Small RNAs in plasma extracellular vesicles define biomarkers of premanifest changes in Huntington's disease

Abstract

Despite the advances in the understanding of Huntington's disease (HD), there is a need for molecular biomarkers to categorize mutation carriers during the preclinical stage of the disease preceding functional decline. Small RNAs (sRNAs) are a promising source of biomarkers since their expression levels are highly sensitive to pathobiological processes. Here, using an optimized method for plasma extracellular vesicles (EVs) purification and an exhaustive analysis pipeline of sRNA sequencing data, we show that EV-sRNAs are downregulated early in mutation carriers and that this deregulation is associated with premanifest cognitive performance. Seven candidate sRNAs (tRF-Glu-CTC, tRF-Gly-GCC, miR-451a, miR-21-5p, miR-26a-5p, miR-27a-3p and let7a-5p) were validated in additional subjects, showing a significant diagnostic accuracy at premanifest stages. Of these, miR-21-5p was significantly decreased over time in a longitudinal study; and miR-21-5p and miR-26a-5p levels correlated with cognitive changes in the premanifest cohort. In summary, the present results suggest that deregulated plasma EV-sRNAs define an early biosignature in mutation carriers with specific species highlighting the progression and cognitive changes occurring at the premanifest stage.

Keywords: Huntington's disease; biomarker; extracellular vesicles; miRNA; premanifest; small RNA; tRF.

FIGURE 1

sRNA biotypes are differently distributed between plasma subfractions. (a) Schematic representation of plasma‐EV isolation procedure by SEC and UF and selected fractions for analysis: cEVs, Filtrate, NonEVs‐Low, and NonEVs‐High fractions. (b) Representative SEC profile of plasma samples showing EV elution in low‐protein fractions as measured by Bradford protein assay and enriched in CD9, CD63, and CD81 markers as analyzed by flow cytometry. (c) Representative NTA size distribution profile of cEVs fraction (n = 5). (d) Western Blot analysis of isolated fractions. Fractions were analyzed for the presence of EV markers (Alix, Flotillin1, TSG101, Syntenin and CD9) and negative EV‐markers (Calnexin and COX IV). SH‐SY5Y cell lysate (SH) was used as positive controls. (e) Representative images of cryo‐EM of the isolated fractions. Upper images, bar = 1 µm. Lower images, bar = 200 nm. Using SeqCluster tool: (f) Fraction of reads that align to small RNA types are shown per plasma subfraction. Mean ± SD is shown. (g) Volcano plot showing differentially expressed sRNAs in cEVs versus NonEVs‐High fractions. Pink dots represent DE miRNAs and green dots represent DE gene fragments. (|log2FoldChange| > 0.58, adj p < 0.05). Samples of each subfraction, n = 5.

FIGURE 2

Characterization of EV fraction and NonEV fraction isolated from P‐HD, M‐HD and CTL plasma. (a) Schematic representation of the workflow followed in both plasma subfractions. (b) Representative profiles of CTL and HD plasma SEC fractions by protein determination and EVs fractions eluted in low protein fractions were detected using EV‐markers CD9, CD63 and CD81 by flow cytometry analysis. (c) Fold change of MFI values (MFI values compared to the isotype control—IgG‐) for CD9, CD63, and CD81 in EVs pooled fractions (n = 8–10 per group). (d) Representative Western Blot analysis of EVs fractions for the presence of EV markers (Flotillin1, TSG101 and Alix) and negative EV‐markers (COX IV and Calnexin). SH‐SY5Y cell lysate (SH) was used as a positive control (n = 3 per group). (e) NTA size distribution profiles of EVs fractions (left) and quantification of EV particle concentration (right) (CTL vs. P‐HD, adj p = 0.984; CTL vs. M‐HD, adj p = 0.348; P‐HD vs. M‐HD, adj p = 0.432; n = 5 per group). (f) Representative images of cryo‐EM of P‐HD, M‐HD and CTL EVs. Bar = 200 nm (n = 5 per group). (g) Size distribution of EVs diameter by cryo‐EM (n = 5 per group). (h) Protein content of NonEVs fractions (n = 8–10 per group). sRNA profiles analyzed by SeqCluster tool showing the fraction of reads that align to small RNA types per plasma group (i) in EVs fractions and (j) NonEVs fractions (n = 10 per group). Coefficient of variance (CV) in EV fraction = 0.09; CV in NonEVs fraction = 1.1; Levene's test used to assess the equality of variances, between EVs and NonEVs (p‐value = 0.0027). All data are represented as Mean ± SD.

FIGURE 3

sRNAs from HD patients are deregulated in EVs fractions in comparison to CTL samples, using SeqCluster tool. (a) Volcano plots showing DE sRNAs‐clusters in P‐HD‐EVs versus CTL‐EVs fractions and M‐HD‐EVs versus CTL‐EVs fractions. Dark pink dots and dark blue dots represent significantly upregulated and downregulated genes, respectively (|log2FoldChange| > 0.58, adj p < 0.05, n = 10 per group). (b) Total number of sRNAs‐clusters DE between P‐HD‐EVs versus CTL‐EVs fractions and M‐HD‐EVs versus CTL‐EVs fractions. (c) Venn diagram of DE sRNAs‐clusters between P‐HD versus CTL and M‐HD versus CTL, showing the number of overlapped dysregulated sRNAs‐clusters between both comparisons. (d) PCA plots constructed with top DE sRNAs and top sRNA that contribute to discriminating between disease conditions and healthy individuals based on PLS‐DA models. The solid line is the linear discriminant function that separates disease conditions.

FIGURE 4

sRNAs deregulated in premanifest stages correlate with cognitive symptomatology. (a) The framework of Huntington's Disease stages. The ideal candidate molecular biomarker would be measurable prior to the appearance of P‐HD and M‐HD symptoms and would correlate with the disease course. The current patients’ classification system together with the Integrated staging system⁷⁴ is depicted. (b) Workflow followed to select relevant sRNAs for correlation analyses. PLSR analysis depicting the relation between the set of sRNAs specifically DE in P‐HD patients and the performance of the SWRT test (c) in HTT mutation carriers (P‐HD and M‐HD; R ² = 0.97; p‐value < 0.0001); (d) only in the P‐HD group (R ² = 0.93; p‐value = 0.035); and (e) only in the M‐HD group (R ² = 0.289; p‐value = 0.735).

FIGURE 5

Validation and diagnostic potential analysis of selected sRNAs at premanifest stages. (a) Boxplots representing the relative expression of sRNAs validated with qRT‐PCR in the group of samples used for sRNA sequencing (n = 10 per group) and in an added independent biological group of samples (n = 10–11 per group). Significant differences between P‐HD and M‐HD vs‐ CTL groups are presented with *** (adj p < 0.001), ** (adj p < 0.01) and nominally significant differences are presented with (p‐value < 0.05). (b) ROC curves analysis of the sensitivity and specificity of plasma NfL between P‐HD and CTL (p‐value = 0.001). (c) Representative ROC curve analysis of the sensitivity and specificity of an individual validated sRNA: tRF‐Glu‐CTC between P‐HD and CTL (p‐value = 0.002). (d) ROC curves analysis of the sensitivity and specificity of a novel 2‐sRNAs‐biosignature: an ensemble of tRF‐Gly‐GCC and miR‐27a‐3p between P‐HD and CTL (p‐value = 0.0001). (e) ROC curves analysis of the sensitivity and specificity of a novel 3‐sRNAs‐biosignature: the ensemble of tRF‐Glu‐CTC, tRF‐Gly‐GCC and miR‐27a‐3p between P‐HD and CTL (p‐value = 0.0001). n = 20‐21 per group.

FIGURE 6

Longitudinal changes validation of selected sRNAs at premanifest stages over a 1.5‐year follow‐up. Boxplots representing the relative expression of sRNAs validated by qRT‐PCR in P‐HD paired samples from baseline and 1.5‐year follow‐up visits. Squares represent mean values. Significant differences between longitudinal samples are presented with *** (adj p < 0.001), ** (adj p < 0.01), * (adj p < 0.05) and nominally significant differences are presented with (p‐value < 0.05). n = 9 samples per group.