Spatial transcriptomics of human placentas reveal distinct RNA patterns associated with morphology and preeclampsia

Spatial transcriptomics (ST) maps RNA level patterns within a tissue. This technology has not been previously applied to human placental tissue. We demonstrate analysis of human placental samples with ST. Unsupervised clustering revealed that distinct RNA patterns were found corresponding to different morphological structures. Additionally, when focusing upon terminal villi and hemoglobin associated structures, RNA levels differed be-tween placentas from full term healthy pregnancies and those complicated by preeclampsia. The results from this study can provide a benchmark for future ST studies in placenta.


Introduction
Preeclampsia is one of the great obstetrical syndromes, affecting placental morphology and transcriptome [1][2][3][4][5].Since some effects on the placenta are localized, bulk RNA sequencing may not reflect important alterations in distinct cell types or anatomical structures.
Characterization of individual cells became possible with single-cell RNA sequencing (scRNA-seq).This technology has revealed a more nuanced view and new insights on the diversity of those cells present in placenta [6][7][8][9].Limitations of scRNA-seq include potentially important information regarding spatial localization and interaction being lost.Also, cells might be affected or lost by the single-cell preparation protocol.
Spatial transcriptomics (ST) combines histology with RNA sequencing [10].Although resolution is not on the single-cell level, ST allows for detection of differences in RNA levels associated with morphological features and gain insight into spatial organization of tissues.Investigating RNA levels together with spatial information presents an excellent opportunity for further understanding of placental physiology, especially for preeclampsia where impact on the tissue is heterogenous.The aim of this study was application of the ST method to obtain a spatially resolved molecular characterization of human placenta, investigation of transcriptional profiles for differing morphological structures in preeclampsia and normotensive controls.

Placental samples
Placenta tissue from cases of term preeclampsia and normotensive controls were included in the study (n = 6; Table S1).Preeclampsia was defined as blood pressure ≥140/90 mmHg and proteinuria ≥300 mg/L [11].Placental tissue collection with informed, written consent from participants was approved by the Ethics Committee Review Board for the studies on human subjects at Lund University and Skåne University Hospital (LU 803-02).

Spatial transcriptomics (ST)
All tissue samples had an RNA integrity number (RIN) ≥7.The ST analysis was performed as previously described [10] with 12 μm thick sections and permeabilization time of 5 min optimized for placental tissue.Detailed descriptions of the methods are available in the supplementary material.Two replicate sections from each biopsy were analyzed.An overview of the study is given in Fig. 1a, and the analysis of the spatial transcriptomics dataset is outlined in Fig. S7.

Results and discussion
In this study, all women were primiparous, non-smokers, and delivered vaginally at term (≥37 weeks of gestation).The placental tissue sections placed on ST slides resulted in sequencing data from 12 sections (6 placental biopsies), encompassing sequencing data from a total of 11830 spots, detecting on average 2090 unique genes, and 4331 unique molecular identifiers (UMIs) per spot (Fig. S1).There were no major discrepancies between the technical replicate tissue sections (Figs.S3-S4), and the ST data correlated well with corresponding bulk RNA-seq data (Fig. S2).Unsupervised clustering divided the spots' RNA profiles into five different clusters (Fig. 1b and c).These clusters roughly corresponded to different morphological structures present in the sections: terminal villi (cluster 0), larger villi (cluster 3) and fibrin deposits (cluster 4).We also detected two unspecific clusters (cluster 1 and 2).Cluster 1 had a high expression of hemoglobin genes spread across the tissue sections.Cluster 2 morphologically mostly consisting of terminal villi but with a predominant expression of pseudogenes.The top marker genes from each cluster were mostly expected given the morphological features they correspond to (Fig. 1c).For instance, cluster 0 consisting of mostly terminal villi had a dominant gene profile expressing hormonal genes (Table S2).This was expected since hormone producing syncytiotrophoblasts constitute a large portion of the terminal villi [12], strengthening the validity of the obtained data.The results from the unsupervised clustering were verified by performing an independent non-negative matrix factor (NMF) analysis (Fig. S5, Table S3).
To compare case and control, we pooled all the spots from all the samples and conducted a pseudo-bulk differential gene expression analysis using DESeq2 [13] which revealed a clear distinction in RNA levels between cases and controls (Fig. 2a) (Table S4).Genes downregulated in preeclampsia compared to control had a higher distribution in clusters 0 and 1 among controls (Fig. 2b).Genes upregulated in preeclampsia compared to control had a higher distribution in cluster 4 (Fig. 2b).With a continued focus on only the clusters common amongst all samples i.e. cluster 0 (terminal villi) and cluster 1 (hemoglobin associated regions), we performed another pseudo-bulk differential gene expression analysis on spots under cluster 0 and cluster 1 using DESeq2 (Table S5) in order to highlight the differences between case and control that are more relevant to the common clusters.Consistent with previously reported results we found lower RNA levels of AQP3 [14] and elevated levels of IGFBP1 [15] with preeclampsia (Fig. 2c and d).However, we did not notice a significant difference in the expression of LEP and FLT1, two genes commonly reported as dysregulated in preeclampsia [5,16], between case and control in our dataset.With pathway analysis, genes associated with oxidative stress in the oxidoreductase pathway [GO:0016684], which include the hemoglobin genes, were elevated in preeclampsia (Fig. 2e and f).This is consistent with previous findings where preeclampsia was associated with increased placental oxidative stress and induction of hemoglobin gene HBA1 expression in syncytiotrophoblasts [17].
To our knowledge, this is the first study using ST on human placental tissue.We have successfully applied the ST protocol to placental samples.Through this, the sensitivity of the ST method on placental tissues has been determined and can now be applied on larger clinical cohorts in future studies.The unsupervised clustering results proved that ST can identify and distinguishing between different morphological structures.Therefore, ST can be used to study RNA patterns and pathway enrichment scores of varied morphologies on the same tissue sections.As terminal villi were the dominant structure in all the tissue sections, the Duplicate sections from each placenta selected based on the spot coverage by the tissue sections and morphology were placed on ST slides with a spot diameter of 100 μm and 100 μm distance between spots.After staining with hematoxylin and eosin, images were captioned for each of the sections.Thereafter, the tissue was permeabilized with the enzyme pepsin allowing RNA to be released from the tissue and captured onto probes attached to the ST slide.After library preparation, captured mRNA was sequenced on the Illumina NextSeq 500/550 platform.b.Manual annotation of two morphological regions (left) and classification by the unsupervised clustering (right) corresponded well especially for the terminal and larger villi structures.Five clusters (cluster 0-4) were identified and visualized using Uniform Manifold Approximation and Projection (UMAP).c.Top marker genes across the five identifiable clusters are presented in a heat-map with the top-five characterizing genes for each of the clusters generated by the unsupervised clustering.
analysis focused primarily on this feature.
Several of the findings in the study are confirmed by previously published works describing the impact of preeclampsia on placental tissue [14,15,17], thereby validating our findings.Although no absolute claims can be made about the results comparing preeclampsia and controls, this study can provide a starting point for future exploratory studies on the relationship between dysregulated RNA profiles and morphological structures in preeclampsia.An advantage of using ST over bulk RNA-seq is that the RNA patterns can be directly correlated with morphological structures within the tissue sections, allowing additional insights into the origin of the disorder.
Owing to the heterogeneous nature of the placenta, future work in the field would include exploration of placental histopathology via ST.Further research of the mechanism of preeclampsia can be performed via including a plethora of samples from different subclasses.A continuous effort towards sharing multimodal datasets will aid the community by the increasing collective understanding of complex placental diseases.
In conclusion, the ST technology can be readily used on human placental samples.With this technology, we demonstrate that predominant RNA transcripts differ in morphological structures and are impacted in preeclampsia.S1).e. Results from pathway analysis from cluster 0 (terminal villi), hemoglobin associated pathways in cluster 0, and cluster 1, where the heatmap corresponds to the adj.p-value and the dot size represents the number of genes in the gene set of the pathway.f.The oxidoreductase pathway, which is linked to oxidative stress, comprised of hemoglobin genes and present in both cluster 0 and 1, compared spatially and up-regulated in PE (section id PE_1_2 and Control_3_2).

Fig. 1 .
Fig. 1.Overview of the study and transcriptional profiles associated with placental morphological structures.a. Placental tissue from six full term pregnancies (control and preeclampsia) were cryosectioned (12 μm thick sections) for ST analysis.

Fig. 2 .
Fig. 2. Differentially expressed genes in preeclampsia (PE) and controls.a.The most significant differentially expressed genes in PE versus controls when applying a pseudo-bulk analysis approach to the dataset with the top 10 differentially expressed genes (adjusted p-value <0.0001 and log2FoldChange>0.5)labelled.b.General distribution of the most significantly up-regulated and down-regulated genes in the five different clusters.c.The top 10 down-regulated and up-regulated genes in PE (n = 3) versus control (n = 3) with distribution in cluster 0 and cluster 1 per placenta sample.The datapoints represents the average expression of the gene in cluster 0 and cluster 1 in samples from each condition.d.Plotting one up-regulated (IGFBP1) and one down-regulated (AQP3) gene of interest spatially and comparing distribution between PE (section id PE_2_2) and control (section id Control_3_1) (see TableS1).e. Results from pathway analysis from cluster 0 (terminal villi), hemoglobin associated pathways in cluster 0, and cluster 1, where the heatmap corresponds to the adj.p-value and the dot size represents the number of genes in the gene set of the pathway.f.The oxidoreductase pathway, which is linked to oxidative stress, comprised of hemoglobin genes and present in both cluster 0 and 1, compared spatially and up-regulated in PE (section id PE_1_2 and Control_3_2).