key: cord-0297470-4ejkk2a1
authors: Ricci, C. A.; Reid, D. M.; Sun, J.; Santillan, D. A.; Santillan, M. K.; Goulopoulou, S.; Phillips, N. R.
title: Mitochondria-mediated Maternal-fetal Interactions and Consequences of Mitochondrial Dysregulation Indicate New Roles for Mitochondria in Hypertensive Pregnancies
date: 2021-12-21
journal: nan
DOI: 10.1101/2021.12.18.21268029
sha: 466da48559707104a4c0cd80532686e701850510
doc_id: 297470
cord_uid: 4ejkk2a1

Oxidative stress, placental mitochondrial morphological alterations, and impaired bioenergetics are associated with hypertensive disorders of pregnancy. Here we examined mitochondrial DNA mutational load in pregnant women with pregnancy-induced hypertension and reanalyzed publicly available high-throughput transcriptomic datasets from maternal and fetal tissues from normotensive and hypertensive pregnancies. Mitochondrial dysregulation was indicated by aberrant mitochondrial gene expression, and putative consequences were examined. Women with hypertensive pregnancy had elevated mitochondrial DNA mutational load. Maternal mitochondrial dysregulation in hypertensive pregnancies was associated with pathways involved in inflammation, cell death/survival, and placental development. In fetal tissues from hypertensive pregnancies, mitochondrial dysregulation was associated with increased extracellular vesicle production. Our study demonstrates mitochondria-mediated maternal-fetal interactions during healthy pregnancy and maternal mitochondrial dysregulation in hypertensive pregnancy development.

Oxidative stress, placental mitochondrial morphological alterations, and impaired 19 bioenergetics are associated with hypertensive disorders of pregnancy. Here we 20 examined mitochondrial DNA mutational load in pregnant women with pregnancy-21 induced hypertension and reanalyzed publicly available high-throughput transcriptomic 22 datasets from maternal and fetal tissues from normotensive and hypertensive 23 pregnancies. Mitochondrial dysregulation was indicated by aberrant mitochondrial gene 24 expression, and putative consequences were examined. Women with hypertensive 25 pregnancy had elevated mitochondrial DNA mutational load. Maternal mitochondrial 26 dysregulation in hypertensive pregnancies was associated with pathways involved in 27 inflammation, cell death/survival, and placental development. In fetal tissues from 28 hypertensive pregnancies, mitochondrial dysregulation was associated with increased 29 extracellular vesicle production. Our study demonstrates mitochondria-mediated 30 maternal-fetal interactions during healthy pregnancy and maternal mitochondrial 31 dysregulation in hypertensive pregnancy development. 32

Impairments in mitochondria fusion/fission dynamics and oxidative 34 phosphorylation [2, 3] are implicated in new-onset hypertensive disorders of pregnancy 35 like gestational hypertension and preeclampsia. However, consequences of 36 mitochondrial dysfunction in hypertensive pregnancies are not fully described, nor is the 37 role of maternal versus fetal mitochondrial dysfunction clearly delineated. In support of 38 the hypothesis that mitochondrial dysfunction underlies pregnancy-induced 39 hypertension, we recently demonstrated impaired circulating cell-free mitochondrial 40 DNA (ccf-mtDNA) dynamics in pregnant mothers with preeclampsia [4] . This finding is 41 highly relevant to preeclampsia pathophysiology, as ccf-mtDNA is often used as a non-42 invasive proxy for cellular stress and systemic inflammation due to its ability to activate 43

Toll-like receptor 9 and other DNA-sensing receptors [5, 6] . Here we examined ccf-44 mtDNA integrity in maternal samples from normotensive and new-onset hypertensive 45 pregnancies. We additionally analyzed publicly available maternal and fetal high-46 throughput transcriptomic datasets from normotensive and new-onset hypertensive 47 encoded genes are limited (Complex I/III/IV/V subunits and rRNAs/tRNAs for 78 mitochondrial translation [9] [10] [11] ) and are less regulated than nuclear-encoded 79 mitochondrial genes [9, 12, 13] . Additionally, mitochondrial fidelity is largely driven by 80 nuclear-encoded genes (e.g., master regulators of mitochondrial biogenesis [10] and 81 dynamics [10, 11] ). Therefore, compensation may easily occur for ccf-mtDNA mutations. 82

Mitochondrial dysregulation was present in hypertensive maternal and fetal tissues, 83

where 30 mtDEGs were detected in maternal blood ( Figure 1B) , and 9 were detected in 84 fetal placenta ( Figure 1C ). Only 2 mtDEGs were shared between maternal gestational 85 ages: MRPL38 (1 st trimester and at delivery) and BCL2L1 (3 rd trimester and at delivery). 86

Paired with increased ccf-mtDNA mutational load, this finding suggests generalized 87 mitochondrial dysregulation in hypertensive mothers (versus a specific point of 88 breakdown) and may not be specific to pregnancy. Though fetal gene expression was 89 Figure 1 Maternal mitochondrial DNA mutational load and mtDEGs in maternal and fetal tissues. A) maternal ccf-mtDNA mutational load; "HT": hypertensive pregnancies (n = 10); "NT": normotensive pregnancies (n = 13); "#": trending on significance (upaired t-test, P-value = 0.08); medium effect size observed (Cohen's d = 0.60), indicating biological significance. B) maternal mtDEGs in blood plasma throughout gestation and at delivery (no mtDEGs were detected in 2 nd trimester). C) fetal mtDEGs in placenta at delivery; P-adjusted values of significance and expression directions provided in the original publication were converted to z-scores; error bars not present because standard error was not reported in original publication [1] Figure 2 Maternal mitochondrial interaction gene expression through gestation. A) mitochondrial interaction gene sets significantly affected by time and gestational age; subscript for COX7B2 indicate 2 separate but significant gene expression trends were detected. B) hierarchical clustering of differentially regulated pathways in MRPL38 and FKBP8 interaction genes; teal circles represent groupings of cell death/survival pathways and immune system pathways.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint Likewise, the 9 mtDEGs detected in fetal placentas were associated with only 13 158 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. or that mitochondria-mediated hypertensive processes are alleviated. However, fetal 166 mitochondrial interaction genes were enriched for 21 cellular components that were 167 Figure 3 Expression patterns and functional enrichment analysis of maternal and fetal mitochondrial interaction genes at delivery. A) biological process enrichments for hypertensive maternal and fetal tissues at delivery. B) cellular component enrichments for hypertensive maternal and fetal tissues at delivery. Terms condensed for plotting purposes if genes comprising the term shared ≥ 99% overlap. "Count" represents number of genes comprising term.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. and fetal tissues is likely an artifact of sample tissue origin, it may also reflect the 173 fundamentally different biological priorities at delivery between mothers and neonates. 174

The prominent signal for EV production in hypertensive fetal placentas is notable. 

We provide evidence that mitochondrial dysregulation contributes to pregnancy-induced 188 hypertension. We demonstrate increased ccf-mtDNA mutational load associated with 189 hypertensive pregnancy, aberrant expression of mitochondrial genes, and aberrant 190 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 21, 2021. expression of mitochondrial interaction genes, through pregnancy and at delivery. We 191 connect pregnancy-specific mitochondrial dysregulation with established preeclampsia-192 associated processes (VEGF and PPAR signaling) and inflammation. We also show 193 fetal placental mitochondrial dysregulation at delivery is associated with increased EV 194 production. Though this may be specific to delivery, we hypothesize that increased 195 placental EV production observed in other studies during preeclamptic pregnancy may 196 also be mitochondria-mediated. Following our findings, we propose a two-pronged In mothers with pregnancy-induced hypertension, mitochondrial dysregulation leads to angiogenic imbalance and differential regulation of cell death/survival pathways and inflammatory/immune response pathways, in addition to dysregulation of VEGF (increased in early gestation) and PPAR (increased in late gestation) signaling. In the developing fetus, inheritance of mitochondrial dysregulation affects maternal-fetal communication through increased extracellular vesicle (EV) release from the placenta.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. UL1TR002537-S1). The content is solely the responsibility of the authors and does not 214 necessarily represent the official views of funders. 215 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10. 1101 /2021 used to index the reads followed by somatic variant calling via GATK4 Mutect2 utilizing 239 the mitochondrial mode and excluding read orientation base qualities below 30 [34, 35] . 240

Variants confirmed in both read directions were tallied. The data were tested and 241 confirmed for normality ( Supplementary Figure 1) , and an unpaired t-test (one-sided) 242 was used to test the hypothesis that mothers with PE have elevated mtDNA mutational 243 load. T-test and graphs were generated using Prism v.9. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10. 1101 /2021 and body mass index (BMI) were separately correlated (Supplementary Figure 1) . 285

Correlated fetal and placental metrics were subsequently collapsed into a summary 286 statistic by conducting principal component analysis (PCA) and taking the eigenvector 287 associated for PC1 (>60% variance explained, data not shown) for each patient, and 288 this process was also carried out on BMI and maternal age for each patient (>60% 289 variance explained, data not shown). Covariates determined as significantly stratifying 290 data were identified by conducting PCA on raw counts for each timepoint and using 291 general linear models to determine covariate effects (main effects and interaction with 292 preeclampsia diagnosis) on PC1 loadings. P-values ≤ 0.1 were used to allow for greater 293 sensitivity in DESeq2 models (Supplementary Table S1 ). DEGs for each contrast were 294 considered those with a P-adjusted value ≤ 0.05 (Supplementary File 1). An in-house 295

DESeq analysis could not be conducted for fetal placental samples as patient 296 characteristics were not provided in the original publication. We therefore relied on the 297

identify fetal DEGs. In these data, only P-adjusted value of significance and direction 299 was provided. Therefore, to understand fetal mtDEG dynamics, the P-adjusted values 300

for DEG list genes identified as mtDEGs were converted to Z-scores using one-tailed 301 standard P-value distributions. We reported the lower tail for down-regulated gene 302 expression and the upper tail for upregulated gene expression. Standard error was not 303 provided by authors. 304

The MitoCarta 3.0 database[43], a database of all verified mitochondrial genes 305 encoded by both the mitochondrial and nuclear genomes, was used to pull out mtDEGs 306 by identifying mitochondrial gene matches to the MitoCarta database in all DEGs for 307 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10.1101/2021.12.18.21268029 doi: medRxiv preprint both maternal and fetal samples (Supplementary Table S2 ). Once mtDEGs were 308 identified, their respective interaction genes present in DEGs were found using String 309 databases and an interaction confidence of "medium" or better (interaction score ≥ 310 0.400) [44] . For placental samples, interaction genes were pulled from the DEG list 311 provided using the same interaction criteria cutoff (≥ 0.400). All possible interaction 312 genes identified from String databases were used to conduct longitudinal gene set 313 analyses for collections prior to delivery (1 st trimester, 2 nd trimester, and 3 rd trimester, 314 details below). https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis). Delivery 322 collection data were excluded from this analysis due to the unique physiology of labor 323 and delivery. Transcripts counts were total count normalized[46] by counts per million. 324

Outliers were identified and removed via Euclidean distance. 325

Default settings for TcGSA were used. DESeq was used to normalize gene counts 326 across maternal gestation (covariates for DESeq model were chosen as described 327 above). A custom gene matrix transposed (GMT) file (Supplementary File 2) was made 328 using the interaction genes for each mtDEG identified during gestation, where gene sets 329 were comprised of the interaction genes identified in String DB for each specified 330 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10.1101/2021.12.18.21268029 doi: medRxiv preprint mtDEG. A null model was tested by making a second, size-matched GMT file 331 comprised of non-mitochondrial and non-differentially expressed genes to ensure that 332 significant gene sets identified in TcGSA were not an artifact of gene set size. 333

Significant gene sets were considered those meeting a P-adjusted value ≤ 0.05 after 334

TcGSA. Significant gene sets were probed for functional enrichment, details below. set and specifying Human as the sample species. All other parameters followed default 353 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 settings. Once the primary analysis was conducted, an additional comparison analysis 354 was performed in IPA to determine pathway enrichment changes through time. 355

Due to the limited number of interaction genes present during delivery for both 356 maternal and fetal tissues, IPA was not feasible. Therefore, to understand the functional 357 consequences of aberrant mitochondrial gene expression in maternal and fetal tissues 358 at delivery, GO enrichment analyses were conducted for the interaction genes using 359 default settings on the Gene Ontology web server. Biological process and cellular 360 component GO enrichments were performed and full terms list can be found in 361

For plotting purposes, GO terms were matched to individual mtDEG interaction 363 genes, and corresponding P-adjusted values for GO term enrichments were converted 364 to Z-scores as described above, using the log-fold change of the mtDEG interaction 365 gene to inform the direction of the Z-score (Supplementary File 4) . Terms were then 366 condensed using the "reduce_overlap" function in the GO Plot package[47] available for 367 R, specifying a 99% overlap as the criteria for term collapsing. 368 369 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 Table S1 . Data structuring for maternal DESeq2 contrasts. PC1 eigen values for each collection were investigated for significant stratification by covariates using linear models. Covariates with a significant effect on PC1 eigen value stratification were included in DESeq model for respective collection contrasts (columns). P-value ≤ 0.1 used to allow for greater sensitivity in DESeq model. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 21, 2021. ; https://doi.org/10.1101/2021.12.18.21268029 doi: medRxiv preprint Table S4 . References supporting classification of MRPL38 and FKBP8 gene set pathway enrichments under the broader classifications of "Cell death/survival" and "Immune system".

Pathway

How does p53 induce apoptosis and how does this relate to p53-mediated tumour suppression?

Inter-and intracellular signaling in amyotrophic lateral sclerosis: role of p38 mitogen-activated protein kinase

Necroptosis: a regulated inflammatory mode of cell death

Sirtuin regulation of mitochondria: energy production, apoptosis, and signaling

Regulation of innate immune cell function by mTOR

Regulatory mechanisms in T cell receptor signalling

Aberrant T Cell Signaling and Subsets in Systemic Lupus Erythematosus

Significance of the IL-8 pathway for immunotherapy

Cytokine networks in neuroinflammation

IL-2 and IL-15 signaling complexes: different but the same'

Pathogenesis of COVID-19 from a cell biology perspective

Role of p38 MAPK in UVB-induced inflammatory responses in the skin of SKH-1 hairless mice

Integrin-Linked Kinase Expression in Myeloid Cells Promotes Inflammatory Signaling during Experimental Colitis

IL-17 Signaling: The Yin and the Yang

Osteoclast precursors, RANKL/RANK, and immunology

Altered B cell receptor signaling in human systemic lupus erythematosus

B cell receptor and Tolllike receptor signaling coordinate to control distinct B-1 responses to both self and the microbiota

Erythropoietin and its derivatives: from tissue protection to immune regulation

The role of JAK-STAT signaling pathway and its regulators in the fate of T helper cells

ERBB signaling attenuates proinflammatory activation of nonclassical monocytes

CNTF-STAT3-IL-6 Axis Mediates Neuroinflammatory Cascade across Schwann Cell-Neuron-Microglia

Liver fibrosis: mechanisms of immune-mediated liver injury

Uniting the neurodevelopmental and immunological hypotheses: Neuregulin 1 receptor ErbB and Toll-like receptor activation in firstepisode schizophrenia