Blood Cell DNA Fingerprints Could Expose Hidden Inflammation in Diseased Tissue

Every cell in the human body carries essentially the same genome, yet each cell type behaves in a distinctive way because of chemical marks layered on top of the DNA. One of the most stable of these marks is DNA methylation, the addition of a methyl group to cytosine bases, most commonly at cytosine-guanine dinucleotides known as CpG sites. Because the enzyme DNA methyltransferase 1 copies methylation patterns from the template strand to the newly synthesised strand during cell division, these marks can persist across generations of cells and act as a long-term biological memory. The result is that each tissue carries a methylation signature reflecting its developmental history, and researchers can exploit those signatures to work out what kinds of cells are present in a mixed sample. A new study published in Epigenetics Communications has taken this principle and applied it to one of the most clinically important questions in pathology: how to detect when immune cells have invaded otherwise healthy tissue.

The research team, led by M. J. Dunnet and Tim Hore of the University of Otago in New Zealand, set out to find genomic regions that could cleanly separate leukocytes, the white blood cells that drive inflammation, from every other tissue in the body. Their starting point was a comprehensive human methylation atlas containing data for more than 423,000 CpG sites across blood cell preparations and nine other tissue types that were believed to be free of immune and vascular contamination. For each CpG site, the team calculated the difference between the average methylation level in leukocytes and the average across all non-leukocyte tissues. Sites with a difference of 0.8 or more on a scale from 0 to 1 were classified as leukocyte-specific differentially methylated positions, or ls-DMPs. The screen yielded 77 such sites, of which 19 were highly methylated in blood cells and 58 were essentially unmethylated.

From this list, the researchers focused on the only two CpG islands containing multiple leukocyte-specific sites. The first sits within the HOXA3 locus on chromosome 7, and the second lies within exon 26 of the MAP4K1 gene on chromosome 19. At both loci, the top CpG sites showed very high methylation in leukocytes, moderately low methylation in adipocytes, lung epithelial cells and vascular endothelial cells, and virtually no methylation in colonic epithelium, cortical neurons, hepatocytes and pancreatic cell types. When the team examined methylation across ten different immune cell populations, all mature leukocyte subpopulations were highly methylated at both markers. Intriguingly, the pattern appears to be laid down early in development: foetal liver CD34-positive haematopoietic stem cells showed significant demethylation compared with their adult counterparts, suggesting that methylation accumulates specifically in the foetal stem cells of the blood-forming lineage and is then maintained through every subsequent stage of differentiation and ageing.

Array-based methylation platforms, which measure thousands of CpG sites simultaneously, are powerful discovery tools but are expensive per sample and rely on single probes that risk under-sampling each region. To turn their discovery into something closer to a diagnostic test, the researchers adapted a rapid, cost-effective amplicon bisulphite sequencing assay. Bisulphite treatment converts unmethylated cytosines into uracils while leaving methylated cytosines untouched, so sequencing the treated DNA reveals the methylation state of every CpG within the amplified fragment. Because neighbouring CpGs in a CpG island tend to share the same methylation state, reading multiple sites per DNA molecule dramatically increases the discriminatory power of the test. The team amplified the HOXA3 and MAP4K1 regions using a dual-index, four-primer PCR strategy and sequenced the products on a benchtop instrument.

The validation strategy was elegantly simple. The researchers purified DNA from peripheral blood mononuclear cells, which represent the nucleated fraction of blood, and from cultured intestinal organoids, three-dimensional structures grown from rectal crypt stem cells that contain no blood or vascular contamination. Pure organoid DNA showed mean methylation of just 0.83 percent at HOXA3 and 1.02 percent at MAP4K1, while pure blood cell DNA showed 94.0 percent and 93.1 percent respectively. When the two DNA sources were mixed at seven precisely defined ratios and subjected to the assay, individual sequencing reads fell into two clean clusters with almost no intermediate methylation. A classification system that assigned reads as blood-derived or organoid-derived based on the number of methylated CpGs per fragment produced correlations with the input DNA ratios of R-squared 0.999 for HOXA3 and 0.9985 for MAP4K1, demonstrating near-perfect quantitative accuracy.

The team then moved to a genuinely heterogeneous, uncultured human sample: saliva. Saliva is attractive for epigenetic epidemiology because collection is non-invasive, but it contains an unpredictable mixture of leukocytes secreted from the oral gingiva and buccal epithelial cells sloughed from the cheek. Using flow cytometry and cellular filtration, the researchers purified buccal cells and salivary leukocytes to purities of 97.1 percent and 99.4 percent respectively, and confirmed that the two populations carried the expected opposite methylation patterns. In a mixed saliva sample that microscopic cell counting showed to be 38 percent leukocytes, the methylation-based test predicted leukocyte proportions of 38.1 percent using HOXA3 and 39.3 percent using MAP4K1, with linear regression R-squared values of 0.998 and 0.997. For researchers who use saliva in large epidemiological studies, this offers a way to correct for cellular composition that previously required complex bioinformatic deconvolution of whole-genome data.

Perhaps the most provocative part of the study is what the biomarkers reveal about published disease research. Because immune cells drive inflammation, genuine leukocyte-specific methylation markers should appear repeatedly in methylation studies of inflammatory diseases, even when the diseased tissue itself is not blood-derived. Searching the EWAS Atlas database, the team found 28 disease traits associated with their markers, with psoriasis, inflammatory bowel disease and Alzheimer disease topping the list. A deeper PubMed search of twelve publications found that seven of the eleven studies using the standard methylation array platform reported at least one differentially methylated position or region overlapping the leukocyte-specific sites. Strikingly, many of these studies never mentioned leukocyte infiltration as a possible confound, and only one investigated the proportion of leukocytes in its samples at all.

The Alzheimer disease connection is particularly striking. Several independent studies have reported that a 48-kilobase region spanning the HOXA gene cluster from HOXA2 to HOXA6 becomes progressively hypermethylated in the brain as disease severity increases, measured by Braak stage. When the Otago team overlaid that hypermethylation pattern with methylation data from purified leukocytes and cortical neurons, the genomic locations matched almost exactly: leukocytes are heavily methylated across the HOXA cluster, neurons are not. The authors hypothesise that the methylation signal in the Alzheimer brain comes from infiltrating leukocytes, a phenomenon known to occur as the ageing blood-brain barrier becomes leaky, rather than from any change in the methylation of the brain cells themselves. A simple calculation based on a single CpG site suggested that by Braak stage 6, an additional 13.9 percent of cells in the prefrontal cortex could be leukocyte-derived.

The biomarkers also performed well in cancer. Using more than 8,000 tumour samples from 30 cancer types in The Cancer Genome Atlas, the team compared methylation at the HOXA3 marker with published estimates of leukocyte content. Across all cancers combined the correlation was modest, because cancer cells frequently acquire their own hypermethylation at the HOXA cluster. But in six tumour types, including uveal melanoma, mesothelioma, thyroid carcinoma, testicular germ cell tumours, bladder urothelial carcinoma and ovarian serous cystadenocarcinoma, a single CpG site showed a strong linear relationship with the leukocyte estimate, meaning that in these cancers one methylation measurement may be enough to quantify immune infiltration without any complex deconvolution.

The implications reach well beyond a single laboratory technique. The study demonstrates that methylation differences attributed to diseased cells may sometimes simply reflect the immune cells living among them, and it offers a concrete remedy: researchers studying inflammatory diseases could sort pure cell populations, spike in known amounts of leukocyte DNA as controls, or compare their samples against reference methylation data from healthy tissues and blood. Looking forward, the authors suggest that because colonic epithelium is virtually unmethylated at both loci, a stool-based test for intestinal inflammation could one day complement or even replace invasive colonoscopy for diagnosing inflammatory bowel disease. What began as a search for molecular fingerprints of blood cells has ended up casting doubt on a whole class of disease methylation signals, and providing the tools to tell them apart.

Subject of Research: Leukocyte-specific DNA methylation biomarkers for detecting immune cell infiltration in tissue samples

Article Title: Leukocyte-specific DNA methylation biomarkers and their implication for pathological epigenetic analysis

Article References: Dunnet, M. J., Ortega-Recalde, O. J., Waters, S. A., Weeks, R. J., Morison, I. M., & Hore, T. A. (2022). Leukocyte-specific DNA methylation biomarkers and their implication for pathological epigenetic analysis. Epigenetics Communications, 2(1), Article 5. https://doi.org/10.1186/s43682-022-00011-z

Image Credits: AI Generated

DOI: 10.1186/s43682-022-00011-z

Keywords: DNA methylation, leukocytes, epigenetics, biomarkers, HOXA3, MAP4K1, inflammation, Alzheimer disease, inflammatory bowel disease, cancer immunology, bisulphite sequencing, saliva