{"paper_id":"1a267665-4ae4-4e20-92ab-4c926158d3bc","body_text":"Deletion of endothelial KLF4 as a model for preeclampsia 1 \n 2 \nEmily Meredith1, Andrew T. Meredith1, Arya Mani1,2, Martin A. Schwartz1,3 4* 3 \n 4 \n 5 \n 6 \n 7 \n1 Yale Cardiovascular Research Center, Section of Cardiovascular Medicine, 8 \nDepartment of Internal Medicine, School of Medicine, Yale University, New Haven, CT 9 \n06511, USA. 10 \n2Department of Genetics, Yale University, New Haven, CT 06510, USA. 11 \n3 Department of Cell Biology, Yale University, New Haven, CT 06510, USA. 12 \n4 Department of Biomedical Engineering, Yale University, New Haven, CT 06510, USA. 13 \n 14 \n 15 \n 16 \n*Corresponding author: martin.schwartz@yale.edu 17 \n 18 \n 19 \n 20 \n 21 \nKeywords: Hypertension, gestational hypertension, sFLT1, preeclampsia, endothelial 22 \ncell, KLF2/4 23 \n  24 \n  25 \nRunning title: Deletion of endothelial KLF4 as a model for Preeclampsia 26 \n  27 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n 28 \nAbstract 29 \nPreeclampsia (PE), or gestational hypertension, affects around 5% of pregnancies and 30 \nleads to approximately 70,000 maternal and 500,000 fetal deaths per year worldwide, 31 \nwith increased cardiovascular and metabolic disease in survivors.  PE is associated with 32 \nelevated circulating levels of the alternative splice isoform of VEGF receptor 1 (sFlt1), 33 \ndefects in placental vasculature, kidney damage and, in severe disease, fetal growth 34 \nrestriction. Current mouse models induce PE via direct expression of sFlt1 or elevation 35 \nof blood pressure, which bypass the natural risk factors for human disease, such as 36 \nage, obesity, hypertension and diabetes.  These risk factors have in common reduced 37 \nexpression of Krüppel-like factors 2 and 4 (KLF2/4), the endothelial transcription factors 38 \nthat protect against cardiovascular disease. We now report that inducible deletion of 39 \nKLF4 in maternal endothelium (KLF4iECKO) results in gestational hypertension, elevated 40 \nsFlt1, defective placental vasculature, kidney damage and fetal growth restriction.   41 \nKLF4iECKO may thus serve as a mouse PE model suitable for mechanistic analysis and 42 \nscreening of treatments that address upstream risk factors.  43 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nIntroduction 44 \n Despite advances in detection and classification, preeclampsia (PE) remains a 45 \nmajor cause of fetal death and maternal morbidity and mortality worldwide [1, 2]. In 46 \naddition to short term effects on health, women after preeclamptic pregnancy show 47 \nlarge increases in cardiovascular disease (CVD) incidence and mortality [3]. CVD risk 48 \nscales with PE severity [4], defined as late (> 34wks) vs early-onset (< 34wks), as mild 49 \n(140mmHg < BP < 160mmHg) vs severe (BP > 160mmHg), and with vs without fetal 50 \ngrowth restriction (FGR)[4]. The most dangerous form of PE is early-onset, severe, with 51 \nFGR, which raises the incidence of maternal CVD later in life as well as the incidence of 52 \nmetabolic syndrome and neurodevelopmental delays in the child [5]. Methods to 53 \nmanage PE are limited, indeed, the only cure is delivery of the placenta and fetus. 54 \nManagement is possible, but treatment options are limited out of concern for the 55 \ndeveloping fetus. 56 \nThe principal risk factors for PE – age, obesity, hypertension and diabetes – mirror 57 \nthose for CVD. Widely used PE mouse models induce hypertension by infusion of 58 \nangiotensin II or elevation of sFLT1 through viral overexpression, bypassing the 59 \nendogenous regulatory pathways that govern PE (ref?). Thus, they do not account for 60 \nthe upstream risk factors in human disease.  These human risk factors, however, share 61 \na common feature: they are opposed by endothelial cell (EC) expression of Krüppel-like 62 \nfactors 2 and 4 (KLF2/4), homologous transcription factors with overlapping (but 63 \nnonidentical) gene targets and functions in ECs  [6-8]. Klf2/4 expression in ECs declines 64 \nwith age [9, 10] and in diabetes [11-13] including gestational diabetes [14]. Elevating 65 \nKlf2 or 4 confers resistance to multiple CVDs in mouse models [15, 16].  KLF2/4 66 \nfunction is vital to blood pressure regulation by inducing eNOS to control vascular tone 67 \nand limiting vascular inflammation[17, 18].  While mouse EC-specific knock out (ECKO) 68 \nof all four alleles of Klf2 and 4 is lethal, ECKO of 1-2 copies of these genes did not 69 \naffect mouse survival [17]. A recent study showed that ECKO of Klf4 had little effect in 70 \nyoung mice but markedly accelerated vascular aging (Jain, soon, I hope).   71 \nMeasurement of blood pressure and circulating sFLT1 is the current standard for 72 \ndiagnosing PE, while sFlt1 is also critical to its pathophysiology [19, 20]. The placenta is 73 \nthought to be the main source of circulating sFLT1, though a contribution from the 74 \nmaternal vascular has not been ruled out and could explain features of disease 75 \nprogression [21, 22]. Here, we report the development of a genetic model of PE driven 76 \nby EC-specific KLF4 deletion, which offers a physiologically relevant platform that more 77 \nclosely mimics the human risk factors that drive PE [12]. 78 \n 79 \nMethods 80 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nsiRNA KD of KLF2 or 4 in HUVECs 81 \nHUVECs were plated such that they were between 50-80% confluent at the time 82 \nof siRNA treatment. Lipofectamine RNAiMAX (ThermoFisher Cat #13778075) was used 83 \nto transfect 10nM KLF2, KLF4 or control siRNA into cells in Opti-Mem, as 84 \nrecommended by the manufacturer. The mixture was incubated on cells overnight in 85 \ncomplete media (Lonza EGM-2 Endothelial Cell Growth Medium-2 BulletKit cat #CC-86 \n3162) before they were replenished and allowed to recover for 24h prior to collection for 87 \nqPCR analysis.  88 \nFor all qPCR analysis, cDNA was generated using the iSCRIPT cDNA Synthesis 89 \nKit (BioRad Cat#1708890) and qPCR was performed using SSO Advanced SYBR 90 \nGreen (BioRad Cat#1725270). Primer sequences used are as follows: 91 \nTarget FWD REV \nhuKLF2 AAGAGCTCGCACCTAAAGGC CTTTCGGTAGTGGCGGGTAA \nhuKLF4  CTATGCAGGCTGTGGCAAAACC TTGCGGTAGTGCCTGGTCAGTT \nhuGAPD\nH \nGTCTCCTCTGACTTCAACAGCG ACCACCCTGTTGCTGTAGCCAA \nhuFLT1 TGGCAGCGAGAAACATTCTTTTAT\nC \nCAGCAATACTCCGTAAGACCACA\nC \nhusFLT1 ACAATCAGAGGTGAGCACTGCAA TCCGAGCCTGAAAGTTAGCAA \n 92 \nRNAscope on Placenta Sections 93 \nPlacentas were dissected at GD18.5, separated from the yolk sac and pup and 94 \nimmersion-fixed in 4% PFA overnight at 4C with gentle rocking. Images were taken of 95 \nthe underside (pup side) for visceral placenta vascularization analysis. Placentas were 96 \nthen incubated in 30% sucrose overnight at 4C with gentle rocking. Once placentas 97 \nwere saturated, as indicated by their sinking in the sucrose solution, they were cut in 98 \nhalf longitudinally and frozen cut side down in OCT at -80C. Rapid freezing was 99 \naccomplished by using acetone chilled by dry ice. A cryostat was then used to cut 10um 100 \nlongitudinal sections.   101 \nSlides containing placenta sections were then baked at 60C for 1hr, washed in PBS 102 \nand post-fixed in 4% PFA for 1hr. Slides were then dehydrated in EtOH prior to 103 \nbeginning H2O2 treatment. ACD’s Multiplex Fluorescent V2 protease-free protocol was 104 \nfollowed. sFLT1 probe: FLT1 probe: multiplex kit:. Following completion of the 105 \nhybridization protocol, sections were then subjected to FN protein staining. Briefly, 106 \nsections were washed in TBS/0.1%Tween 20 (TBST) three times, 5mins each. The 107 \nsections were then blocked in TBS/5%BSA for 1hr followed by FN antibody incubation 108 \n(Millipore cat#F3648, 1:500) for 2h at RT. Primary antibody was then washed off with 109 \nTBST and secondary antibody (Alexa Fluor series from ThermoFisher; 1:1000) plus 110 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nDAPI (ThermoFisher Cat#D1306, 1:10,000) was added in TBS/5% w/v BSA for 1hr at 111 \nRT. Sections were then washed a final time before using TrueView Autofluorescence Kit 112 \n(Vector Laboratories cat#SP-8400-15) diluted 1:10 to decrease background staining. 113 \nSlides were then mounted using Vector Lab’s VactaShield vibrance antifade mounting 114 \nmedium (cat#H-1700-2). Imaging was done on a Zeiss Leica Confocal Microscope 115 \nusing the Leica Application Suit X software. A 40x oil objective was used for all images. 116 \nZ-stacks of approximately 10um were taken with at least 5-6 images taken per sample. 117 \nWith RNAscope, it can be challenging to identify real versus background signal. Two 118 \nways we distinguished signal from noise were by firstly setting imaging parameters 119 \nbased off sections that were stained with positive and negative control probes provided 120 \nby ACD. Secondly, by utilizing z-stacks, areas that displayed real signal were more 121 \napparent if RNAscope spots appeared in multiple z-sections. Maximal projections were 122 \nused to enhance spots that were real versus not real.   Image analysis was done using 123 \nFiji with maximal z-projections. 124 \nsFLT1 ELISA Measurements 125 \nBlood samples were taken from dams at GD18.5 via cardiac puncture. Blood was 126 \ncollected into EDTA-coated K2 tubes (VWR cat#76343-512) and spun at 10,000g for 127 \n10mins at 4C to separate the serum, which was then isolated and flash frozen in LN2 128 \nbefore being stored at -80C. Samples were then thawed on ice prior to analysis. sFLT1 129 \nserum levels were analyzed using R&D’s VEGR1 ELISA kit (Cat #MVR100) with the 130 \nDuoSet ELISA Ancillary Reagent Kit (Cat #DY008B). Serum samples were diluted from 131 \n1:5-1:20 to ensure samples landed within the standard curve. Raw pg/mL values are 132 \nplotted. 133 \nMice 134 \niECKO and Timed matings 135 \n8-9wk old KLF4f/f;CDH5cre-ERT2+ dams were injected with 80mg/kg/day tamoxifen for 5 136 \ndays to induce ECKO. Dams were rested for 1wk following injections as tamoxifen can 137 \nreduce fertility [23]. At the same time, male B6 littermates were separated 1wk prior to 138 \ncombination with females. Dams were then provided with dirty male bedding for 72hrs 139 \nprior to introduction of males to induce estrus. 1-2 females were then introduced to male 140 \ncages for 3 days before separating. Pregnancy was confirmed via plug checking, 7-day 141 \nweight gain ≥2g or by blood pressure status.   142 \nBlood pressure monitoring via the CODA System 143 \nMice were trained for 3-5 days prior to data collection. Training consisted of regular 144 \nblood pressure measurements within the system. To reinforce training, foraging mix 145 \n(VWR cat #76628-302) or other treats were given following any blood pressure 146 \ncollection. Blood pressure was taken every other day for 2 weeks following confirmation 147 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nof pregnancy (GD7.5-GD17.5). Data per day consists of at least 3 measurements 148 \naveraged. Data for non-pregnant control mice consists of at least 3 averaged 149 \nmeasurements every other day for 2 weeks giving at least 18 averaged measurements 150 \nper control mouse. To quantify acquisition stability, we calculated the percent coefficient 151 \nof variation (%CV) for repeated tail-cuff cycles within each session for each mouse. 152 \nEarly acclimation sessions exhibited higher variability, consistent with handling-related 153 \nstress, whereas CV progressively decreased with training. Sessions demonstrating 154 \n%CV values within the expected physiological range (<10%) were considered stable.  155 \nKidney Analysis 156 \nSections and IF staining 157 \n Kidneys were isolated from perfusion-fixed mice at GD18.5 and incubated in 4% 158 \nPFA overnight at 4C with gentle rocking. Kidneys were then soaked in 30% sucrose 159 \novernight at 4C with gentle rocking or until the kidney was saturated as indicated by it 160 \nsinking in the solution. Kidneys were then cut in half longitudinally, dabbed of excess 161 \nsucrose and embedded in OCT with the cut-side down. Rapid freezing was 162 \naccomplished using acetone chilled with dry ice. 10um sections were taken using a 163 \ncryostat set at -20C. For staining, slides were thawed at room temperature (RT) and 164 \nwashed three times in PBS for 5mins to remove excess OCT. Samples were then 165 \nblocked and permeabilized in PBS/0.1% TritonX/5% BSA for 1hr at RT. Primary 166 \nantibodies were diluted in PBS/TritonX/BSA (FN (Millipore cat#F3648, 1:500), CD31 167 \n(Fisher Scientific cat#AF3628, 1:1000), CD34 (Abcam cat#AB81289)) and incubated 168 \novernight in a humidity chamber at 4C. Primary antibodies were then washed off with 169 \nPBS/0.1% tween20 (PBST) three times for 5mins. Alexa Fluor Secondary antibodies 170 \nwere diluted in PBS/TritonX/BSA at 1:1000 and left for 1hr at RT. Slides were then 171 \nwashed again in PBST before the excess was blotted off. Slides were mounted using 172 \nProlong Gold antifade mounting medium (Invitrogen cat#P36980) and allowed to cure 173 \novernight prior to imaging. Imaging was done on a Zeiss Leica Confocal Microscope 174 \nusing the Leica Application Suit X software. A 20x air objective was used for all kidney 175 \nimages. Image analysis was done using Fiji. 176 \nH&E staining 177 \nKidney sections were taken as described in the previous section. H&E staining of 178 \ntissue sections was done by the Yale Research Histology Core using standard 179 \ntechniques 180 \nUrine collection and ACR 181 \n Urine was collected from GD17-18.5 dams prior to dissection using a clean cage 182 \nwithout bedding, covered with saranwrap. The mouse was allowed to freely walk around 183 \nuntil naturally voiding at which point the urine was quickly collected via clean syringe 184 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nand stored at -80C prior to analysis. ACR was then measured using two kits: Albuwell M 185 \n(Ethos biosciences cat#CHM03V055, and the creatinine companion kit (Ethos 186 \nbiosciences cat# CHM03V058) Samples were diluted 1:2-1:5.   187 \nResults 188 \nKLF2 association with hypertension 189 \nGiven that both PE and cardiovascular disease share key risk factors, we 190 \ninterrogated the CVD Knowledge Portal (CVDkP) for variants in KLF2/4. Although 191 \nassociations between KLF2/4 SNPs and PE did not reach statistical significance, likely 192 \nbecause of the limited statistical power of the study, variant rs3745318 in KLF2 is 193 \nassociated with blood pressure and coronary artery disease (Figure 1A and [24]). 194 \nFurthermore, this variant is identified in the GTEx database as an eQTL for KLF2 in 195 \nwhole blood (p = 1.27e-6), supporting a potential regulatory role in vascular pathways 196 \nlinked to PE (figure 1B). 197 \nTo address function, we suppressed KLF2 or 4 expression in HUVECs in vitro 198 \nand used QPCR to quantify both total Flt1 (tFlt1) and the soluble isoform (sFlt1). siRNA-199 \nmediated knockdown (KD) of Klf2 (validated in Fig 1C) increased the sFlt1/tFlt1 ratio by 200 \napproximately 2-fold, without significantly changing total Flt1 (tFlt1) mRNA (Fig 1D). Klf4 201 \nknockdown (validated in Fig 1E) similarly increased the sFlt1/tFlt1 ratio (Fig 1F). 202 \nKnockdown of Klf2 or 4 thus induces a shift in mRNA splicing rather than expression. 203 \nThe established causal role of sFlt1 in PE, together with genetic links between Klf2/4 204 \nand hypertension, prompted us to examine a possible role for KLF signaling in PE [24].  205 \nKLF4 iECKO triggers severe PE. 206 \n In mice, early-onset, severe preeclampsia is characterized by elevated blood 207 \npressure (average BP>130) before the third trimester, elevated circulating sFLT1, and 208 \nevidence of kidney or other organ damage [4, 25]. KLF4f/f;CDH5Cre-ERT2 dams were treated 209 \nwith tamoxifen to specifically delete endothelial KLF4 (KLF4iECKO) then bred with wild-210 \ntype B6 males, producing WT offspring. Thus, any observed phenotype is attributable to 211 \nthe maternal genotype rather than the fetus.  212 \nBlood pressure was measured using a CODA tail-cuff telemetry system. To ensure 213 \naccuracy, within-session variability, expressed as percent coefficient of variation (%CV) 214 \nof tail-cuff measurements was calculated. %CV declined across acclimation sessions, 215 \nmarked on all graphs as a purple box, indicating effective mouse training (supplemental 216 \nfigure 1A-C). Percent CV was calculated per individual per experimental session and 217 \nsessions above 10% CV were discarded (Supplemental figure 1A-C). We observed no 218 \ndifference in resting blood pressure between KLF4iECKO and KLF4f/f dams prior to the 219 \nonset of pregnancy (Figure 2A, time point -1; supplemental figure 1D-F), as is often the 220 \ncase in human PE.  However, GD6.5, KLF4iECKO dams had a trend toward elevated BP 221 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nby day 2.5, did not go through the mid-gestation nadir at around GD8.5, and remained 222 \nelevated over controls at later times (Figure 2A; supplemental figure 1G-H).  KLF4iECKO 223 \nthus elevates blood pressure within the first trimester of pregnancy in mice.  224 \nTo assess similarities to human preeclampsia, we next measured circulating sFLT1 225 \nand indicators of end organ damage. At GD18.5, pregnant KLF4iECKO dams had strongly 226 \nelevated sFLT1 compared to controls (Figure 2B; supplemental figure 2A). Kidney 227 \ndamage was analyzed using CD34 staining, a proteinuria assay and H&E glomeruli 228 \nstaining. These metrics revealed increased proteinuria, CD34 expression and capillary 229 \nocclusion in KLF4iECKO dams compared to controls (figure 2C-D, supplemental figure 230 \n1H-I) [26].  231 \nWhile evidence concerning which cell types contribute to elevated sFLT1 in the 232 \nmaternal circulation is mixed, it is known that maternal ECs can produce sFLT1 [27]. 233 \nThis source potentially creates a positive feedback loop, promoting maternal endothelial 234 \ndysfunction and CVD risk, which can worsen fetal sFlt1 production.  We therefore 235 \nassayed sFlt1 mRNA in arterial endothelium of KLF4iECKO dams using RNAscope 236 \nprobes specific for sFLT1 and total FLT1(supplemental figure 2B-D).  Examination of 237 \naortic endothelium en face revealed elevated sFLT1 mRNA in ECs in KLF4iECKO dams 238 \ncompared to cre- controls (Figure 2E-F). Together, these data indicate that KLF4iECKO 239 \ndams display a severe, early-onset PE phenotype. 240 \nKLF4iECKO triggers fetal growth restriction  241 \n Severe PE is associated with poor placental perfusion, leading to fetal growth 242 \nrestriction (FGR). Measures for FGR in mice include reduced pup weight and litter size, 243 \nwhich were both strongly decreased in KLF4iECKO litters (Figure 3A-B). Additionally, 244 \nplacental vascular area, measured both by longitudinal sections and macroscopically, 245 \nwas decreased in KLF4iECKO placentas (Figure 3C-D). RNAscope analysis of placenta 246 \nsections showed increased sFLT1 mRNA in the dense labyrinth in KLF4iECKO placentas 247 \n(Figure 3E; supplemental figure 3). Overall, these data confirm that maternal deletion of 248 \nendothelial Klf4 results in severe preeclamptic pregnancy and FGR.  249 \nDiscussion  250 \nPreeclampsia is a severe disease of pregnancy leading not only to maternal or fetal 251 \ndeath but to life-long increased incidence of CVD in surviving mother [28, 29] and 252 \nincreased metabolic syndrome, impaired vascular function and neurodevelopmental 253 \nproblems in surviving offspring [5, 30]. Mild PE is associated with a ~2-fold increased 254 \nincidence of maternal CVD while severe, early onset PE increases future CVD by 255 \nalmost 10-fold [31]. Our data identify maternal KLF4iECKO as a novel genetic model of 256 \nearly onset PE with FGR, the most severe form. Unlike viral sFLT1 overexpression, this 257 \napproach preserves endogenous regulation and maternal–placental interactions, 258 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nproviding a physiologically relevant framework for studying how maternal vasculature 259 \ndefects contribute to development of PE. It may also better represent disease in 260 \nhumans where the main risk factors are known to decrease endothelial Klf2 and 4 261 \nexpression.  PE was also observed in a spontaneously hypertensive rat model [32] but 262 \nthe mouse KLF4iECKO model offers greater opportunity for reverse genetic functional 263 \nanalysis as well as better recapitulating PE predisposition in otherwise healthy women 264 \n[30].  265 \nTreating PE is a considerable challenge, due to a large extent to the risk of harm to the 266 \nfetus.  Hypertension management through aspirin administration is currently the only 267 \nwidely used therapy to mitigate PE symptoms. Our model suggests that the maternal 268 \nvasculature may offer viable targets. Indeed, the KLF2/4 target gene eNOS, a major 269 \ncomponent of blood pressure regulation and target of hypertensive risk factors, is under 270 \nconsideration as a target for PE treatment [33, 34]. Recently developed antibodies [6] 271 \nand siRNaS [35] that target vascular endothelium and elevate Klf2/4 expression, 272 \nperhaps in conjunction with delivery via nanoparticles, Fab fragments and some 273 \nantibody subtypes, e.g., IgA and IgM, that do not efficiently cross the placental barrier 274 \noffer some promise here. The Klf4iECKO model therefore offers a suitable platform to test 275 \nthese interventions in future studies.  276 \n 277 \n 278 \n 279 \n 280 \n 281 \n 282 \n 283 \n 284 \n 285 \n 286 \n 287 \nAcknowledgments 288 \nWe thank the Yale Research Histology Core, M. Jain (Brown University) for the 289 \nKLF4f/f;CHD5cre-ERT2 mice, H. Aldrich for sample OCT embedding, and the Schwartz Lab 290 \nmembers for the extensive discussions.    291 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nSources of Funding 292 \nThis work was supported by a National Institutes of Health grant no. RO1 HL171773 to 293 \nM.A.S and a T32 Fellowship no 5T32HL007950 to E.M. 294 \nDisclosures 295 \nThe authors declare no competing interests. 296 \n 297 \n 298 \n 299 \n 300 \n 301 \n 302 \n 303 \n 304 \n 305 \n 306 \n 307 \n 308 \n 309 \n 310 \n 311 \n 312 \n 313 \n 314 \nReferences 315 \n1. Cresswell, J.A., et al., Global and regional causes of maternal deaths 2009-20: a 316 \nWHO systematic analysis. Lancet Glob Health, 2025. 13(4): p. e626-e634. 317 \n2. Ma'ayeh, M. and M.M. Costantine, Prevention of preeclampsia. Semin Fetal 318 \nNeonatal Med, 2020. 25(5): p. 101123. 319 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n3. Bellamy, L., et al., Pre-eclampsia and risk of cardiovascular disease and cancer 320 \nin later life: systematic review and meta-analysis. BMJ, 2007. 335(7627): p. 974. 321 \n4. Roberts, J.M., et al., Subtypes of Preeclampsia: Recognition and Determining 322 \nClinical Usefulness. Hypertension, 2021. 77(5): p. 1430-1441. 323 \n5. Huang, K.H., et al., Prediction of pre-eclampsia complicated by fetal growth 324 \nrestriction and its perinatal outcome based on an artificial neural network model. 325 \nFront Physiol, 2022. 13: p. 992040. 326 \n6. Joshi, D., et al., Endothelial gamma-protocadherins inhibit KLF2 and KLF4 to 327 \npromote atherosclerosis. Nat Cardiovasc Res, 2024. 3(9): p. 1035-1048. 328 \n7. Hsieh, P.N., et al., Aging and the Kruppel-like factors. Trends Cell Mol Biol, 2017. 329 \n12: p. 1-15. 330 \n8. Deng, H., et al., A KLF2-BMPER-Smad1/5 checkpoint regulates high fluid shear 331 \nstress-mediated artery remodeling. Nat Cardiovasc Res, 2024. 3(7): p. 785-798. 332 \n9. Hsieh, P.N., et al., A conserved KLF-autophagy pathway modulates nematode 333 \nlifespan and mammalian age-associated vascular dysfunction. Nat Commun, 334 \n2017. 8(1): p. 914. 335 \n10. Dykxhoorn, D.M., et al., MicroRNA-29c-3p and -126a Contribute to the 336 \nDecreased Angiogenic Potential of Aging Endothelial Progenitor Cells. Int J Mol 337 \nSci, 2025. 26(9). 338 \n11. Samak, M., et al., Dysregulation of Kruppel-like Factor 2 and Myocyte Enhancer 339 \nFactor 2D Drive Cardiac Microvascular Inflammation and Dysfunction in 340 \nDiabetes. Int J Mol Sci, 2023. 24(3). 341 \n12. Thakar, S., et al., Intermittent High Glucose Elevates Nuclear Localization of 342 \nEZH2 to Cause H3K27me3-Dependent Repression of KLF2 Leading to 343 \nEndothelial Inflammation. Cells, 2021. 10(10). 344 \n13. Zhong, F., et al., Reduced Kruppel-like factor 2 expression may aggravate the 345 \nendothelial injury of diabetic nephropathy. Kidney Int, 2015. 87(2): p. 382-95. 346 \n14. Zhang, H., Z. Chen, and X. Wang, Differentiated serum levels of Kruppel-like 347 \nfactors 2 and 4, sP-selectin, and sE-selectin in patients with gestational diabetes 348 \nmellitus. Gynecol Endocrinol, 2022. 38(12): p. 1121-1124. 349 \n15. Gao, J., et al., Endothelial Kruppel-like factor 2/4: Regulation and function in 350 \ncardiovascular diseases. Cell Signal, 2025. 130: p. 111699. 351 \n16. Boon, R.A., et al., Kruppel-like factor 2 improves neovascularization capacity of 352 \naged proangiogenic cells. Eur Heart J, 2011. 32(3): p. 371-7. 353 \n17. Sangwung, P., et al., KLF2 and KLF4 control endothelial identity and vascular 354 \nintegrity. JCI Insight, 2017. 2(4): p. e91700. 355 \n18. Hamik, A., et al., Kruppel-like factor 4 regulates endothelial inflammation. J Biol 356 \nChem, 2007. 282(18): p. 13769-79. 357 \n19. Ng, K.W., et al., Biomarkers and point of care screening approaches for the 358 \nmanagement of preeclampsia. Commun Med (Lond), 2024. 4(1): p. 208. 359 \n20. MacDonald, T.M., et al., Clinical tools and biomarkers to predict preeclampsia. 360 \nEBioMedicine, 2022. 75: p. 103780. 361 \n21. Slehria, T., et al., A Gap in Knowledge-Sudden Death and Preeclampsia. Am J 362 \nCardiol, 2023. 202: p. 199-200. 363 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n22. Atluri, N., et al., Challenges to diagnosing and managing preeclampsia in a low-364 \nresource setting: A qualitative study of obstetric provider perspectives from 365 \nGhana. PLOS Glob Public Health, 2023. 3(5): p. e0001790. 366 \n23. Akduman, A.T., et al., Effect of tamoxifen on ovarian reserve: A randomized 367 \ncontrolled assessor-blind trial in a mouse model. J Turk Ger Gynecol Assoc, 368 \n2014. 15(4): p. 228-32. 369 \n24. Kivioja, A., et al., Increased Risk of Preeclampsia in Women With a Genetic 370 \nPredisposition to Elevated Blood Pressure. Hypertension, 2022. 79(9): p. 2008-371 \n2015. 372 \n25. Sones, J.L. and R.L. Davisson, Preeclampsia, of mice and women. Physiol 373 \nGenomics, 2016. 48(8): p. 565-72. 374 \n26. Dupont, V., et al., Impaired renal reserve contributes to preeclampsia via the 375 \nkynurenine and soluble fms-like tyrosine kinase 1 pathway. J Clin Invest, 2022. 376 \n132(20). 377 \n27. Huang, X., et al., sFlt-1-enriched exosomes induced endothelial cell dysfunction 378 \nand a preeclampsia-like phenotype in mice. Cytokine, 2023. 166: p. 156190. 379 \n28. Yang, C., et al., Long-Term Impacts of Preeclampsia on the Cardiovascular 380 \nSystem of Mother and Offspring. Hypertension, 2023. 80(9): p. 1821-1833. 381 \n29. Wojczakowski, W., et al., Preeclampsia and Cardiovascular Risk for Offspring. J 382 \nClin Med, 2021. 10(14). 383 \n30. Turbeville, H.R. and J.M. Sasser, Preeclampsia beyond pregnancy: long-term 384 \nconsequences for mother and child. Am J Physiol Renal Physiol, 2020. 318(6): p. 385 \nF1315-F1326. 386 \n31. Brown, M.C., et al., Cardiovascular disease risk in women with pre-eclampsia: 387 \nsystematic review and meta-analysis. Eur J Epidemiol, 2013. 28(1): p. 1-19. 388 \n32. Davisson, R.L., et al., Discovery of a spontaneous genetic mouse model of 389 \npreeclampsia. Hypertension, 2002. 39(2 Pt 2): p. 337-42. 390 \n33. Ssengonzi, R., et al., Endothelial Nitric Oxide synthase (eNOS) in Preeclampsia: 391 \nAn Update. J Pregnancy Child Health, 2024. 6. 392 \n34. Kim, S., et al., Alleviation of preeclampsia-like symptoms through PlGF and 393 \neNOS regulation by hypoxia- and NF-kappaB-responsive miR-214-3p deletion. 394 \nExp Mol Med, 2024. 56(6): p. 1388-1400. 395 \n35. Zhou, Z., et al., Targeted polyelectrolyte complex micelles treat vascular 396 \ncomplications in vivo. Proc Natl Acad Sci U S A, 2021. 118(50). 397 \n 398 \n  399 \n 400 \nFigure Legends 401 \nFigure 1. KLF2/4 are involved in hypertension and sFLT1 elevation. A CVD 402 \nKnowledge Portal (CVDkP) analysis revealed SNPs within KLF2 linked to hypertension, 403 \na major causal factor in the development of PE. B This variant is identified in the GTEx 404 \ndatabase as an eQTL for KLF2 in whole blood. C-F siRNA-mediated KD of KLF2 (B-C) 405 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nor 4 (D-E) in HUVECs significantly increased the sFLT1/FLT1 ratio in comparison to 406 \ncontrol siRNA. We did not observe a change in total FLT1 levels, indicating that KLF2/4 407 \nnegatively regulates the splicing of sFLT1 and not the overall transcript levels of FLT1.  408 \nFigure 2. KLF4iECKO mice develop early, severe preeclamptic pregnancies. A mean 409 \narterial pressure (MAP) was tracked using tail-cuff telemetry over the course of KLF4f/f 410 \nor KLF4iECKO pregnancies. Dams lacking EC KLF4 showed significant hypertension in 411 \ncomparison to cre- controls for the duration of their pregnancy. Prior to pregnancy, mice 412 \nshowed no difference between genotypes (GD-1). Days covered by purple are training 413 \ndays. B Blood samples were collected at GD18.5 via cardiac puncture. Samples were 414 \nthen run at dilutions ranging from 1:5-1:20 to ensure values landed within the standard 415 \ncurve. 6 B6 and 9 KLF4-/- dams were analyzed using R&D’s VEGFR1 ELISA kit. KLF4-416 \n/- mice had significantly higher sFLT1 in comparison to B6 controls. C To determine the 417 \nextent of kidney damage during KLF4iECKO preeclamptic pregnancies, we collected urine 418 \nprior to sacrificing on GD18.5 and calculated the albumin:creatine (ACR). KLF4iECKO 419 \ndams had significantly more proteinurea in comparison to B6 controls. D This was 420 \nfurther confirmed via kidney H&E staining, which showed severe occlusion of the 421 \nglomeruli in KLF4iECKO versus cre- controls. E-F RNAscope analysis of ECs in the aorta 422 \nshowed that KLF4iECKO dams had significantly more sFLT1 splicing in their aortic 423 \nendothelium in comparison to cre- animal controls.  424 \nFigure 3. Pups from KLF4iECKO Dams experience fetal growth restriction and 425 \ndecreased placental vascular area. A Following dissection from the yolk sac and 426 \nplacenta, pups were weighed as a group and the average pup weight was determined 427 \nfor each litter by dividing the total weight by the number of pups. Data points represent 428 \nlitters. Pups from KLF4iECKO pregnancies were significantly smaller in comparison to B6 429 \ncontrols. A representative pup comparison is pictured to the right of the quantification. B 430 \nIn addition to being smaller, litters from KLF4iECKO pregnancies had significantly fewer 431 \npups in comparison to B6 litters. C-D Placental vascular area was analyzed in two 432 \nways: firstly, via gross analysis of the underside (pup side) of the placenta where blood 433 \nperfusion can be used to distinguish between the vascular area (VA) and the total area 434 \n(TA) (C), and secondly via sectioning and comparing the dense labyrinth area to the 435 \ntotal placental area (D). Representative images with area tracings are shown for both 436 \nanalyses. Gross images were taken using an iPhone 8. Placenta section images were 437 \ntaken using a 20x air objective with tiling used to capture the entire section.  Data is 438 \nrepresented as per placenta (C) with at least 6 sections being analyzed per genotype in 439 \n3 independent staining (D). E RNAscope analysis with FN protein staining on placenta 440 \nsections reveals increased sFLT1 splicing and FN deposition in KLF4iECKO versus B6 441 \ncontrols. Representative images are shown with zoomed insets (Ei-ii) and 442 \nquantification. Data points are placentas that are an average of 3 stained sections, 443 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\nplacentas from at least 3 different litters were used for each genotype.  All data is shown 444 \nas the mean +/- SEM. 445 \nSupplemental Figure 1. Additional assays on hypertension and kidney damage. A-446 \nC percent coefficient of variation (%CV) calculated per session per individual to ensure 447 \nrobust measurements. For each animal and session, BP was calculated as the mean of 448 \n≥3 accepted tail-cuff cycles. Session stability was quantified using the coefficient of 449 \nvariation (CV = SD/mean × 100). Cycle-level outliers exceeding ±2 SD were excluded to 450 \nremove technical artifacts. Sessions demonstrating CV values within the expected 451 \nphysiological range (<10%) were considered stable. Days covered by purple are training 452 \ndays. D-F average diastolic (D) and systolic (E) and mean (F) blood pressure 453 \ncomparison between nonpregant B6, KLF4f/f and iECKO dams showing no difference 454 \nprior to pregnancy. G-H Time course of diastolic (G) and systolic (H) blood pressure 455 \nchanges during pregnancy in f/f or iECKO KLF4 animals. Days covered in purple are 456 \ntraining days. I-J Representative CD34 staining in KLF4iECKO and f/f controls with 457 \nquantification (I). 458 \nSupplemental Figure 2. RNAscope Schematic and Analysis. A confirmation of 459 \nKLF4iECKO following tamoxifen injection. Aortas were stained en face for KLF4, CD31 460 \nand DAPI to confirm iECKO. B Mouse FLT1 mRNA schematic is shown with RNAscope 461 \nprobe locations and the sFLT1 i13 splice site (red line). C RNAscope probe control 462 \nstaining in en face aorta pieces. Laser power and exposure for each repeat experiment 463 \nwas determined by control probe fluorescence before imaging experimental samples. D 464 \nSplit channel representative RNAscope en face aortas with FN protein staining and 465 \nsample calculation.   466 \nSupplement Figure 3. Placenta RNAscope validation using mouse positive and 467 \nnegative control probes. Placenta sections from 2-3 individual litters were used to 468 \nstain for positive and negative control probes. Imaging parameters (laser power, 469 \nexposure) for experimental samples were determined based on detection of signal only 470 \nin the positive control set, but not the negative control set, as represented here. 471 \nChannels HRP-C1 and –C2 were optimized in this experiment, we did not develop 472 \nHRP-C3 since that channel was taken by FN protein staining.  473 \n 474 \nFigures 475 \nFigure 1 476 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n 477 \nFigure 2 478 \n 479 \n 480 \nFigure 3 481 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n 482 \nSupplemental Figure 1 483 \n 484 \nSupplemental Figure 2 485 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n 486 \nSupplemental Figure 3 487 \n 488 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.30.715448doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}