime scene and continuing through every stage of evidence analysis.

* People of the State of California v. Orenthal James Simpson. LA County Superior Court no. BA097211. Mazzola 4/25/95 at 24192 & 24325. Yamauchi 5/30/95 aaat 29739-297342. Fung 4/3/95 to 4/17/95.

		% of Total by allele
Total	Month	1.1	1.2	1.3	2	3	4	1.2/1.3/4
32	May-93	0.0	0.0	0.0	0.0	0.0	0.0	0.0
50	Jun-93	4.0	0.0	14.0	0.0	0.0	0.0	14.0
56	Jul-93	1.8	3.6	11.0	0.0	0.0	0.0	14.6
38	Aug-93	2.6	0.0	2.6	0.0	0.0	2.6	5.2
136	Sept-93	2.2	0.0	2.2	0.0	0.0	6.6	8.8
97	Oct-93	15.5	0.0	11.3	0.0	0.0	5.2	16.5
11	Nov-93	18.0	0.0	18.0	0.0	0.0	18.0	36.0
17	Dec-93	11.8	0.0	11.8	0.0	0.0	0.0	11.8
50	Jan-94	34.0	2.0	12.0	40.0	0.0	36.0	50.0
8	Feb-94	37.5	0.0	0.0	0.0	0.0	0.0	0.0
5	Mar-94	0.0	40.0	0.0	40.0	0.0	20.0	60.0
46	Apr-94	15.2	8.7	15.2	37.0	0.0	39.0	62.9
45	May-94	17.8	2.2	6.7	4.4	0.0	8.9	17.8
16	Jun-94	6.3	6.3	6.3	0.0	0.0	6.3	12.5
12	Jul-94	8.3	0.0	0.0	0.0	0.0	8.3	8.3
61	Aug-94	1.6	3.3	8.2	3.3	4.9	16.4	26.2