Sensitive detection of highly fragmented cytomegalovirus nucleic acid in human cfDNA

Vikas Peddu, Benjamin T. Bradley, Raj Shree, Brice G. Colbert, Hong Xie, Tracy K. Santo, Meei‐Li Huang, Kwang‐Ting Cheng, Eric Q. Konnick, Stephen J. Salipante, Keith R. Jerome, Christina M. Lockwood, Alexander L. GreningerbioRxiv 2019 · DOI

TL;DR

This paper explores using cell-free DNA (cfDNA) sequencing from maternal blood to detect Cytomegalovirus (CMV) infection, offering a non-invasive alternative to current diagnostic methods. They found CMV cfDNA is highly fragmented and shorter than human cfDNA, which could explain variations in qPCR results and pave the way for improved screening.

What Problem Does This Paper Address?

Congenital Cytomegalovirus (CMV) infections are a major cause of lifelong disabilities in newborns, including hearing and central nervous system impairments. Currently, there's no routine prenatal screening for CMV in the US, and confirming suspected cases requires invasive procedures like amniocentesis. This delay and invasiveness hinder early intervention. The authors propose that detecting CMV DNA in maternal blood plasma (cell-free DNA or cfDNA) could offer a non-invasive way to screen for congenital CMV infection.

Methods

The researchers analyzed cfDNA sequencing data from 2,208 individuals undergoing routine prenatal aneuploidy screening. Their approach involved:

  • Sequencing and Alignment: cfDNA was sequenced, and reads were aligned to the human herpesvirus 5 Merlin strain reference genome to identify CMV.
  • Quantification: CMV detection was quantified using reads per million sample reads (RPM).
  • Thresholding: Samples were classified as "strong positive" (RPM > 0.3) or "intermediate positive" (0.01 < RPM < 0.3).
  • qPCR Validation: A subset of positive and negative samples were tested using quantitative PCR (qPCR) to confirm CMV detection and compare viral loads.
  • Fragment Length Analysis: They specifically analyzed the insert size (fragment length) of CMV cfDNA and compared it to human cfDNA. This involved deep sequencing of a single high-CMV sample to get robust fragment length data.
  • Bioinformatics Pipeline: A detailed bioinformatics pipeline was used for alignment, filtering, and quantification of CMV reads.

Datasets

The study utilized cfDNA sequencing data from 2,208 pregnant individuals who underwent clinically-indicated aneuploidy screening at the University of Washington. This dataset was collected between May 2017 and November 2018.

Key Insights and Ideas

  • CMV cfDNA is Highly Fragmented: A key finding is that CMV cfDNA fragments are significantly shorter (median 103 bp) than human cfDNA fragments (median 172 bp). This makes CMV cfDNA appear at the 3rd percentile of human cfDNA fragment sizes.
  • Fragment Length Explains qPCR Variability: The authors hypothesize that this high fragmentation of CMV cfDNA is a major reason for the variability observed in CMV viral load measurements across different qPCR assays. Assays designed for smaller amplicons might be more efficient at capturing these fragmented viral DNA pieces.
  • Non-Invasive Screening Potential: The study demonstrates that CMV can be detected in cfDNA from routine prenatal screening samples, suggesting that leveraging existing NIPT (non-invasive prenatal testing) infrastructure could provide a dual-purpose screening tool.
  • HHV-6 Comparison: The study also briefly explored the detection of Human Herpesvirus 6 (HHV-6) in cfDNA, finding that its fragment length was more similar to human cfDNA, highlighting differences in viral cfDNA characteristics.

Key Results

  • CMV Detection Rate: CMV was detected in 117 out of 2,208 samples (5.3%).
  • Classification: 10 samples were classified as "strong positive" (RPM > 0.3) and 107 as "intermediate positive" (0.01 < RPM < 0.3).
  • qPCR Confirmation: qPCR confirmed CMV in 9 out of 10 strong positive samples and 2 out of 32 intermediate positive samples.
  • Correlation with qPCR: A weak positive correlation (R² = 0.42) was observed between CMV cfDNA RPM and viral load measured by qPCR.
  • Fragment Length Difference: The median CMV cfDNA fragment length (103 bp) was significantly shorter than human cfDNA (172 bp) (p < 4.1e-102).

Limitations and Future Work

  • Sample Size for Fragment Length: The estimation of CMV cfDNA fragment length was based on a single specimen with the highest CMV amount. This might bias the estimation for the general population. More cfDNA sequencing in other clinical contexts (e.g., stem-cell or solid organ transplant) is needed.
  • Clinical Outcome Data: Clinical outcome data was only available for a small number of patients (five mothers and five offspring), making it difficult to draw conclusions about the clinical significance of positive CMV cfDNA results.
  • Maternal Data Gaps: Maternal data was missing for several strongly positive CMV cfDNA samples because mothers received care at outside institutions.
  • Pediatric Hospital Separation: The separation of the hospital system from the major pediatric hospital limited access to significant clinical history or post-natal outcomes for some positive cases.
  • Future Directions: Future studies need to examine more samples to generalize the findings and explore whether a specific level of CMV cfDNA can predict congenital infection. Further research is also needed to understand the effect of cfDNA in vivo on CMV DNA quantitation.

Summary updated

Abstract

Abstract Congenital human cytomegalovirus (CMV) infections are the leading cause of newborn hearing and central nervous system impairments worldwide. Currently, routine prenatal screening for congenital CMV is not performed in the United States and confirmation in suspected perinatal cases requires invasive sampling by amniocentesis. We hypothesized that detection of CMV from maternal cell-free DNA (cfDNA) plasma could provide a non-invasive indicator of congenital CMV infection. We analyzed sequence data from 2,208 individuals undergoing routine non-invasive prenatal aneuploidy screening at the University of Washington. CMV reads were identified in 117 (5.3%) samples. Positive samples were stratified based on CMV reads per million sample reads (RPM), resulting in ten samples being classified as strong positive (RPM &gt; 0.3) and 107 as intermediate positive (0.01&lt;RPM&lt;0.3). Subsequent qPCR testing identified CMV in 9/10 strong positive samples and 2/32 intermediate positive samples. Median cfDNA insert size derived from CMV was significantly shorter than cfDNA derived from human chromosomes (103 vs 172 bp, p&lt;0.0001), corresponding to the 3 rd percentile of human cfDNA insert size. In addition, CMV cfDNA fragment lengths were distributed over a wider range than human cfDNA reads. These studies reveal the highly fragmented nature of CMV cfDNA and offer precise measurements of its length: these features likely explain discrepancies in serum CMV viral loads measurements determined by different qPCR assays, despite widespread efforts to standardize results. More work is required to determine how detection of CMV from maternal cfDNA can be best used as tool for congenital CMV screening or diagnosis.
BibTeX
@misc{peddu2019sensitive,
  author = {Vikas Peddu and Benjamin T. Bradley and Raj Shree and Brice G. Colbert and Hong Xie and Tracy K. Santo and Meei‐Li Huang and Kwang‐Ting Cheng and Eric Q. Konnick and Stephen J. Salipante and Keith R. Jerome and Christina M. Lockwood and Alexander L. Greninger},
  title = {{Sensitive detection of highly fragmented cytomegalovirus nucleic acid in human cfDNA}},
  year = {2019},
  doi = {10.1101/748434},
  howpublished = {bioRxiv},
}

Loading related content...

Figure