Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Oct 3.
Published in final edited form as: J Mol Endocrinol. 2025 Aug 22;75(2):e250008. doi: 10.1530/JME-25-0008

TNF but not VEGF induces secretion of multiple chemokines and cytokines by uterine artery endothelial cells – potential implications for preeclampsia

L Clemente 1, C Zhou 3, K Chaiyakul 2, JH Adams 1, J Jacobson 1, JL Austin 1, DS Boeldt 1, IM Ong 1,2,**, IM Bird 1,**,+
PMCID: PMC12490415  NIHMSID: NIHMS2102818  PMID: 40762350

Abstract

While pregnancy is known to be an inflammatory condition, preeclampsia (PE) is a more extreme state associated with higher cytokines and/or altered growth factors. It is generally assumed these PE elevated factors come from stimulation of immune cells and/or hypoxic uterine tissue, but several studies have shown that endothelial cells may also be a source. The goal of this study was to determine to what extent TNF, a factor over produced by uteroplacental tissue in PE pregnancy, may influence uterine artery endothelial cells to contribute to these other PE specific factors in the maternal circulation. Herein we use multiple analytical methods to show that uterine artery endothelial cells from pregnant sheep (P-UAEC) on exposure to cytokines can secrete multiple cytokines and chemokines seen in PE women that may contribute to production of Th17 cells and attraction of these and other cells to the vessel surface. Furthermore, the factors not significantly increased by TNF include those known to be specifically secreted by proinflammatory T cells. This begs the question if endothelium itself is the initial primary orchestrator of chemokine and cytokine imbalance, acting directly and indirectly to promote the symptoms of impaired vasodilation and reduced uteroplacental blood flow. If so, future preventive therapies for PE should be targeted at endothelium as well as immune cells.

Keywords: TNF-alpha, pregnancy, inflammation, uterine artery, endothelial dysfunction, hypertension, preeclampsia

Introduction

In pregnancy, blood flow through the uterine artery increases dramatically. This is facilitated in large part by the uterine artery endothelial cells undergoing substantial physiologic adaptation to achieve dramatically increased vasodilation. While this adaptation begins within weeks of conception, it is during the second half of pregnancy that enhanced uterine artery vasodilation is key to fetal well-being (Sladek et al 1997). We have shown that uterine artery endothelial cells adaptation of function occurs through a higher Cx43 gap junction communication, so leading to more coordinated and sustained Ca2+ signaling across the monolayer (Yi et al 2010, Reviewed Bird et al 2013). Conversely, preeclampsia (PE) is a hypertensive condition of pregnancy associated with a higher degree of inflammation (Redman & Sargent 2003) and the failure of this adaptation is replicated by cytokine mediated closure of Cx43 channels (Bird et al 2013). Longer term exposure of P-UAEC to TNF can also cause selective breakdown of adherins junctions through secretion of metalloproteases. (Ampey et al 2021), and consistent with the additional proteinuria and edema commonly observed in PE (Myatt & Webster 2009).

With the recent advent of multiplex assays, PE has been further characterized as an excessive elevation of multiple chemokines and inflammatory cytokines. Indeed, PE and other hypertensive conditions (systemic and pulmonary hypertension) share the same inflammatory markers, together with a pronounced increase in Th1/Th17 dominance in the T cell population (Jafri & Ormiston 2017). It has generally been assumed these elevated cytokines are of immune or uteroplacental origin (Jafri & Ormiston 2017, Benyo et al 1997), but little attention has been paid to the possibility that endothelium is a source. Significant increases in cytokine and chemokine output can occur on exposure of multiple endothelial cell types to inflammatory cytokines like TNF, (Mai et al 2013). As such, the key question addressed by this study is if TNF treated P-UAEC can act as a significant source of the many chemokines and cytokines known to be further elevated in PE subjects. To that end we report that TNF does specifically drive changes in mRNA transcripts of multiple endocrine factors in general, and chemokines and cytokines in particular. We also show that many of the corresponding proteins are increased in culture media. We undertook this study in purified endothelial cells in vitro to assure that these altered transcripts and corresponding secreted factors could not be from other sources. Our findings are consistent with a more substantial role for TNF inducing endothelial secretion of chemokines (Liu et al 2023, Wang et al 2022) and cytokines (Szarka et al 2010, Ma et al 2019) elevated in PE subjects than previously assumed.

Methods

General Reagents 1:

General cell culture reagents were from Life Technologies (Carlsbad, CA), and other basic chemicals and reagents were from Sigma (St. Louis, MO) unless otherwise stated. Gentamycin, penicillin/streptomycin were from Gibco (Grand Island, NY). TNF and IL1B were from R&D Systems, Inc. (Minneapolis, MN). For P-UAEC, growth media was prepared from Minimum Essential Medium (MEM), Fetal Bovine Serum, Penicillin/Streptomycin, and Gentamycin. Otherwise, all other ‘treatment’ reagents were made up per manufacturer recommendations in sterile H2O, sterile PBS, or DMSO. Note: Final DMSO was < 0.1%, which has no measurable effects on protein expression/phosphorylation, calcium signaling, or barrier function (unpublished data).

P-UAEC Cell Culture and Treatment:

Pregnant sheep at 120 days gestation were terminated and uterine arteries collected in accordance with Institutional Animal Care and Use Committees (IACUC) protocol. Uterine artery endothelial cells (P-UAEC) were isolated and grown in culture as described to passage 3 (Bird et al 2000). Growth media consisted of minimum essential medium (MEM) containing 20% fetal bovine serum (FBS), 1% penicillin-streptomycin, and 4 ug/ml gentamycin. Cells were then stored in liquid nitrogen for future use. Purity was verified at 99% by morphology and AcLDL uptake. Cells from 4 sheep were then grown to near confluence, pooled, split 1:4 and refrozen at passage 4. For experimental use, pooled cells were cultured in a T75 flask to 90% confluence and passaged to 6 × 60mm dishes. Treatments were performed on cells at 90% confluence in serum free (SF) MEM, 0,1% BSA, 1% penicillin-streptomycin, and 4 ug/ml gentamycin. Serum withdrawal was performed for one hour, and treatments added: control - media alone, TPA (10nM), VEGF (10ng/ml), TNF (10ng/ml), PP2 (10uM) or U0126 (10uM) to a final volume of 2.5 ml. Treatments were for 12 hours for transcriptomics and 18 hours for proteomics/endocrine secretion analysis (18h was chosen to allow completion of protein synthesis following the transcriptional changes by 12h). Dishes were then placed on ice and processed accordingly. For transcriptomics, media was removed, and cells snap frozen for RNA extraction and analysis at the Biotechnology Center (below). For media protein analysis, media was phenol red free, and recovered conditioned media was centrifuged at 300g for five minutes to remove detached cells. The supernatant was removed, and spun again at 12000 g to remove debris. 2ml was sent for proteomic analysis at the University of Wisconsin – Madison Biotechnology Center. The remainder was placed in 100ul aliquots and snap-frozen for storage at −80°C until Multiplex analysis.

Bulk Transcriptome Analysis by Affymetrix array:

An initial full spectrum analysis of ovine mRNA was performed using the Ovine ST 1.1 Gene expression chip (Affymetrix). P-UAEC were grown in 2.5ml minimum essential medium(MEM), 20% fetal bovine serum (FBS), penicillin and gentamycin to 80% confluence on 60mm dishes (passage 4), and serum withdrawn for 1 hour. Cells were then treated for 12h with vehicle (Control), TPA (10nM), VEGF (10ng/ml), TNF (10ng/ml), PP2 alone (10uM) or U0126 alone (10uM). Cells were gently rinsed twice with serum free cold media containing: Minimum Essential Medium, 1% penicillin-streptomycin, and 4 ug/ml gentamycin, and media aspirated before dishes were flash frozen using liquid nitrogen for subsequent (within <7 days) RNA analysis. A Trizol based extraction of the adherent cell monolayer and column purification using the Direct-zol™ RNA MiniPrep kit (Zymo Research Corp) was used as described by the manufacturer. Recovered RNA was assessed for purity on an Agilent RNA PicoChip at 5 ng/ul. RNA with a RIN number of above 8.5 and a 260:280 ratio above 2.0 was used for probe generation using the Ambion WT Expression Kit (as described by the manufacturer) using 400 ng RNA in each case. cRNA probe generation yielded products in the range 0–1000 nt as expected with a RIN number of 2 or greater. The ovine ST 1.1 chipset was used. Separate chips were run for 4 independent experiments (each experiment run on different days), and combined data initially compared using Transcriptome Analysis Consul V3.0 (Affymetrix). Detected transcripts were verified at the level of corresponding protein detected previously using a variety of immune based methods including ICC and ‘Western’ based techniques, including BD Powerblot (Gifford et al 2003). Transcript data was then further analyzed whereby the 1,416,100 Oligo probe data were aggregated and normalized using robust multichip average algorithm (RMA) (Irizarry et al 2003). The final 24,595 genes were then log2 transformed. To more reliably identify significantly differentially expressed genes, fold changes and p-values were recalculated using significance analysis of microarrays (SAM) (Tusher et al 2001). SAM was run using mostly default parameters with “Two class unpaired” for resp.type, “array” for assay.type, “standard” for testStatistic, “slope” for time.summary.type, “standard” for regression.method, and 26 for random.seed. The Benjamini-Hochberg method was then applied to control false discovery rate (Benjamini & Hochberg, 1995). Finally, the significantly expressed genes were selected for the combination of at least 2 FC and FDR < 0.05. The pathway analysis was performed using PANTHER.db. All analyses were performed in R (v. 4.3.1) using oligo (v. 1.64.1) (Carvalho & Irizarry, 2010), pd.ovigene.1.0.st (v. 3.12.0) (Carvalho, 2015), and samr (v. 3.0) (Tibshirani et al 2018) for differential expression and GO Terms with each treatment.

Milliplex Assays of Endocrine secretions to media by P-UAEC:

The same experimental treatment of P-UAEC for Affymetrix analysis was repeated in serum free medium without BSA for 18h stimulation time. Media was collected and stored frozen at −80C for subsequent Milliplex assay and proteomics analysis. We used the MILLIPLEX MAP Human Cytokine/Chemokine magnetic bead panel (HCYTOMAG-60K) and MILLIPLEX MAP Human Cytokine/Chemokine magnetic bead panel III (HCYP3MAG-63K), followed by the Ovine Cytokine/Chemokine Panel 1 kit (Millipore Sigma, Burlington, MA) for our analysis of media from cells treated overnight with TPA (10nM), VEGF (10ng/ml), TNF (10ng/ml), PP2 alone (10uM) or U0126 alone (10uM). The standard curves and conditioned media from each treatment in each experiment were tested in duplicate in a 96 well plate. Analytes bound to beads, were recovered with a magnetic washer as per the manufacturer’s instructions. After incubation with secondary and detection antibodies, beads were analyzed with the Luminex 200 (Luminex, Austin, TX). Curves were generated using Luminex 200 software, and results reported as pg/ml (mean and SE from four or more replicate experiments). The combined human kits targeted many more analytes than the ovine kit. Where an analyte was successfully measured by multiple kits, the ‘ovine’ value was used.

Proteomics analysis of secretions to media by P-UAEC:

The serum free and BSA free conditioned media from P-UAEC treated overnight (18h) with TPA (10nM), VEGF (10ng/ml), TNF (10ng/ml), PP2 alone (10uM) or U0126 alone (10uM) was also sent to the University of Wisconsin – Madison Biotechnology Center for analysis. Media was de-salted and split into two aliquots. The first was used to determine protein concentration of the conditioned media. The second sample underwent tryptic enzyme digestion based on the concentration determined in the first. Digests were cleaned using OMIXC18SPE cartridges (Agilent, Palo Alto, CA), eluted, dried, and reconstituted in 10µl of 70/30/0.1% ACN/H2O/TFA, dried to completion in the speed-vac and finally reconstituted in 0.1% formic acid to 1µg/ul concentration. Peptides were analyzed by nanoLC-MS/MS using the Agilent 1100 nanoflow system (Agilent) connected to a hybrid linear ion trap-orbitrap mass spectrometer (LTQ-Orbitrap Elite™, Thermo Fisher Scientific) equipped with an EASY-Spray™ electrospray source. Chromatography of peptides prior to mass spectral analysis was accomplished using capillary emitter column (PepMap® C18, 3µM, 100Å, 150×0.075mm, Thermo Fisher Scientific) onto which 2µl of extracted peptides was automatically loaded. NanoHPLC system delivered solvents A: 0.1% (v/v) formic acid, and B: 99.9% (v/v) acetonitrile, 0.1% (v/v) formic acid at 0.50 µL/min to load the peptides (over a 30 minute period) and 0.3µl/min to elute peptides directly into the nano-electrospray with gradual gradient from 3% (v/v) B to 30% (v/v) B over 155 minutes and concluded with 10 minute fast gradient from 30% (v/v) B to 50% (v/v) B, followed by a 7 minute flush-out from 50–95% (v/v) B. As peptides eluted from the HPLC-column/electrospray source, survey MS scans were acquired with a resolution of 120,000 followed by MS2 fragmentation of 20 most intense peptides detected in the MS1 scan from 350 to 1800 m/z; redundancy was limited by dynamic exclusion. The proteomics data from ‘batch 1’ and ‘batch 2’ experiments were combined and preprocessed using the proBatch tool from Bioconductor (Cuklina 2018). The data were log2 transformed, and quantile normalized. Continuous drift correction was used to correct batch effects by adjusting the trend and correcting with ComBat. For ANOVA, the preprocessed data were z-scaled across features.

Results

Affymetrix data analysis was undertaken on RNA from P-UAEC treated with VEGF or TNF. We used TPA, a broad kinase activator (including ERK and Src), as a positive control. As opposing ‘negative’ controls (ie factors having the opposite effect of cytokines or TPA on signaling kinases), we used suppressors of Src and ERK signaling (PP2 or U0126 alone) respectively. An initial examination of the transcripts detected in P-UAEC regardless of fold change was consistent with known endothelial cell biology, including transcripts relevant to endocrine control of vasodilatory/mitogenic processes (endocrine factors and receptors, junctional proteins, connexin isoforms as identified previously). A summary is given in Suppl Table 1. There was a high degree of match for isoforms detected here vs those detected previously by GF211 microarray or as proteins by BD Powerblot antibody panel (Gifford et al 2003). These included eNOS, cyclooxygenase 1, CD31/PECAM, VE cadherin (VE-Cad), connexins GAPJ37, 40 and 43, and multiple isoforms of G protein coupled receptors (including adenosine, adrenergic (alpha and beta), cholinergic, GABA, purinergic subtypes X and Y, and serotonin, as well as angiotensin Type 1, bradykinin B2, dopamine D3, and endothelin A). We also detected multiple expected isoforms of growth factor receptors to VEGF, FGF, PDGF and EGF as reported previously.

Further analysis using TAC consol (Fig 1) showed that exposure to TNF for 12 hours was far more effective at increasing differential expression (DE) of multiple transcripts compared to VEGF, which had almost no effect. The positive control TPA generated hundreds of transcript changes >2 fold relative to control, as expected for an agent working through a wide variety of kinases. TNF predominantly generated increases in transcripts, with very few negatively regulated. PP2 and U0126 had mixed effects on multiple transcripts as would be expected for inhibitors of multiple kinases. Further data processing and statistical analysis as described in methods verified that TNF can significantly promote DE of 32 transcripts, and we provide an illustration of how they distribute relative to other treatments (Fig 2). Of note, 12 of the 32 transcripts are differentially regulated by TNF, but not any other treatment. Out of the 32 identified transcripts with >2 fold increase with FDR<0.05 (Fig 2), at least half were associated with chemokine and cytokines secretion or other inflammatory processes. ‘Pathway’ analysis of the 32 genes is shown in Table 1, with 12 GO terms achieving significance.

Figure 1: Volcano plots for DE transcripts in P-UAEC following treatment with TNF, VEGF, and controls.

Figure 1:

Open in a new tab

Following treatment with TNF (10 ng/ml), VEGF (10 ng/ml), TPA (10 nM), PP2 (10 uM) and U0126 (10 uM) for 12 h, media was removed and total RNA recovered from attached cells for analysis as described. Data is from n=4 independent experiments in each case. DE transcripts are defined as those significant at 2 fold or more DE with an FDR <0.05 Changes for all transcripts are shown, but those that did not acheive 2 fold and significance are greyed out. Those that are >2.0 fold DE and significant are shown as solid color. While VEGF had no significant effect, TNF impacted multiple transcripts.

Figure 2: Identity of transcripts induced by TNF, and relationship to the effect of other treatments.

Figure 2:

Open in a new tab

Using the same data set described in Fig 1 (n=4 each) this representation further details the identity and magnitude of changes in individual transcripts induced by TNF (10 ng/ml) for 12 hours. Also shown is a Venn diagram that summarizes DE transcript changes relative to other treatments (other than VEGF which had no effect). Gene names are given. Red indicates a relationship to immune modulation, blue to chemokine or cytokine signaling, and purple to extracellular protease activity.

Table 1: [SAM] PATHWAY Analysis of TNF regulated transcripts using Multiple DATABASE.

GO terms generated by Pathway analysis of the 33 DE transcripts in response to TNF are shown, along with significance (P<0.05, indicated by *). Note 12 of the 20 terms are significant.

Category Term Count Genes Benjamini
adjusted P-value
KEGG_PATHWAY Cytokine-cytokine receptor interaction 9 CCL20, CCL5, CCL8, CXCL10, CXCL11, CXCL6, IL1A, IL1R1, TNFSF15 3.9E-6*
KEGG_PATHWAY Chemokine signaling pathway 7 CCL20, CCL5, CCL8, CXCL10, CXCL11, CXCL6, NFKBIA 1.4E-4*
REACTOME_PATHWAY R-HSA-380108:Chemokine receptors bind chemokines 5 CCL20, CCL5, CXCL10, CXCL11, CXCL6 3.6E-4*
KEGG_PATHWAY Cytosolic DNA-sensing pathway 5 CCL5, CXCL10, DDX58, NFKBIA, ZBP1 3.5E-4*
BBID 109.Chemokine_families 6 CCL20, CCL5, CCL8, CXCL10, CXCL11, CXCL6 9.8E-4*
KEGG_PATHWAY Influenza A 6 CCL5, CXCL10, DDX58, NFKBIA, IL1A, RSAD2 9.0E-4*
KEGG_PATHWAY Rheumatoid arthritis 5 CCL20, CCL5, CXCL6, IL1A, MMP3 7.5E-4*
KEGG_PATHWAY TNF signaling pathway 5 CCL20, CCL5, CXCL10, NFKBIA, MMP3 1.3E-3*
REACTOME_PATHWAY R-HSA-909733:Interferon alpha/beta signaling 4 ISG15, MX2, IFIT3, RSAD2 9.3E-3*
KEGG_PATHWAY RIG-I-like receptor signaling pathway 4 CXCL10, DDX58, ISG15, NFKBIA 5.9E-3*
BIOCARTA NF-kB Signaling Pathway 3 NFKBIA, IL1A, IL1R1 6.5E-2
REACTOME_PATHWAY R-HSA-418594:G alpha (i) signalling events 5 CCL20, CCL5, CXCL10, CXCL11, CXCL6 2.9E-2*
KEGG_PATHWAY Toll-like receptor signaling pathway 4 CCL5, CXCL10, CXCL11, NFKBIA 1.7E-2*
BIOCARTA Signal transduction through IL1R 3 NFKBIA, IL1A, IL1R1 6.7E-2
REACTOME_PATHWAY R-HSA-1169408:ISG15 antiviral mechanism 3 DDX58, ISG15, MX2 1.1E-1
KEGG_PATHWAY NF-kappa B signaling pathway 3 DDX58, NFKBIA, IL1R1 1.2E-1
REACTOME_PATHWAY R-HSA-168928:DDX58/IFIH1-mediated induction of interferon-alpha/beta 2 DDX58, ISG15 2.7E-1
KEGG_PATHWAY Osteoclast differentiation 3 NFKBIA, IL1A, IL1R1 2.3E-1
KEGG_PATHWAY Measles 3 DDX58, NFKBIA, IL1A 2.1E-1
REACTOME_PATHWAY R-HSA-1810476:RIP-mediated NFkB activation via ZBP1 2 NFKBIA, ZBP1 2.8E-1
Open in a new tab

We then went on to examine the release of such factors by analysis of proteins in conditioned media. Our initial evaluation was undertaken with a human targeted (HCYTOMAG-60K) Milliplex kit, with mixed success. Of note, several analytes not detected show only ~80% homology to human proteins. Because of this, we further supplemented this analysis with a more limited but newly available ‘ovine’ kit, and backed this up with media proteomics (which did not depend on antibody specificity at all). Our combined assay results (Fig 3–5) were consistent with the transcripts detected by Affymetrix analysis of transcripts (Supplementary Table 1), and some but not all hormone secretions following TNF treatment were significantly increased. Overall our combined methods confirmed that TNF was a far more effective and indeed selective stimulant of multiple growth factors, cytokines, and chemokines in contrast to VEGF. Of particular note, Figs 3 and 4 show the dramatic increases in production of EGF, IL1a, IL6 and IL8 observed with the ovine kit, and dramatic increases in IL7 and IL12p70 were observed with the ‘human’ kit. Of the limited number of chemokines detected by the Milliplex kits (Fig 5), CXCL10 and fractalkine (CX3CL1) were significantly increased. In addition, several chemokines (CCL2, CCL5, CCL17, CCL20) as well as cytokines IL6 and IL8 (CXCL8) were detected by proteomics, with CCL2, CCL5, CCL20 and IL6 achieving significance (Fig 6).

Figure 3: Milliplex assay of Growth Factors secreted into media by P-UAEC treated for 18 h.

Figure 3:

Open in a new tab

Media from cells treated with TNF (10 ng/ml), VEGF (10 ng/ml), TPA (10 nM), PP2 (10 uM) and U0126 (10 uM) for 18 h was recovered and assayed for secreted endocrine factors by Multiplex assay as per Methods. Each box and whisker plot shows data from n=6–7 independent replicates. Mean values are indicated by the bar within the box, and * above the bar indicates significant differences at P<0.05. Where both ovine and human kit values were obtained, ovine kit values are reported. Where human kit data only was available, the graph title indicating analyte shows an additional ‘*’.

Figure 5: Milliplex assay of Chemokines secreted into media by P-UAEC treated for 18 h.

Figure 5:

Open in a new tab

Media from cells treated with TNF (10 ng/ml), VEGF (10 ng/ml), TPA (10 nM), PP2 (10 uM) and U0126 (10 uM) for 18 h was recovered and assayed for secreted endocrine factors by Multiplex assay as per Methods. Each box and whisker plot shows data from n=6–7 independent replicates. Mean values are indicated by the bar within the box, and * above the bar indicates significant differences at P<0.05. Where both ovine and human kit values were obtained, ovine kit values are reported. Where human kit data only was available, the graph title indicating analyte shows an additional ‘*’.

Figure 4: Milliplex assay of Cytokines secreted into media by P-UAEC treated for 18 h.

Figure 4:

Figure 4:

Open in a new tab

Media from cells treated with TNF (10 ng/ml), VEGF (10 ng/ml), TPA (10 nM), PP2 (10 uM) and U0126 (10 uM) for 18 h was recovered and assayed for secreted endocrine factors by Multiplex assay as per Methods. Each box and whisker plot shows data from n=6–7 independent replicates. Mean values are indicated by the bar within the box, and * above the bar indicates significant differences at P<0.05. Where both ovine and human kit values were obtained, ovine kit values are reported. Where human kit data only was available, the graph title indicating analyte shows an additional ‘*’.

Figure 6: Proteomics analysis of condition media following 18h treatment.

Figure 6:

Open in a new tab

The same media analyzed in Fig 3,r4 and 5 (for n= 4 replicates each due to cost) was also subjected to proteomics analysis as described. In this figure we show the DE proteins that achieved significant (P<0.05) fold changes against control, presented in the same way as for transcripts in Fig 1.

In Fig 7 we looked further at a possible relationship between the observed DE of transcripts in response to each treatment and the correspondingly detected increase in protein. There are limitations to this analysis given we are looking at secretions to media, and not directly at the cells, but the results are nonetheless informative. Given many of the ovine chemokines and cytokines were not detected by the human Milliplex kit, the data in Fig 7 is limited. Nonetheless, in Fig 7 we see that for Growth Factors (EGF, bFGF, VEGFA, PDGF-AA, and PDGF-AB/BB), responses to TPA or TNF are predominantly among the biggest fold increases, and responses to PP2 or U0126 (triangles) are predominantly the biggest fold decreases below 1. Nonetheless, there is only a weak positive relationship by linear regression of all data points, and a Spearman’s Rank correlation coefficient of only 0.255 fails to reach significance. In contrast, analysis of the more extensive cytokines data (CSF2, CSF3, IFNA2, IFNG, TNF, IL1A, IL1B, IL4, IL6, IL7, IL8, IL10 and IL13) showed the strongest fold responses were again related to the action of TPA and TNF, and the strongest decreases below 1 were related to PP2 and U0126. In this case, while a stronger positive linear regression gradient failed to reach significance, the Spearman’s Rank Correlation Coefficient was 0.457 and this was highly significant at P<0.001). Lastly, an analysis of the chemokines data was limited since it only included CXCL10, CX3CL1, CCL3, and CCL4. While CCL5 and CCL20 were detected as increased by Proteomics, they were not detected by Milliplex Assay. The semiquantitative nature of the proteomics analysis prevented us from including the data in this graph, and the resulting lack of any significant trend was unsurprising given the limited data points. Thus the negative result should be considered with caution.

Figure 7: Comparison of fold change in protein secreted to medium vs DE of mRNA transcript in P-UAEC following treatments.

Figure 7:

Open in a new tab

Each plot shows combined data for all treatments on Growth factors (Left Panel), Cytokines (Middel Panel) or Chemokines (Right Panel), where both transcript and secreted protein data was available. Outcomes of positive stimuli (TPA, TNF) were plotted as circles, and negative controls (PP2, U0126) as triangles. Combined data on each plot was then fitted with a regression line with 95% confidence intervals. A possible relationship between the two parameter was tested first by regression analysis, and then by Spearman’s rank order correlation. Significance in each case is as shown.

Beyond confirmation of secreted endocrine factors, proteomics sequence analysis also provided information on the state of the protein being detected. Intracellular junctional proteins ZO-1 and CTNNA1, CTNNA2, CTNNB1 were detected as peptide fragments matching throughout the coding sequence (not shown), consistent with them not being accessible to extracellular proteases and likely present at low levels as exosomes not fully removed by centrifugation. In contrast, the peptide fragments detected for VE-Cad and Junctional Adhesion Molecule A (JAM1) in particular were predominantly detected in regions confined to the extracellular domain alone (see supplementary figures 1 & 2), consistent with them being ‘shed’ by protease activity. Consistent with this, proteomics also detected MMP1, 2, 3, 9 and 28 along with their regulators TIMP1, 2, and 3, consistent with our prior transcript and antibody-based analysis of P-UAEC (Ampey et al 2021). In addition we detected ADAM9, 10, 15, 17, 19, and ADAMTS4, and S7. Beyond those proteins of interest, multiple isoforms of collagen were detected at comparatively high levels, as expected (Not shown).

Discussion.

In this study we examine if TNF or VEGF can alter expression and/or secretion of proinflammatory chemokines and cytokines otherwise reported to be elevated in PE subjects. These treatments were chosen based on our previous studies showing while VEGF and TNF can both act through ERK and Src signaling to acutely (1 hr) inhibit CX43 function via ERK and/or Src kinases, to inhibit vasodilation (Bird et al 2013), only longer term (overnight) treatment with TNF can break down the P-UAEC monolayer (Ampey et al, 2021). It is therefore relevant that the impact of TNF on the transcript profile of P-UAEC is also substantially greater than the effect of VEGF. Of the 32 transcripts significantly increased by TNF, 8 encode cytokines or chemokines as expected, and one encodes MMP3, consistent with the proteins reported to be elevated in PE subjects (Szarka et al 2010, Ma et al 2019, Liu et al 2023, Wang 2022). Of the Go Terms generated, it is unsurprising that a response to TNF would promote the term ‘Cytokine – cytokine receptor interaction’ or TNF signaling pathway’, but CCL20, CCL5, CXCL10, NFKBIA, and MMP3 combined also generated a GO term for another Th17 associated disease, rheumatoid arthritis, which is a known risk factor for PE. Several of the other differentially expressed transcripts (CD274, BST-2B, CD83, ISG15, ZBP1, GPA33, IF144, IFI44L, DDX58, RSAD2, MX2, IFIT3 (shown red in Fig1) are associated by ‘PANTHER’ with ‘immune response/inflammatory response’.

While the results of transcriptomic analysis alone are interesting, it is important to confirm that increases in DE transcripts were indeed indicative of increases in secreted protein. A direct comparison of mRNA to protein where protein is specifically detected shows TNF was very effective in promoting multiple endocrine factor secretions into media than VEGF. Multiple growth factors (VEGF, EGF, FGF2, and PDGFAA/AB) were significantly increased by TNF, along with multiple cytokines (IL1a, IL6, IL8, IL7 and IL12p70), and chemokines (CXCL10 and CXCL1, and by proteomics CCL2, CCL5 and CCL20). So it is clear the answer to our basic question if P-UAEC can secrete multiple proinflammatory cytokines and chemokines in response to local uteroplacental TNF is ‘yes’. Furthermore, our analysis of the relationship between DE mRNA levels across treatments (ie all except VEGF) versus changes in secreted protein confirm that for cytokines at least, there is a significant relationship. As such, signaling to transcriptional regulators may be key to why TNF has this ability to increase multiple proinflammatory mediators, while VEGF does not. Certainly it is true that the specific kinases mediating TNF vs VEGF signaling in P-UAEC are different (Ampey et al 2019, Boeldt et al 2015), and that TNF signaling through Src is more sustained. Src in turn is known to act through the JAK/Stat pathway via GP130, and such signaling is known to regulate several of these transcripts (Jalali et al 2022, McLoughlin et al 2005, Bae et al 2018). One other DE expressed transcript invoked by TNF but not VEGF encodes NFKBIA, a member of the NFKB transcriptional complex, and imbalance of JAK/Stat vs NFKB signaling has also been associated with altered MMP secretion profiles in other cell types (Smola-Hess et al, 2001).

Clinical Implications of our Findings:

While pregnancy is in itself an inflammatory condition, the development of PE in humans is associated with further elevation of a number of inflammatory markers (Redman & Sargent 2003). We have previously pointed out elevation of these cytokines in PE pregnancy is consistent with abnormal wound healing (Bird et al 2013). Several other studies to date relate to Th1>Th2 associated cytokines including TNF, IL1B and IL6, and more recently reports of proinflammatory alterations in IL10 and IL17 have also been made (Catarino et al 2012, Ma et al 2019). More recent studies have now added chemokines, and particularly CCL5 and CCL20 to that list (Liu et al 2023, Wang et al 2022). The assumption has been that most of these factors are coming from the immune system or perhaps the uteroplacental unit, but there has been little attention given to the possibility uterine artery endothelium itself may be a key source. Our studies herein show clearly that elevation of many of these factors in PE subjects in response to local TNF is as likely to be from uterine artery endothelium as it is from uterine tissue in general. Our findings agree with those reported previously for HUVEC treated longer term (Mai et al 2013), so while we acknowledge this is not a response specific to uterine artery, it is physiologically the most relevant to maternal health given uterine artery endothelium directly controls maternal blood flow entering the uteroplacental system (Sladek et al 1997).

Looking more specifically at the factors secreted, there is evidence that VEGF, FGF2 and PDGF AA/AB are altered in PE subjects, but there is little evidence to suggest a direct role in inflammation (Bird et al, 2013). The growth factor most elevated by TNF is EGF, and we have suggested recently that while EGF is not usually a strong stimulant of endothelial function or dysfunction due to low endogenous receptor numbers, it is possible EGFR rich exosomes known to be elevated in PE subjects could fuse with P-UAEC, so making an adverse response possible. If autocrine release of EGF then occurred secondary to TNF elevation, further damage to vasodilatory function could be mediated locally through this additional autocrine axis (Clemente & Bird 2022).

Several of the chemokines secreted by P-UAEC (CCL5 and CCL20) are known to attract proinflammatory T cells (Li et al 2022), and serum CCL5 and CCL20 detected as elevated by proteomics are both increased in PE subjects (Liu et al 2023, Wang et al 2022). Of note, Affymetrix analysis and proteomics analysis confirmed the previously reported increase in multiple MMP isoforms (Ampey et al 2021). The presence of the shed extracellular domains for two of their known targets (VE-Cad and JAM1) is consistent with this, and shed VE Cad is elevated in the circulation of PE subjects (Acikgoz et al 2024).

While chemokines may be important to trafficking immune cells to the endothelial surface, the soluble mediators of paracrine interaction between the two cell types are cytokines. TNF stimulation of P-UAEC caused a rise in multiple cytokines reported to be elevated in PE subjects (IL6, IL7, IL8 and IL12) (Szarka et al 2010, Ma et al 2019). Of those, IL6 is widely known to be secreted by many cell types in the uteroplacental space and can act with TNF to amplify its effects via GP130. While IL8 is elevated, it may actually have anti-inflammatory properties so it is not necessarily a concern. The more curious result is that for IL7 and IL12. It is possible secretion of endothelial IL7 in particular could help promote the emergence of maternal Th17 dominance (Webster et al (2014), Arbelaez et al (2015), Chen et al (2017)). Of note, P-UAEC did not show any cytokine secretory response to TNF for factors otherwise secreted by proinflammatory T cells, (namely TNF, IL1B, IFNg, IL17). This is physiologically relevant because it means that only by bringing T cells in close proximity with the vessel luminal surface itself would ‘docked’ endothelial cells experience near saturating levels of these particular cytokines that otherwise circulate at far lower concentrations.

In summary, our studies on P-UAEC show that endothelial cells are far from quiescent victims in PE and may in fact mediate the initial attraction of Th cells to the vessel surface through chemokines secretion. Further damage to P-UAEC may also occur through the co-secretion of cytokines including but not limited to IL6. Cytokines such as IL7 and IL12p70 secreted by P-UAEC may also impact on the expansion of adverse proinflammatory cell populations known to predominate in PE subjects. If those T cells are recruited and actually bind to P-UAEC, additional paracrine factors secreted by local T cells could cause further damage. While this study has only studied the ‘endocrine’ secretions of P-UAEC in response to TNF, the proposed significance of our findings also depends on if P-UAEC are equipped on their cell surface with the necessary proteins to dock to and interact with specific classes of immune cells, so opening the possibility of further paracrine interactions. This question, and the possibility that different P-UAEC cell subtypes mediate different responses, are addressed in our companion paper (Companion Paper 2**).

Supplementary Material

01

Suppl Figures: The consensus sequence for VE Cad (Fig S1) and JAM A (Fig S2) are shown along with peptide sequences detected by Proteomics. In each case, the transmembrane peptide sequence is indicated as underlined text.

02

Suppl Table 1: Initial Affymetrix Data Summary (from TAC Consol). Data was generated using n=4 replicate experiments. Data in each tab relate to the corresponding protein’s biological functions of interest, ie hormones, receptors, surface proteins etc. All data are shown as Log2 values. ‘Anova’ P value and ‘FDR’ values are also shown.

Acknowledgements:

The authors have no disclosures or conflicts of interest to declare. The authors would also like to acknowledge major grant support from the Unity Point-Meriter Research Foundation (Award #764), and the Dept ObGyn Research and Development Fund. Luca Clemente were funded by NIH Postdoctoral T32 HD101384. We would also like to thank the Gene Expression Center at UW Madison for providing equipment and staff to support the execution of these studies. In closing, we dedicate this study to the memory of J Austin, who was a central member of this team and lost to us before his time. Our heartfelt thanks go out to his family.

References

  1. Acikgoz AS, Gok K, Tuten A, Kaya B, Hamzaoglu K, Gayretli TK, Malik E, Guralp O (2024). Evaluation of serum levels of soluble vascular endothelial (sVE)-Cadherin in early- and late onset preeclampsia serum: sVE-Cadherin levels in preeclampsia. J Maternal Fetal Medicine. 37(1), 2326304. [DOI] [PubMed] [Google Scholar]
  2. Ampey AC, Boeldt DS, Clemente L, Grummer MA, Yi F, Magness RR, Bird IM (2019). TNF-alpha inhibits pregnancy-adapted Ca2+ signaling in uterine artery endothelial cells. Mol Cell Endocrinol 488:14–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Ampey AC, Dahn RL, Grummer MA, Bird IM (2021). Differential control of uterine artery endothelial monolayer integrity by TNF and VEGF is achieved through multiple mechanisms operating inside and outside the cell - Relevance to preeclampsia. Mol Cell Endocrinol 534:111368. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bae HC, Jeong SH, Kim JH, Lee H, Ryu WI, Kim MG, Son ED, Lee TR, Son SW (2018). RIP4 upregulates CCL20 expression through STAT3 signalling in cultured keratinocytes. Exp Dermatol 27(10):1126–1133. [DOI] [PubMed] [Google Scholar]
  5. Benjamini Y, and Hochberg Y (1995). Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological) 57, 289–300. [Google Scholar]
  6. Benyo DF, Miles TM, Conrad KP (1997). Hypoxia stimulates cytokine production by villous explants from the human placenta. J Clin Endocrinol Metab 82(5):1582–8. [DOI] [PubMed] [Google Scholar]
  7. Bird IM, Boeldt DS, Krupp J, Grummer MA, Yi FX and Magness RR. (2013). Pregnancy, programming and preeclampsia: Gap junctions at the nexus of pregnancy-induced adaptation of endothelial function and endothelial adaptive failure in PE. Current Vascular Pharmacology. 11:712–29. [DOI] [PubMed] [Google Scholar]
  8. Bird IM, Sullivan JA, Di T, Cale JM, Zhang L, Zheng J, Magness RR (2000) Pregnancy-dependent changes in cell signaling underlie changes in differential control of vasodilator production in uterine artery endothelial cells. Endocrinology. Mar;141(3):1107–17. [DOI] [PubMed] [Google Scholar]
  9. Boeldt DS, Grummer MA, Yi F, Magness RR, Bird IM (2015). Phosphorylation of Ser-279/282 and Tyr-265 positions on Cx43 as possible mediators of VEGF-165 inhibition of pregnancy-adapted Ca2+ burst function in ovine uterine artery endothelial cells. Mol Cell Endocrinol 412:73–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Catarino C, Santos-Silva A, Belo L, Rocha-Pereira P, Rocha S, Patrício B, Quintanilha A, Rebelo I (2012). Inflammatory disturbances in preeclampsia: relationship between maternal and umbilical cord blood. J Pregnancy. 2012:684384. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Carvalho B (2015). pd.ovigene.1.0.st: Platform Design Info for Affymetrix OviGene-1_0-st. R package version 3.12.0. [Google Scholar]
  12. Carvalho BS, Irizarry RA (2010). A Framework for Oligonucleotide Microarray Preprocessing. Bioinformatics, 26(19), 2363–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Chen Y, Chauhan SK, Tan X, Dana R. (2017). Interleukin-7 and -15 maintain pathogenic memory Th17 cells in autoimmunity. Journal of Autoimmunity 77, Pages 96–103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Clemente L, Bird IM (2022). The epidermal growth factor receptor in healthy pregnancy and preeclampsia. JME 70(1):e220105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Cuklina J (2018). Computational challenges in biomarker discovery from high-throughput proteomic data. Master of Applied Mathematics and Physics, DISS. ETH NO. 25390. Moscow Institute of Physics and Technology. https://www.research-collection.ethz.ch/handle/20.500.11850/307772. [Google Scholar]
  16. Gifford SM, Cale JM, Tsoi S, Magness RR, Bird IM (2003). Pregnancy-specific changes in uterine artery endothelial cell signaling in vivo are both programmed and retained in primary culture. Endocrinology 144(8):3639–50 [DOI] [PubMed] [Google Scholar]
  17. Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, Speed TP (2003) Exploration, Normalization, and Summaries of High Density Oligonucleotide Array Probe Level Data. Biostatistics. Vol. 4, Number 2: 249–264. [DOI] [PubMed] [Google Scholar]
  18. Jafri S, Ormiston ML (2017). Immune regulation of systemic hypertension, pulmonary arterial hypertension, and preeclampsia: shared disease mechanisms and translational opportunities. Am J Physiol Regul Integr Comp Physiol 313(6):R693–R705. [DOI] [PubMed] [Google Scholar]
  19. Jalali BM, Likszo P, Lukasik K (2022). STAT3 in porcine endometrium during early pregnancy induces changes in extracellular matrix components and promotes angiogenesis. Biol Reprod 107(6):1503–1516. [DOI] [PubMed] [Google Scholar]
  20. Liu T, Zhang T, Yu H, Shen H, Xia W (2014). Adjudin protects against cerebral ischemia reperfusion injury by inhibition of neuroinflammation and blood-brain barrier disruption. J Neuroinflammation 11:107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Liu M, Niu Y, Ma K, Leung PCK, Chen ZJ, Wei D, Li Y (2023). Identification of novel first-trimester serum biomarkers for early prediction of (preeclampsia. J Transl Med 21(1):634. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Ma Y, Ye Y, Zhang J, Ruan CC, Gao PJ. (2019) Immune imbalance is associated with the development of preeclampsia. Medicine (Baltimore) 98(14):e15080. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Mai J, Virtue A, Shen J, Wang H, Yang XF (2013), An evolving new paradigm: endothelial cells--conditional innate immune cells. J Hematol Oncol 6:61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. McLoughlin RM, Jenkins BJ, Grail D, Williams AS, Fielding CA, Parker CR, Ernst M, Topley N, Jones SA. (2005). IL-6 trans-signaling via STAT3 directs T cell infiltration in acute inflammation. Proc Natl Acad Sci USA;102(27):9589–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Myatt L, Webster RP (2009). Vascular biology of preeclampsia. J Thromb Haemost 7: 375–84. [DOI] [PubMed] [Google Scholar]
  26. Pober JS, Merola J, Liu R, Manes TD (2017). Antigen Presentation by Vascular Cells. Front Immunol 8:1907. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Redman CWG, Sargent IL (2003). Pre-eclampsia, the Placenta and the Maternal Systemic Inflammatory Response - A Review. Placenta 24, Supplement A, Trophoblast Research, Vol. 17, S21–S27. [DOI] [PubMed] [Google Scholar]
  28. Sladek SM, Magness RR, Conrad KP (1997). Nitric oxide and pregnancy. Am J Physiol 272(2 Pt 2):R441–63. [DOI] [PubMed] [Google Scholar]
  29. Smola-Hess S, Schnitzler R, Hadaschik D, Smola H, Mauch C, Krieg T, Pfister H (2001). CD40L induces matrix-metalloproteinase-9 but not tissue inhibitor of metalloproteinases-1 in cervical carcinoma cells: imbalance between NF-kappaB and STAT3 activation. Exp Cell Res 267(2):205–15. [DOI] [PubMed] [Google Scholar]
  30. Szarka A, Rigó J Jr, Lázár L, Beko G, Molvarec A (2010). Circulating cytokines, chemokines and adhesion molecules in normal pregnancy and preeclampsia determined by multiplex suspension array. BMC Immunol 11:59. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Tibshirani R, Seo MJ, Chu G, Narasimhan B, Li J (2018). samr: SAM: Significance Analysis of Microarrays. R package version 3.0. [Google Scholar]
  32. Tusher V, Tibshirani R and Chu G (2001). Significance analysis of microarrays applied to the ionizing radiation response. PNAS 98: 5116–5121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Wang X, Yip KC, He A, Tang J, Liu S, Yan R, Zhang Q, Li R (2022). Plasma Olink Proteomics Identifies CCL20 as a Novel Predictive and Diagnostic Inflammatory Marker for Preeclampsia. J Proteome Res 21(12):2998–3006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Webster KE, Kim HO, Kyparissoudis K, Corpuz TM, Pinget GV, Uldrich AP, Brink R, Belz GT, Cho JH, Godfrey DI, Sprent J. (2014) IL-17-producing NKT cells depend exclusively on IL-7 for homeostasis and survival. Mucosal Immunol 7(5):1058–67. [DOI] [PubMed] [Google Scholar]
  35. Yi FX, Boeldt DS, Gifford SM, Sullivan JA, Grummer MA, Magness RR, Bird IM (2010) Pregnancy enhances sustained Ca2+ bursts and endothelial nitric oxide synthase activation in ovine uterine artery endothelial cells through increased connexin 43 function. Biol Reprod 82(1):66–75. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

01

Suppl Figures: The consensus sequence for VE Cad (Fig S1) and JAM A (Fig S2) are shown along with peptide sequences detected by Proteomics. In each case, the transmembrane peptide sequence is indicated as underlined text.

NIHMS2102818-supplement-01.pdf (410.7KB, pdf)
02

Suppl Table 1: Initial Affymetrix Data Summary (from TAC Consol). Data was generated using n=4 replicate experiments. Data in each tab relate to the corresponding protein’s biological functions of interest, ie hormones, receptors, surface proteins etc. All data are shown as Log2 values. ‘Anova’ P value and ‘FDR’ values are also shown.

NIHMS2102818-supplement-02.xlsx (50.2KB, xlsx)

RESOURCES