The circular clues in MS blood: why women’s signals shine—and six RNAs stand out

If you watch the RNA world, you’ve seen circular RNAs (circRNAs) go from curiosity to serious biomarker contenders. A Human Molecular Genetics study by Iparraguirre et al. pushes that arc forward with a simple but gutsy question: do circRNAs in blood leukocytes systematically change in multiple sclerosis (MS)—and can any of them help diagnose disease? The short answer is yes, with a twist: the signal is strikingly female-skewed.

What they did—at a glance
Samples. Three cohorts from whole blood leukocytes: (i) discovery RNA-seq (30 MS; 20 healthy controls), (ii) RT-qPCR validation (70 MS; 46 controls), and (iii) relapse/remission pairs (20 MS; sampled in both states). All MS diagnoses were standard; ethics, consent, and biobanking are covered.

Library & mapping. rRNA-depleted, strand-specific RNA-seq (HiSeq X Ten, ~10 Gb/sample). circRNAs called with find_circ v1.0 and CIRI2, intersected for stringency (≥2 BSJ reads, both callers). Linear transcripts quantified with STAR→HTSeq. DE called with DESeq2.

Validation. Divergent-primer RT-qPCR across the BSJ, single-amplicon specificity verified by melt curves, gels, and Sanger sequencing. ROC analyses via DeLong’s approach.

TcircRNAs are globally up in MS blood—especially in women
From the RNA-seq discovery set, intersecting callers yielded 22,835 bona fide circRNAs. MS samples had more unique circRNAs (19,781) than controls (16,772), and a higher overall circRNA abundance (normalized counts, P < 0.0001). Differential analysis (|FC| > 1.5, P < 0.05) flagged 464 circRNAs, and here’s the jaw-dropper: 96.1% were upregulated in MS (446 up / 18 down).

The sex breakdown sharpened things:

Females: 498 up vs 14 down (97.3% upregulated).

Males: 193 up vs 73 down (72.5% upregulated) and effects were smaller.

Only 34 circRNAs overlapped between sexes—most of the signature is sex-specific.

Importantly, this global tilt did not show up in the linear transcriptome. Among 45,736 detected linear RNAs, 1,372 were DE and split roughly 50 / 50 up vs down. GO terms there screamed innate immunity (myeloid/neutrophil activation, degranulation), but parent genes of DE circRNAs did not enrich specific processes, arguing the circRNA shift is specific and system-wide, not a byproduct of the same immune programs visible in mRNAs.

Take-home: circRNA upregulation is robust, disease-associated, and heavily sex-dependent, whereas mRNA changes look like familiar immune activation.

Six circRNAs rose to the top—and they classify MS well
From 68 “high-confidence” DE circRNAs (BaseMean >10, present in >50% of samples), the team shortlisted seven candidates and validated six by RT-qPCR in the independent cohort. The validated set:

PADI4 (chr1:17,668,437–17,668,897)

ABCA13 (chr7:48,541,721–48,542,148)

AFF2 (chrX:147,743,428–147,744,289)

NEIL3 (chr4:178,274,461–178,274,882)

AGFG1 (chr2:228,356,262–228,389,631)

ATP8B4 (chr15:50,294,349–50,311,173)

(RELL1 trended up but missed significance in validation.) In women, all six validated as up; in men, three did, with PADI4 and ABCA13 showing larger fold changes than in females.

Diagnostic performance (AUC): PADI4 0.843; AFF2 0.753; NEIL3 0.741; ABCA13 0.748; AGFG1 0.733; ATP8B4 0.722. Combined six-marker model: AUC 0.852—promising for a minimally invasive blood signature.

Does MS stage matter? (RR vs SP) Not much.
Comparing RR-MS (n=20) vs SP-MS (n=10) in discovery RNA-seq found 121 DE circRNAs—just 0.61% of circRNAs detected in patients—implying that the global circRNA signature is largely shared across subtypes. Only seven of those were “high confidence.”

What about relapse vs remission?
In paired blood from 20 patients sampled in both states, 14/20 showed notable within-person shifts (|FC|>1.5 on ≥half the candidates), but direction split: 7 trending up, 7 down during relapse. Subtle clinical patterns (not statistically significant) hinted that the “up in relapse” group had lower EDSS and longer disease duration, and was mostly male. Mechanistically, this points to state-dependent and patient-specific dynamics rather than a single relapse “switch.”

Bioinformatics angles worth noting
Caller intersection matters. The team combined find_circ and CIRI2 and required ≥2 BSJ reads, a conservative strategy that still yielded 22,835 circRNAs. It’s a nice template for reducing false positives in biomarker screens.

Sex as a covariate. DESeq2 models included sex when comparing MS vs controls, which likely helped isolate the global disease effect before diving into sex-stratified analyses.

Biology ≠ biogenesis artifact. Known RBP regulators of circularization (e.g., ADAR1, DHX9, QKI, FUS, HNRNPL) weren’t differentially expressed at the RNA level, arguing the circRNA surge isn’t simply altered biogenesis via these factors. Cell-composition effects were explored indirectly and pointed toward higher granulocyte signatures in MS, but the authors conclude these don’t explain the circRNA-specific global increase.

Why this matters
Clinically feasible matrix. Whole-blood leukocytes are a practical source; circRNAs are stable, making them attractive for real-world testing (shipping delays, freeze–thaws).

Sex-aware precision. The female-skewed signature dovetails with the epidemiology of MS and with prior sex effects in blood transcriptomes—critical for any diagnostic aiming for high PPV across populations.

Orthogonal to mRNA readouts. Because the circRNA shift is global and specific, a circRNA panel could complement mRNA immune-activity scores rather than duplicate them.

A few caveats
Multiple testing & thresholds. DE was defined at P < 0.05 with |FC|>1.5; the paper doesn’t emphasize adjusted q-values for circRNAs. Replication across centers and platforms would help lock down the global-up trend.

Cell heterogeneity. Bulk leukocytes blur cell-type signals. Cell-sorted or single-cell circRNA profiling (or deconvolution) could pinpoint whether the sex difference reflects composition vs per-cell regulation.

Mechanism. The authors speculate that widespread circRNA changes might intersect innate immune sensors (e.g., PKR-related dsRNA recognition) as suggested by other work; that’s testable in primary immune cells.

Clinical fit. AUC 0.852 is encouraging, but real deployment needs prospective studies, confounder checks (infection, steroids, comorbid autoimmunity), and inter-lab assay harmonization.

Blood leukocyte circRNAs show a global, disease-specific, and sex-dependent upregulation in MS. Six validated circRNAs—PADI4, ABCA13, AFF2, NEIL3, AGFG1, ATP8B4—classify MS vs controls well (combined AUC 0.852), with stronger signals in women and patient-specific shifts across relapse/remission. It’s one of the clearest cases yet that circRNAs carry orthogonal information to mRNAs in autoimmune disease—and they may be ready for serious biomarker development.

Disclaimer: This blog post is based on the provided research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.

References:
Iparraguirre, L., Alberro, A., Sepúlveda, L., Osorio-Querejeta, I., Moles, L., Castillo-Triviño, T., ... & Otaegui, D. (2020). RNA-Seq profiling of leukocytes reveals a sex-dependent global circular RNA upregulation in multiple sclerosis and 6 candidate biomarkers. Human Molecular Genetics, 29(20), 3361-3372.