World War II - Cabanatuan

When the U.S. entered World War II on 7 December 1941, the U.S. already had forces spread across the islands and nations of the Pacific. In particular, over 10,000 U.S. service members were based in the island chain of the Philippines. This country was considered a pivotal location, not only for the Americans, but for the Japanese, who were intent on using the country as a refueling location for possible attacks on Malaysia and Australia. It was considered of such import, that General Douglas McArthur based the Allied command in the Pacific on the fortress island of Corregidor located at the mouth of Manila Bay.

After the attack on Pearl Harbor, Japanese troops descended upon the Philippines, intent on taking Corregidor. Fighting against the U.S. forces and nearly 65,000 Filipino soldiers, Japanese forces were able to force the Allied troops down the island chain until they were isolated on the Bataan Peninsula (Figure 1). By this time, the troops were beset with limited support and rations, as MacArthur had abandoned Corregidor on 11 March 1942, leaving the troops with little means of escape. Facing starvation, malnutrition, and malaria, the troops were forced to surrender to the Japanese on 8 April 1942. Afterwards, the fortress of Corregidor was set upon, and fell one month later, 6 May 1942. Corregidor and the Philippines remained under Japanese control until 26 February 1945.

Figure 1. The northern region of the Philippine archipelago. The Bataan Peninsula is marked with a square. The Cabanatuan Prison Camp #1 location is marked with a star. The map is sourced from Beckenbaugh and Harris (2005) with some embellishments added by the author.

While these facts may not seem relevant to the DNA testing of remains, it does set up the story for what happened next. The opposing forces expected to take control of troops that were relatively hale and hearty and planned on marching them sixty-five miles across the Luzon Peninsula, from Bataan to Camp O’Donnell. As there was no other means of transport, the ailing prisoners were indeed marched through the jungle over eleven days in what is known as the “Bataan Death March.” An undetermined thousands of Allied troops perished along the way. Once at Camp O’Donnell, another 1500 U.S. service members and 26,000 Filipinos died in the next seventy-one days.

With the high mortality rate, the Filipino troops were granted amnesty with the agreement that they would not take up arms against the occupying forces. The remaining Allied troops were moved again, 42 miles across the Luzon Peninsula, to Cabanatuan Camp #1. At around the same time, Corregidor fell, and those taken prisoner there were remanded to Cabanatuan as well. These troops were far healthier than those that were marched from Bataan. Regardless, the daily fatality rate was substantial. By the time the camp was liberated in January 1945, there were 2,764 deaths, resulting in over 1,100 distinct gravesites.

Of particular note about the burials at Cabanatuan was that the U.S. service members took it upon themselves to initiate a sort of graves registry service, even appointing a “Graves Registration Officer.” Detailed records were taken on who perished on any given day. If possible, identification media was placed with the deceased upon burial; however, many of those transported from Camp O’Donnell had their identification removed. Rather than individual gravesites, all of those who died in a 24-hour period were placed within a single grave, referred to as a Common Grave (CG) (Trotter, 1951). Some of these might be a single individual, but more commonly the grave might contain anywhere from a dozen to three dozen individuals. The deadliest month on record at Cabanatuan resulted in 799 Allied deaths.

Once the camp was liberated in 1945, the American Graves Registry Service (AGRS) took the records from the prisoners and mapped the prison graveyard (Figure 2). The Common Graves were numbered by AGRS and the individuals historically believed to be associated with each was noted and used in the initial identification process. The burials were exhumed by AGRS and the remains removed to a temporary burial at the U.S. Army Air Forces Cemetery in Manila. Over a thousand of the deceased were preliminarily identified at this time using dental records and identification tags. Between 1947 and 1950, those as yet to be identified were exhumed and moved to the AGRS Mausoleum at Nichols Filed near Manila. By 1950, the first Cabanatuan Project was established in an effort to corroborate historical and anthropological records. The project was to be short lived, and in 1951 those deemed to be unidentifiable were interred in approximately 1000 graves at the Manila American Cemetery and Memorial (MACM), located in the heart of the city of Manila. It is estimated that between 990 and 1,006 casualties are currently unresolved from Cabanatuan POW Camp and interred at the MACM.

Figure 2. A map of the Cabanatuan Prison Camp cemetery as generated by the American Graves Registry Service (Beckenbaugh and Harris, 2005). The graves are numbered and referred to as “Common Graves” or “CG”.

This may seem to be a rather lengthy historical lead up to the modern identification efforts. However, the point here is to demonstrate that what should be a fairly straight-forward process was confounded by multiple exhumations and problematic identifications during the initial recoveries. The dental identifications were made not by the dentists in the field. Rather dental records were created by these dentists and sent back to Washington, D.C., where two officers were tasked with comparing the field notes to known records. In addition, identification efforts between 1946 and 1951 often resulting in “swapping” elements to match the number of sets of remains to the number of individuals listed in each CG. These practices resulted in a high degree of commingling, thus preventing a full return of all elements of a single individual. While it is clear that there were some misidentifications made at the time, it should also be acknowledged that AGRS was practicing the best scientific processes available at the time in the face of an enormous number of casualties.

Modern Day Efforts at Identification - Initial Testing

The initial recovery from Cabanatuan occurred in 2014 with the exhumation of Common Grave (CG) 717. Our initial information about the burial was rather limited. There was a target list of fourteen individuals, of which AFDIL had valid maternal references for twelve. The samples themselves looked good, appearing to be fairly dense, with no fracturing nor powdering that would be indicative of weathering or other environmental insults. In addition, the extraction protocol being used at the time, a modified version of Loreille, et al., 2007, resulted in a 95% success for Sanger sequencing of mitochondrial (mtDNA) regardless of the skeletal element being tested (Edson, 2019a, b). In short, we did not expect these to be a particular challenge.

In that we were very wrong. Of the 221 samples initially tested, only 86 (or 39%) generated a usable mtDNA profile. However, the results did provide an indication of how much commingling there might be (Table 1). Of the thirteen groupings of remains submitted, all except two contained at least two mtDNA sequences, with the most being eight, and one grouping failing any reportable results at all. So we did what many forensic labs do not have the leisure of doing. We asked why. This isn’t because we, as forensic scientists, don’t want to ask why. We are often restricted by any number of things: the caseload, a desired turn-around time, or our own SOPs. AFDIL has the benefit of having the ETS group as well as the DPAA laboratory full of anthropologists who are also keen on having an explanation why. DPAA also has a section of historians who might be able to provide additional information about the burials and how the remains might have been treated.

Table 1. The original distribution of mitochondrial DNA sequences generated from the samples submitted for Common Grave 717. Black fill indicates that a sample with that number designation (as noted in the first column) was not submitted. Grey fill indicates the tested sample failed to generate a mtDNA sequence. All other colors indicate a sequence was generated.

It is these historians that came across the standard manner by which AGRS would treat remains they were transporting. What is colloquially called a “hardening compound” was used to reduce decomposition until the remains could be properly interred. The composition of the compound varies somewhat depending on the availability of materials. In this particular instance (468 Graves Registration Service, 1948), it was primarily made of Plaster of Parris (25%), creta precipitate [CaCO3 (26%)], and an unspecified zinc compound (32%). The most notable of the trace additives is 2% formalin.

Formalin and formaldehyde are notable in the ability to preserve organic materials by binding to proteins. In general, it is not a friend to DNA in that it binds to the histones, and forms DNA-protein cross links. While DNA can be extracted from remains, it is unable to be “unzipped” during PCR and sufficiently amplified. Formalin/formaldehyde treatment of remains has been a problem for DNA identification of remains of U.S. servicemembers from the Korean War, specifically those who were returned through Camp Kokura and interred in the National Memorial Cemetery of the Pacific (NMCP), aka The Punchbowl, in Hawaii. These remains were saturated with a formaldehyde solution. However, it is this set of remains that has driven us to improve and finally find a solution for these types of samples.

Next Generation Sequencing (NGS, aka MPS) using a Capture based assay has enabled the sequencing of the mitochondrial DNA genome in some of the most damaged samples. The process enables specific fragments of mtDNA to be plucked from a DNA extract using custom made baits. These pieces are then captured and put through an enrichment process so the targeted human DNA can be sequenced (Marshal, et al., 2017). AFDIL implemented NGS into our casework strategies in 2016. CG717 was the first of the cases put through the NGS pipeline, increasing the success from 38% to 60% of the samples tested.

What became abundantly clear at this point was that there was a severe commingling issue, not only between the individual sets of remains being tested but potentially between the common graves themselves. CG717 was believed to contain the remains of 14 individuals. After testing was completed, we had 18 mtDNA sequences. Since mtDNA is a lineage marker, it meant we had a minimum of 18 individuals present. Twelve individuals from CG717 have been identified and returned to their families. However, it was rapidly becoming apparent that Cabanatuan as a whole would be a challenge, not only from the anthropology and DNA sides, but a bioinformatics issue as well.

Modern Day Efforts at Identification - The Next Stage

With the relative DNA testing success for CG717, DPAA decided to move forward the disinterring other caskets associated with additional common graves. With approximately 1000 unresolved and potentially 2,768 total casualties, this has rapidly become the single largest incident within a conflict we’ve collectively worked on. The extent of the commingling indicated that we would need another type of testing strategy in addition to the standard mtDNA analysis. The lineage marker would certainly be of great use and is our preferred screening method; however, it would fail to separate out individuals of the most common mito type. STR analysis at AFDIL typically involves AmpFlSTR® Minifiler™ (Thermo Fisher, Gaithersburg, MD), PowerPlex® Fusion (Promega, Madison, WI), or GlobalFiler™ (Thermo Fisher) with the manufacturer recommended parameters, and a modified AmpFlSTR® Yfiler (Thermo Fisher) protocol with double the recommended polymerase input and 36 cycles (Sturk, et al., 2009). These four strategies work reasonably well with degraded skeletal materials; however, we were not optimistic when using them with Cabanatuan samples.

As we’ve done before, we looked to modify our internal protocols so as to optimize the results we’ve obtained. In this instance, it was expected that a modified extraction protocol would be the best solution. While Loreille, et al. (2007) involves a complete demineralization of the skeletal material and an organic purification step, AFDIL has found that an inorganic purification is optimal for STR testing, regardless of the platform being used (Loreille, et al., 2010; Edson and McMahon, 2016). For Cabanatuan samples, the protocol was modified slightly to increase the sample input from 0.5 g of skeletal material to 1.0 g. The increased volume of powdered osseous material is incubated overnight at 56C in two 50mL conical tubes containing 7.5 mL of demineralization buffer (0.5M EDTA, pH 8.0; 1% N-lauroylsarcosine; 200mg/mL proteinase K). These two tubes are combined during the inorganic purification the QIAquick PCR purification kit (QIAGEN, Hilden, Germany), with a final recovery volume of 50uL (Table 2).

Table 2. The different extraction protocols in use at AFDIL. Modified organic is the technique currently being used on the Cabanatuan samples. Modified organic uses 1.0g of powdered osseous material split evenly into two volumes, effectively doubling the extract. These two volumes are then combined during QIAquick purification with a final recovered volume of 50µL. The other two protocols are commonly used for mtDNA (organic) and STR (inorganic) downstream processing. Further details on these two protocols can be found in Edson and McMahon (2016).

The modified extraction method gave an approximately 20% success for all samples tested with Minifiler™ and modified AmpFlSTR® Yfiler and a 55% success for PowerPlex® Fusion. It is of note that the success of PowerPlex® Fusion is somewhat artificially elevated in that samples tested in PowerPlex® Fusion were those that were successful in Minifiler™. While these numbers are not outstanding when compared to fresh samples or even our regular casework, it was far better than what was expected. Again, we had to ask why.

The scientists leading the STR testing of the Cabanatuan complex decided to make note of any possible metrics that could help us determine the answer to that question. The average quant value as calculated with Plexor® HY DNA Quantification Kit (Promega) was less than 0.05ng/uL, indicating a low probability of success. However, all of these samples were also tested in NGS, allowing us to examine metrics that would not commonly be collected when only performing STR analysis. What they found was that the measure of unique start positions (USP), a common measure for determining the quality of NGS tested samples, tended to be higher for any sample that worked in STRs. In fact, samples that generated a reportable modified AmpFlSTR® Yfiler or Minifiler™ profile of four loci or more almost always had an average USP value of at least 100.

What was not relayed in the introduction is that the final burials were subjected to a high degree of ground water contamination. Manila is located in a tropical rainforest environment, with an average daily temperature of 82°F (27.8°C). The average annual rainfall is approximately 80 inches (2032mm). The MACM is located in the center of a densely populated urban environment, and water may collect in that area. Most caskets upon exhumation are full of water and need to be drained before being lifted from the burial site. Of note, many of the osseous remains tested showed little to no evidence of water damage. Our hypothesis is that the formalin present in the hardening compound actually aided in the preservation of the skeletal materials. By binding to the proteins in the skeletal matrix, the formalin prevented the swelling and fracturing that often occurs when bones are exposed to water for an extended period of time. As a corollary to better structural integrity, a greater amount of high-quality DNA was preserved and was able to be recovered.

Future Testing

At present, we have tested 1,635 samples associated with the complex, representing approximately 300 sets of remains out of the approximately 1,000 total unknowns. Of the potential 2,768 individuals involved in the incident, we have generated 218 mitotypes, less than a mere 10%. Data continues to be collected from the elements tested. We aim to be able to determine additional trends that could streamline the testing of this rather enormous project. Not only are we looking at the materials being tested, but we are planning to examine the soil composition in the MACM itself. Perhaps the fertilizers and pesticides will give us further insight into the trends associated with DNA testing and help us with identification efforts associated with other disinterments.

Testing of the remains recovered from the Cabanatuan prison camp will continue for many years to come. We will continue to update and modify our testing strategies to increase our success and bring more of our fallen service members home.