Placental Vascular Defects and Embryonic Lethality Triggered by TFPIα Deficiency in Factor V Leiden Mice

Abstract

Background:

Tissue Factor Pathway Inhibitor (TFPI) inhibits the initiation of blood coagulation. TFPIα, the only alternatively spliced TFPI isoform in platelets, is abundant in placenta and uniquely inhibits prothrombinase (FXa-FVa). This inhibitory activity is reduced when prothrombinase is assembled with Factor V Leiden (FVL).

Methods:

Effects of TFPIα isoform (Tfpi^Δα) and platelet (Tfpi^fl; Pf4-Cre⁺) specific knockout alleles were characterized in FVL (F5^L) mice to examine the physiological effects of the TFPIα-FV interaction.

Results:

Genotype frequencies were assessed and revealed that Tfpi^+/Δα F5^L/L mice survive to adulthood. However, Tfpi^Δα homozygosity with even a single F5^L allele resulted in embryonic lethality during mid-gestation development regardless of maternal FVL status. In contrast, F5^L/L Tfpi^fl/fl Pf4-Cre⁺ mice were at expected frequencies at weaning, indicating that platelet TFPIα loss alone did not cause mid-gestation lethality in Tfpi^Δα/Δα F5^L mice. Histological analyses showed no fibrin deposition in embryonic or extraembryonic tissues but revealed placental vasculature defects in Tfpi^Δα/Δα F5^L genotypes. Treatment with the direct thrombin inhibitor dabigatran partially rescued the lethality and corrected placental defects, implicating excessive thrombin generation as a factor in Tfpi^Δα/Δα F5^L demise.

Conclusion:

These findings suggest that TFPIα and its inhibition of prothrombinase plays an important role in placental angiogenesis and embryonic survival.

Graphical Abstract

Introduction

TFPI is an alternatively spliced Kunitz-type serine protease inhibitor that is present in plasma¹, on endothelium and monocytes^2–4, and within platelets.^5–8 It is an essential anticoagulant protein⁹ that inhibits FXa, TF-FVIIa¹⁰ and FXa-FVa (prothrombinase)¹¹ at early stages of the coagulation cascade. TFPIα, the primary transcript of the TFPI gene, is evolutionarily conserved from humans to zebrafish¹² and is the only isoform expressed by megakaryocytes and present in platelets.⁶ The other TFPIβ and TFPIγ isoforms localize to distinct pools, with TFPIβ on the endothelium in therian mammals^12,13 and TFPIγ as the predominant isoform in murine plasma.^14,15 TFPIα and TFPIβ are also abundant in placenta.¹² In late-gestation rat placenta, Tfpi gene expression was localized to various trophoblast cell populations strategically positioning it to facilitate blood flow at the uterine-placental interface, and within the junctional and labyrinth zones of the chorioallantoic placenta.¹⁶

All TFPI isoforms inhibit the TF-FVIIa catalytic complex and FXa via the first (K1) and second (K2) Kunitz domains, respectively.^10,13,14,17 Global loss of the TFPI K1 domain in mice (Tfpi^tm1Gjb;Tfpi^−/−) results in embryonic lethality with ~40% dying at mid-gestation (E9.5-E14.5), and the remaining during late-gestation (E15.5-E18.5).^18–20 The Tfpi^−/− late-gestation embryos display cerebrovascular developmental defects that likely contribute to their demise.²¹ Interestingly, the cerebral vascular defects were not observed in Tfpi^−/− embryos lacking endogenous FV, strongly suggesting that regulation of FV procoagulant activity by TFPI is necessary for proper cerebrovascular development.²¹

TFPIα is distinct from the β and γ isoforms via an intimate structural and functional relationship with FV. TFPIα has a C-terminal region encoding a third Kunitz domain (K3), which binds to protein S,²² and a stretch of basic amino acids similar to the basic region in the FV B domain that is well conserved across mammalian species.¹¹ A high-affinity interaction between the TFPIα basic region and an acidic region in the B-domain of partially activated forms of FV mediates inhibition of nascent prothrombinase prior to thrombin generation.^11,23 The importance of this TFPIα-FV interaction is evident in the FV East Texas,^24–26 Amsterdam,²⁷,and Atlanta²⁸ bleeding disorders whereby complex formation with FV-short isoforms increases TFPIα plasma concentration 10- to 20-fold.

Factor V Leiden, an R506Q variant in the F5 gene (FVL; F5^L), is the most prevalent genetic risk factor for venous thromboembolism in Caucasians.^29–32 FVL prolongs prothrombinase activity due to its resistance to inactivation by activated protein C.³³ Mice homozygous for F5^L exhibit a mild prothrombotic phenotype mimicking the human disorder.³⁴ Interestingly, Tfpi^+/− F5^L/L mice die from perinatal thrombosis,^35,36 suggesting that TFPI is an important inhibitor of FVL prothrombotic activity. Consistent with this notion, individuals with FVL have reduced TFPIα-mediated prothrombinase inhibition with kinetic studies indicating that TFPIα is a 1.7-fold weaker inhibitor of prothrombinase assembled with FVL than with FV.³⁷ Furthermore, increased TFPI resistance was confirmed in plasma from both heterozygous and homozygous FVL individuals using an assay measuring FV susceptibility to inhibition by the TFPIα C-terminus.³⁸ Thus, TFPIα anticoagulant activity towards prothrombinase may be particularly important when coagulation is initiated by weak stimuli that are amplified when FVL is present.

Herein, TFPIΔα mice carrying a premature stop codon at the 5’ end of the TFPIα terminal exon³⁹ were used as an in vivo model to examine the isoform-specific role of TFPIα in regulating FVL prothrombotic activity. The TFPIΔα mouse model displays expected Mendelian genotype distribution, normal survival throughout adulthood, and ability to reproduce as homozygotes.³⁹ TFPI antigen and activity is normal in Tfpi^Δα/Δα plasma, but completely absent in platelets; consistent with unaltered TFPIγ expression but altered TFPIα expression. In these studies, generation of TFPIΔα mice expressing FVL revealed that Tfpi^Δα/Δα progeny heterozygous or homozygous for F5^L succumb to mid-gestation embryonic lethality with severe placental labyrinth zone defects.

Methods

Availability of Data:

The datasets and protocols generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Mice

Tfpi^Δα/Δα and Tfpi^fl/fl Pf4-Cre⁺ mice were previously generated.^39,40 F5^L/L mice were provided by Dr. Hardy Weiler (Versiti, Milwaukee, WI). All mice were backcrossed >8 generations to the C57BL/6J background. The F5^L allele was crossed onto the Tfpi^Δα genetic background to generate the TFPIΔα/FVL line. A female Tfpi^+/+ F5^L/L mouse from this line was crossed to a Tfpi^fl/fl Pf4-Cre⁺ male to generate platelet-specific TFPI knockout F5^L mice (pTFPIα-KO/FVL). Genotyping primers are listed in Supplemental Table S1.

RT-PCR and 3’-RACE

RNA extraction, cDNA synthesis, and RT-PCR were conducted on liver, placenta, and freshly isolated platelets using primers listed in Supplemental Table S2.¹⁹ Analysis of polyadenylated liver transcripts was performed using GSP1/GSP2 primers (Supplemental Table S3) and the 3´ RACE System (Thermo Fisher Scientific, Waltham, MA). Bidirectional Sanger sequencing was performed by Retrogen Inc.

Platelet TFPIα Activity Assays

Platelet lysate TFPIα activity was measured using TF-FVIIa-dependent FX assays.^19,40

Histopathology

The uterine vasculature of both uterine horns at the cervical and ovarian ends was ligated, fixed for 24 hours in neutral buffered 10% formalin, and then followed by a midline sagittal bisection of each placentation site. Paraffin-embedded tissue was cut in 4.5 μm sections and evaluated with Gomori’s trichrome and immunohistological staining using Cytokeratin 8 (KRT8, clone: TROMA1; Millipore Sigma, St. Louis, MO), CD34 (clone SI16–01; Novus Biologicals; Centennial, CO), Endomucin (AF4666; R&D Systems, Minneapolis, MN), CD45 (AF114; R&D Systems), laminin (LAMA1, L9393; Millipore Sigma) or Alexa Fluor 647-conjugated fibrin (clone 59D8; Versiti) primary antibodies. Species-specific IgG comparable to the test antibodies served as controls. Immunohistochemistry slides were processed using ImmPRESS secondary polymers (MP-7405 and MP-7401; Vector Laboratories, Burlingame, CA) with NovaRed (#SK-4805, Vector Laboratories) detection, hematoxylin counterstain (H-3401, Vector Laboratories), and the Easy Scan imaging system (Motic, Hong Kong). Immunofluorescence slides were processed using goat anti-rabbit Alexa Fluor 647 (111–605-144, Jackson ImmunoResearch Labs, West Grove, PA), donkey anti-rabbit Alexa Fluor 750 (ab175728, Abcam, Waltham, MA), donkey anti-goat Alexa Fluor 750 (ab175745, Abcam) or donkey anti-rat Dylight 755 (SA510031, Thermo Fisher Scientific) and a Nikon Eclipse Ti2 inverted microscope with a DS-Ri2 high-speed color camera low-magnification 10×/0.45, 20×/0.75, 40×/0.95 or a high-magnification 60× oil/1.4 numerical aperture objectives (Nikon Instruments, Melville, NY). Images were pseudocolor processed with the Nikon NIS-Elements software platform and formatted in Adobe Photoshop 2023 and Adobe Illustrator 2023 (Adobe Systems Inc., San Jose, CA). ImageJ version 1.54i (NIH, Bethesda, MD) quantified KRT8, CD34, and fibrin immunostaining.

Dabigatran Diet and Direct Thrombin Inhibitor Assay

Mice were fed the 5L0D PicoLab Laboratory Rodent Diet (Dyets Inc., Bethlehem, PA) containing dabigatran etexilate (5 g/kg; Adooq Bioscience, Irvine, CA) and non-nutritive peanut flavoring and peanut butter (2 and 10 g/kg, respectively) or the non-supplemented 5LOD PicoLab control diet. Females were fed for 7 days before breeding in timed matings. At birth, the dabigatran containing diet was replaced with the control diet, and progeny counted daily until weaning. The BIOPHEN^™ DTI (Direct Thrombin Inhibitor) assay (Aniara Diagnostica, West Chester, OH) was conducted using maternal plasma obtained at mid-gestation or from 7-day diet-acclimated, female breeders.

Statistical analyses

Statistical analyses were performed in R ver3.6.3 and GraphPad Prism ver 10.2.1 (GraphPad Software, Boston, MA) with an alpha level of 0.05. Statistical differences among mouse progeny were determined using Fisher exact test (FET) or goodness-of-fit tests (χ2 or exact binomial tests (EBT)) using the method of small p-values. For quantitative assays, the Mann-Whitney test or Kruskal-Wallis test and post hoc analyses using the Wilcoxon signed-rank test were performed. Multiple comparison p-values were adjusted using the Benjamini and Hochberg false discovery rate (FDR) with significance reported as the adjusted q-value.

Study Approval

The Institutional Animal Care and Use Committee at Medical College of Wisconsin approved all mouse experiments.

Results

The Tfpi^Δα-F5^L genetic interaction produced complete lethality in Tfpi^Δα/Δα F5^L mice.

The F5^L allele was crossed onto the TFPIΔα genetic background to examine the physiological regulation of FVL procoagulant activity by TFPIα. Analyses of 318 weaning age (P21) progeny derived from Tfpi^+/Δα F5^+/L × Tfpi^+/Δα F5^+/L (dihybrid) crosses revealed discrepancies in Mendelian distributions (Table 1; χ²=106.9, df=8, p=2.2×10⁻¹⁶). Unlike the previously reported perinatal lethality observed in Tfpi^+/− F5^L/L mice^35,36, Tfpi^+/Δα F5^L/L mice were present at near expected Mendelian frequency (q_EBT=0.160; 95% CI: 6.5–13.2%). However, homozygosity of the Tfpi^Δα allele produced completely penetrant lethality when co-segregating with either F5^L/L (q_EBT=8.9×10⁻⁹; 95% CI: 0.0–1.2%) or F5^+/L (q_EBT=2.0×10⁻¹⁵; 95% CI: 0.0–1.2%) genotypes indicating that expression of TFPIα is required for survival of heterozygous or homozygous FVL mice.

Table 1.

Progeny distribution from Tfpi^+/Δα F5^+/L × Tfpi^+/Δα F5^+/L crosses at weaning

Genotype	No. Obs	%Obs	%Exp	95% CIa	q-valuea
Tfpi+/+ F5+/+	33	10.4	6.25	7.25–14.26	0.0089b
Tfpi+/+ F5+/L	48	15.1	12.5	11.34–19.51	0.20
Tfpi+/+ F5L/L	13	4.1	6.25	2.19–6.89	0.17
Tfpi+/Δα F5+/+	71	22.3	12.5	17.87–27.31	<0.001b
Tfpi+/Δα F5+/L	104	32.7	25.0	27.57–38.16	0.0051b
Tfpi+/Δα F5L/L	30	9.4	12.5	6.46–13.19	0.16
TfpiΔα/Δα F5+/+	19	6.0	6.25	3.64–9.17	>0.99
TfpiΔα/Δα F5+/L	0	0.0	12.5	0.00–1.15	<0.001b
TfpiΔα/Δα F5L/L	0	0.0	6.25	0.00–1.15	<0.001b

^aExact binomial test of the genotypic proportions (observed (Obs) and expected (Exp)) and false discovery rate-adjusted probabilities (q-value). 95% confidence intervals (CI) represent the estimated range of observed genotype proportions based on binomial distribution. CIs are not FDR-adjusted.

^bSignificant.

Tfpi^Δα/Δα F5^L embryos died during mid-gestation.

Dihybrid timed matings were performed to examine whether the Tfpi^Δα/Δα F5^L genotypes are embryonic lethal and, if so, whether the lethality predominantly clusters within mid- or late-gestational stages (Table 2 and Supplemental Table S4). There were discrepancies in Mendelian distributions among 83 progenies from 13 litters harvested at late-gestation between E13.5-E17.5 (χ²=19.4, df=8, p=0.013), as no Tfpi^Δα/Δα F5^L/L and only one Tfpi^Δα/Δα F5^+/L embryo were identified (Table 2). At mid-gestation however, normal frequencies in genotype distribution were observed among 64 progeny from 8 litters harvested between E9.5-E11.5 (p_FET=0.8663; Table 2). No hemorrhage was present in the Tfpi^Δα/Δα F5^L conceptuses which displayed embryo and yolk sac pallor (Supplemental Figure S1), a phenotype present in many mid-gestation lethal mouse mutants.⁴¹ Therefore, the Tfpi^Δα/Δα progeny heterozygous or homozygous for the FVL allele succumb to intrauterine lethality during mid-gestation.

Table 2.

Genotypic distribution of embryonic progeny derived from dihybrid crosses at mid- and late-gestation^a

Genotype	%Exp	Mid-gestation (E9.5-E11.5)			Late-gestation (E13.5-E17.5)
		No. Obs (%Obs)	95% CIb	q-valueb	No. Obs (%Obs)	95% CIb	q-valueb
Tfpi+/+ F5+/+	6.25	2 (3)	0.38–10.84	0.66	6 (7)	2.70–15.07	0.83
Tfpi+/+ F5+/L	12.5	7 (11)	4.51–21.25	0.85	11 (13)	6.81–22.48	0.87
Tfpi+/+ F5L/L	6.25	7 (11)	4.51–21.25	0.66	4 (5)	1.33–11.88	0.87
Tfpi+/Δα F5+/+	12.5	11 (17)	8.90–28.68	0.66	16 (19)	11.44–29.41	0.20
Tfpi+/Δα F5+/L	25.0	14 (22)	12.51–33.97	0.75	26 (31)	21.59–42.44	0.46
Tfpi+/Δα F5L/L	12.5	11 (17)	8.90–28.68	0.66	12 (15)	7.70–23.89	0.83
TfpiΔα/Δα F5+/+	6.25	2 (3)	0.38–10.84	0.66	7 (8)	3.46–16.61	0.66
TfpiΔα/Δα F5+/L	12.5	5 (8)	2.59–17.30	0.66	1 (1)	0.03–6.53	0.0028c
TfpiΔα/Δα F5L/L	6.25	5 (8)	2.59–17.30	0.75	0 (0)	0.00–4.35	0.045c

^aA priori power analysis for χ² goodness-of-fit estimated 82.7 observations needed for detecting complete lethality in the Tfpi^Δα/Δα F5^L genotypes from dihybrid crosses (df=8, w=0.48, 90% power, α=0.05).

^bExact binomial test of genotypic proportions (observed (Obs) and expected (Exp)) and false discovery rate-adjusted probabilities (q-value). 95% confidence intervals (CI) represent the estimated range of observed genotype proportions based on binomial distribution. CIs are not FDR-adjusted.

^cSignificant.

Maternal FVL was not required for mid-gestation lethality.

Embryonic development is dependent on maternal-fetal interactions. Tfpi^Δα/Δα F5^+/+ female × Tfpi^+/Δα F5^L/L male crosses were performed to examine whether maternal FVL contributed to the mid-gestation lethality (Table 3). Only two genotypes are produced from this cross: the viable Tfpi^+/Δα F5^+/L and the lethal Tfpi^Δα/Δα F5^+/L. No lethal Tfpi^Δα/Δα F5^+/L progeny were present among 16 examined at weaning. In timed mating studies, only 1 of 10 E15.5 embryos was Tfpi^Δα/Δα F5^+/L, while 5 of 15 were present at E10.5. These data indicated that maternal FVL is not required for lethality in this model and suggested that embryonic/extraembryonic-derived FVL is the major driver of mid-gestation lethality in Tfpi^Δα/Δα progeny.

Table 3.

Progeny distributions from Tfpi^Δα/Δα F5^+/+ × Tfpi^+/Δα F5^L/L crosses

Age	No. Obs				95% CIb	p-valueb
	Tfpi+/Δα F5+/L	TfpiΔα/Δα F5+/L	Resorbinga	Total
P21	16	0	---	16	0.0–20.6	<0.001c
E15.5	6	1	3	10	0.3–44.5	0.022c
E10.5	8	5	2	15	11.8–61.6	0.30

^aGenotype unknown.

^bExact binomial test of the Tfpi^Δα/Δα F5^+/L proportion (observed (Obs)) at each age (50% expected). 95% confidence intervals (CI) represent the estimated range of observed genotype proportions based on binomial distribution.

^cSignificant.

Specific absence of fetal platelet TFPIα (pTFPIα) did not produce lethality in FVL mice.

TFPIα is the only TFPI isoform present in human and mouse platelets and is completely absent in the TFPIΔα mouse model.³⁹ A conditional mutant Tfpi model⁴⁰ was employed to examine if platelet TFPIα contributed to the lethal phenotype. Platelet TFPIα activity was 1.32 ± 0.43 fmoles per mg total protein in Tfpi^fl/fl Pf4-Cre⁺ mice versus 7.24 ± 1.3 fmoles per mg total platelet protein in Tfpi^fl/fl Pf4-Cre⁰ mice (p_MW=0.0002; Supplemental Figure S2A). A Tfpi^+/+ F5^L/L female from the TFPIΔα/FVL line was mated to a Tfpi^fl/fl Pf4-Cre⁺ male⁴⁰ and resulting progeny intercrossed for three subsequent generations to generate the pTFPIα-KO/FVL line (Supplemental Figure S2B). F5^L/L Tfpi^fl/fl Pf4-Cre⁰ female × F5^L/L Tfpi^fl/fl Pf4-Cre⁺ male crosses generated 22 out of 55 mice positive for the transgene at weaning (χ²=2.2, df=1, p=0.138). Thus, F5^L/L Tfpi^fl/fl Pf4-Cre⁺ mice were viable and fertile, indicating that absence of TFPIα in fetal platelets alone was not responsible for the mid-gestation lethality observed in the TFPIΔα/FVL line.

TFPIΔα is a pan-cellular, TFPIα-specific, knockout mouse model.

The differences in TFPIΔα/FVL and pTFPIα-KO/FVL lethality led us to perform additional studies to rigorously assess tissue-specific effects of the targeting mutation on Tfpi transcript production and the mechanism for loss of platelet TFPIα expression in TFPIΔα mice. Initial RT-PCR analysis verified that Tfpiβ and Tfpiγ transcript processing was unaltered in mutant homozygotes (Figure 1). Surprisingly, the predicted TFPIΔα (TFPI-K1K2K3-stop)-encoding transcript was not present in Tfpi^Δα/Δα platelets, placenta, or liver. Analyses of the expressed transcripts revealed that presence of the Tfpi^Δα allele resulted in global loss of canonical Tfpiα terminal exon splicing (Supplemental Figure 3). Instead, three alternative polyadenylation (APA) transcript variants (Variants 1 – 3) are expressed that are predicted to either be degraded or encode TFPI-K1K2K3 followed by an alternative C-terminal region lacking the amino acids that bind FV. Thus, the TFPIΔα model is a pan-cellular, TFPIα-specific knockout mouse that retains TFPIβ and TFPIγ expression.

Figure 1. TFPIΔα is a pancellular, TFPIα-specific, knockout mouse model.

(A) Schematic of the mouse Tfpi gene and position of the Tfpi^Δα targeting site. Positions of the start codon (ATG), Kunitz domains (K(1–3)), transcript variant-specific stop codons (Tfpi(β, γ, α, Δα)) and polyadenylation sites (pA) are indicated. The 5’-UTR variable region of non-coding exons is omitted. (B) The coding sequence spanning RT-PCR analysis of transcripts expressed in Tfpi^+/+ (+/+) and Tfpi^Δα/Δα (Δα/Δα) platelets, placenta, and liver. Placentas were derived from homozygous crosses of Tfpi^+/+ or Tfpi^Δα/Δα mice. Schematic displays the intended PCR products amplified using a common Tfpi forward and transcript-specific reverse primers (arrowheads; Supplemental Table S2) with the hashed lines indicating the intronic regions spliced out between the exons. L is a 1Kb+ ladder. NTC is a no template PCR control.

Tfpi^Δα/Δα F5^L placentas had defects in labyrinth zone formation.

Placental malformations are present in ~68% of knockout mouse lines that are lethal between E9.5-E14.5.⁴¹ Assessment of TFPIΔα/FVL mid-gestation placentation sites from dihybrid and Tfpi^Δα/Δα F5^+/+ female × Tfpi^+/Δα F5^L/L male crosses determined that the overall organization and total placental area were comparable among genotypes (Figure 2A–C), but histopathological abnormalities were prominent in Tfpi^Δα/Δα F5^L labyrinth zones at E10.5 (Supplemental Figure 4 and 5). Branching and expansion of CD34-positive fetal blood vessels into the labyrinth zone was severely impaired in lethal genotypes (Figure 2D and E). There also were stark differences in the trichorial interhemal membrane that separates fetal and maternal blood (Figure 2F and Supplemental Figure 6). The intricate network of fetal blood vessels (FBV) in tight juxtaposition to maternal blood spaces (MBS) observed in viable genotypes was absent in labyrinth zones of Tfpi^Δα/Δα F5^L conceptuses. Effective gas and nutrient exchange between mother and fetus depend on tight apposition of FBV and MBS and the lack thereof may have contributed to the demise of Tfpi^Δα/Δα F5^L conceptuses.

Figure 2. Tfpi^Δα/Δα F5^L placentas had defects in labyrinth zone formation.

(A) Schematic representation of the E10.5 mouse placentation site consisting of: 1) the outer maternal layer of uterine decidual cells and the maternal vasculature; 2) the middle junctional zone containing parietal trophoblast giant cells (p-TGCs) and spongiotrophoblast (SpT) cells; and 3) the inner labyrinth zone (Lab) consisting of a branched network of trophoblast that are bathed by maternal blood that flows through a series of channels, canals and sinusoids. Inset details the interhemal membrane that separates maternal and fetal blood comprising two layers of multinucleated syncytiotrophoblast (SynT-I and SynT-II) that surround the endothelium of the fetal blood vessels (FBV) and large nucleated sinusoidal trophoblast giant cells (S-TGC) that line the maternal blood spaces (MBS). (B) Immunohistochemistry (IHC) of cytokeratin 8 (KRT8) positive (red) trophoblast cells of the junctional zone and labyrinth at E10.5 (scale bar=300um). The general organization and distinct layers of the placenta were evident and intact in both viable and lethal genotypes. Junctional zone SpT and p-TGC appeared normal. (C) Quantification of total placental area. Areas were demarcated by KRT8 positive junctional zone and labyrinth trophoblast cells and were not significantly different averaging 1.7×10⁶ ± 4.5×10⁵ μm² across embryonic and maternal genotypes. Points indicate the placental area in μm² (y-axis) measured from each histological section for each conceptus genotype (x-axis). Black = derived from dihybrid cross, red = derived from Tfpi^Δα/Δα F5^+/+ female × Tfpi^+/Δα F5^L/L male cross. Number of measured placental areas for each embryonic genotype: Tfpi^+/Δα F5^+/+ (N=2), Tfpi^+/Δα F5^+/L (N=3), Tfpi^+/Δα F5^L/L (N=3), Tfpi^Δα/Δα F5^+/L (N=5) and Tfpi^Δα/Δα F5^L/L (N=4). Data from the three viable genotypes are consolidated. (D) CD34 IHC (red) detecting fetal endothelium at the level of the labyrinth zone. Viable genotypes displayed extensive branching of FBV into the labyrinth zone. In Tfpi^Δα/Δα F5^L/L and F5^+/L placenta, initial penetration of angiogenic FBV was apparent, but further expansion and branching appeared arrested or severely stunted amongst densely packed chorionic trophoblast cells. Embryonic and maternal (cross type) genotypes are indicated. Scale bar=100μm. (E) Quantification of CD34 IHC from the same cohort of placentas in (C) and plotted in the same manner. Data is displayed as the percent of the total placental area positive for CD34 immunolabeling measured in pixels. (F) Dual immunofluorescence imaging of viable Tfpi^+/Δα F5^L/L and lethal Tfpi^Δα/Δα F5^L/L labyrinth zones using KRT8 (green) and CD34 (red) with DAPI nuclear counterstain (blue). Viable genotypes displayed thin, elongated layers of Syn-T cells and S-TGC separating the MBS from the FBV that were lined by flat, thin endothelial cells and contained nucleated erythrocytes. In lethal genotypes, cell fusion was evident in trophoblast cells separating the maternal and fetal blood supply, but they lacked the thin, elongated structure of differentiated Syn-T populations. S-TGC appear unaffected and were present lining the MBS. Defects in FBV were apparent with lumens being nearly indiscernible amongst the thick, disarrayed layers of endothelial cells. Lower 40x magnification is shown in merged image on left (scale bar=50μm) with hashed boxes indicating position displayed in the higher 60x magnification images shown on right (scale bar=10μm). In the merged 60x image on far right, M = maternal blood space; FB = fetal blood; arrow = S-TGC; arrowhead = syncytiotrophoblast layers; and * = evidence of cell fusion.

Dabigatran, a direct thrombin inhibitor, attenuates the Tfpi^Δα/Δα F5^L lethal phenotype.

The role of excessive thrombin activity as an underlying mechanism for the mid-gestation lethality observed in Tfpi^Δα/Δα F5^L progeny was examined by treating pregnant dams with dabigatran. Dihybrid crosses were performed while mating pairs were on a laboratory diet containing dabigatran (5 g/kg). Circulating dabigatran plasma levels in adult females were approximately 270.4±165.5 ng/mL (n=6) and were undetectable in adult females receiving the control diet (n=3). Of 84 total pups, six were Tfpi^Δα/Δα F5^+/L with four surviving to weaning (Figure 3A and Supplemental Table S5). No Tfpi^Δα/Δα F5^L/L pups were born, but less than three would be expected in this cohort assuming the ~55% rescue penetrance observed with Tfpi^Δα/Δα F5^+/L. It is also possible that the amount of thrombin generated in Tfpi^Δα/Δα F5^L/L pups was greater than that in Tfpi^Δα/Δα F5^+/L pups requiring a higher dose of dabigatran to achieve rescue. The two male and two female Tfpi^Δα/Δα F5^+/L rescued mice survived >180 days on a standard laboratory diet and were fertile, successfully nurturing two or more litters to weaning when crossed to double heterozygotes. Assessment of placentation sites from dabigatran-treated female timed matings at E10.5 and E11.5 revealed comparable and normal labyrinth zone vasculature and trichorial interhemal membrane development between Tfpi^Δα/Δα F5^L genotypes and their littermates (Figure 3B–D). These findings indicate that the vascular impairment observed in the Tfpi^Δα/Δα F5^L labyrinth zones is caused by excessive thrombin activity, suggesting that TFPIα, and its inhibition of prothrombinase, play a role in embryonic survival of FVL mice.

Figure 3. The direct thrombin inhibitor dabigatran suppresses the Tfpi^Δα/Δα F5^L lethal phenotype.

(A) Table showing the genotypic ratios of progeny obtained at birth and at weaning from dihybrid crosses fed a diet containing dabigatran (5mg/kg). (B) Table showing the number of observed conceptus genotypes from dabigatran-fed timed matings obtained at E10.5 and E11.5 using dihybrid crosses or at E11.5(*) from crossing Tfpi^+/Δα F5^+/L females to Tfpi^Δα/Δα F5^+/L males. In (*), 25% and 12.5% of zygotes are expected to carry Tfpi^Δα/Δα F5^+/L and Tfpi^Δα/Δα F5^L/L genotypes respectively. (C and D) Dual immunofluorescence imaging of labyrinth zones from viable, Tfpi^+/Δα F5^L/L and lethal Tfpi^Δα/Δα F5^L/L conceptuses derived from dabigatran-fed timed matings using cytokeratin 8 (KRT8, green) and CD34 (red) with DAPI nuclear counterstain (blue). In (C), upper panel is 20x magnification (scale bar=100μm) of placentation sites harvested at E10.5 with hashed boxes indicating position displayed in the higher 60x magnification images shown below (scale bar=10μm). In (D), upper panel is 10x magnification (scale bar=100μm) of placentation sites harvested at E11.5 with hashed boxes indicating position displayed in the higher 60x magnification images shown below (scale bar=10μm). M = maternal blood space; FB = fetal blood; arrow = S-TGC; arrowhead = syncytiotrophoblast layers.

Labyrinth zone defects were not associated with excess fibrin deposition.

The role of excessive thrombin activity in the Tfpi^Δα/Δα F5^+/L conceptuses suggested that excessive placental fibrin deposition may contribute to fetal demise. Maternal fibrin deposition at the placental-decidual interface stabilizes uteroplacental attachment and is required for successful pregnancy.⁴² Consistently, fibrin deposits were found at this location in both lethal and non-lethal genotypes (Supplemental Figure S4A). However, notable fibrin deposits were not present in embryonic or extraembryonic fetal vasculature within the three Tfpi^Δα/Δα F5^+/L, five Tfpi^Δα/Δα F5^L/L and eight viable genotypes examined. Only two conceptuses (a viable Tfpi^+/Δα F5^+/L and a lethal Tfpi^Δα/Δα F5^+/L) had fibrin deposits within MBS (661.7 and 407.5 microns respectively) closely associated with the labyrinth zone and located near the apex of the fetal vasculature (Supplemental Figure S4B). However, it is unlikely that they were large or frequent enough to have impaired labyrinth zone vasculature development. Therefore, excess fibrin deposition unlikely contributed to demise of Tfpi^Δα/Δα F5^L conceptuses.

Discussion

TFPIα modulates FV procoagulant activity by inhibiting nascent forms of prothrombinase that assemble before thrombin is generated. The interaction between TFPIα and FV occurs at a physiologically important regulatory point of the blood coagulation cascade. Bleeding disorders, such as FV east Texas, and clotting disorders, such as FVL, occur when the TFPIα-FV interaction is imbalanced. Here, we investigated the TFPIα-FV interaction in vivo using the TFPIΔα/FVL mouse model. Tfpi^Δα/Δα mice represent a global TFPIα knockout model. They are viable, fertile and lack an overall prothrombotic phenotype likely due to continued expression of TFPIβ and TFPIγ. However, when crossed to FVL, Tfpi^Δα/Δα mice carrying only a single F5^L allele succumbed to intrauterine lethality during mid-gestation. Thus, successful murine gestation requires dampening of FVL procoagulant activity by TFPIα. A diet containing dabigatran, a direct thrombin inhibitor, revealed that an underlying mechanism for the embryonic lethality was excess thrombin generation in the placenta that occurred independent of the presence of maternal FVL or the loss of TFPIα from fetal platelets. This excess thrombin produced defects within the placental labyrinth zone and associated vasculature, which were likely the cause of the mid-gestational embryonic demise. These findings indicate that TFPIα has an essential role in regulating FVL procoagulant activity during embryonic development that extends beyond its function as an inhibitor of blood coagulation to a function as a modulator of vascular development, a role not compensated by TFPIβ or TFPIγ isoforms or by other endogenous anticoagulant proteins such as thrombomodulin and antithrombin.

The nearly complete mid-gestation lethality of Tfpi^Δα/Δα F5^L embryos was a striking phenotype that we characterized using a step-by-step approach focusing on what could be learned about the TFPIα-FV interaction from this in vivo model. First, the possibility that maternal FVL contributed to embryonic demise was explored. Human case-control studies suggest that maternal FVL carrier status may double the risk of miscarriage and fetal demise,⁴³ but most prospective studies find no correlation between inherited thrombophilias and adverse pregnancy outcomes.⁴⁴ When analyzing progeny derived from crosses using Tfpi^Δα/Δα female mice expressing wild-type FV, the Tfpi^Δα/Δα offspring heterozygous for FVL continued to experience intrauterine lethality at mid-gestation. This finding suggested that the major source of FVL driving lethality was within the fetal circulation, which is derived predominantly from the liver primordium and yolk sac at mid-gestation.^45,46

Second, the possibility that fetal platelet TFPIα (pTFPIα) deficiency was responsible for embryonic death was explored. TFPIα is the only isoform found in platelets, which begin to circulate in the murine embryonic vasculature by ~E10.5.⁴⁷ The complete survival of F5^L/L Tfpi^fl/fl Pf4-Cre⁺ mice indicated that Tfpi^Δα/Δα F5^L embryonic lethality was not solely mediated by the absence of pTFPIα but involved a broader deficiency of TFPIα across different cell types, including trophoblast cells, which are specialized extraembryonic cells of the placenta. This distinction underlines the multifaceted roles of TFPI isoforms in different cellular contexts, and its essential contributions, not only in maintaining hemostatic balance, but also during critical stages of development.

Third, the possibility that placental defects contributed to embryonic demise was explored. The placenta is essential for fetal survival and growth, acting as the interface between maternal and fetal blood flow, enabling metabolic and gas exchange, and eliminating fetal waste. From analyses of hundreds of knockout mouse lines, it has become clear that a developmental bottleneck exists at mid-gestation with mutants dying between E9.5 – E14.5 often having severe placental malformations.⁴¹ TFPIα is highly expressed in trophoblast cells and placental defects were prominent within the labyrinth zone at E10.5 in the lethal Tfpi^Δα/Δα F5^L genotypes. By E10.5, three compartments within the placentation site begin to take shape: 1) an outer layer of maternal decidual cells and associated vasculature; 2) a centrally located junctional zone containing trophoblast giant cells and spongiotrophoblast cells; and 3) an inner labyrinth zone proximal to the developing fetus, which is designed for nutrient and gas exchange. This latter compartment is bathed by maternal blood and is composed of three layers of trophoblast cells and an inner core of stromal cells and blood vessels derived from the extraembryonic mesoderm of the allantois (hemotrichorial). Development of the labyrinth zone begins shortly after chorioallantoic fusion at ~E8.5 when trophoblast cells of the chorionic plate begin to fold inward and create a path for the invagination of fetal blood vessels from the allantois, a process driven by angiogenesis. During this time, chorionic trophoblast cells differentiate into three labyrinth zone cell types: mononuclear trophoblast cells that line maternal blood sinuses and two multinucleated syncytiotrophoblast (formed by trophoblast cell fusion) that surround fetal capillary endothelium. Together, trophoblast and fetal vasculature create extensively branched network within the labyrinth zone. The data presented here suggest that mid-gestation placental defects arise from disrupted angiogenesis in the Tfpi^Δα/Δα F5^L labyrinth zone. This conclusion is further supported by documented angiogenic defects in the brains of late gestation TFPI-K1 knockout embryos.²¹ Notably, the TFPIα C-terminus has been shown to inhibit angiogenesis in the adult hind limb ischemia model.⁴⁸ While mechanisms of angiogenesis differ between adult pathological contexts and placental vascular development,⁴⁹ it is noteworthy that the loss of TFPIα leads to labyrinth angiogenesis defects only in the presence of FVL.

Fourth, the role of thrombin in embryonic demise was explored. Thrombin is long known to contribute to the activation of angiogenesis in endothelial cells by facilitating their detachment and migration,⁵⁰ potentiating VEGF-induced activity,⁵¹ and upregulating α_Vβ₃ integrin expression.⁵² Notably, these proangiogenic actions are receptor-mediated, independent of fibrin production, and dose-dependent, with higher thrombin concentrations repressing these effects.⁵³ The near complete mid-gestation lethality associated with the combined loss of TFPIα with FVL pointed to excessive thrombin generation. This hypothesis was confirmed as lethal phenotypes were partially rescued by dabigatran, a direct thrombin inhibitor that crosses the placenta. Moreover, dabigatran treatment corrected the placental defects, which is consistent with these defects contributing to the fetal demise. Despite these convincing data showing excessive thrombin contributes to death, the placentas did not have excessive fibrin deposition. At mid-gestation, the higher sialic acid content in fetal fibrinogen results in fibrin clots that are softer, less branched, and degrade faster—properties well-suited for the dynamic and rapidly changing environment within the placenta.⁵⁴ In this mid-gestation lethality model, fibrin clots produced by excess thrombin are possibly being resolved too rapidly by the fibrinolytic system to detect or alternatively, the thrombin-mediated effect may be independent from its fibrin-generating function altogether.

It is possible that the mid-gestation lethality observed in the TFPIΔα/FVL mouse model simply represents an additive procoagulant effect independent of a distinct interaction between TFPIα and FVL. However, the ≥90% lethal penetrance in Tfpi^Δα/Δα F5^L embryos occurring in the presence of unaltered TFPIβ/γ isoforms and additional anticoagulants argues against this, especially when considering that 60% of TFPI K1 domain knockout (Tfpi^−/−)¹⁹ and 100% of Tfpi^+/− F5^L/L embryos^35,36 survive the mid-gestation bottleneck when on a congenic C57BL/6J background. Although we cannot definitively exclude the possibility of modifier gene interactions confounding comparison across these mouse lines, the large differences in the lethal penetrance strongly suggest a critical role for TFPIα and the TFPIα-FV interaction in thrombin-mediated placental angiogenesis, which is essential for maintaining and ensuring embryo survival.

Highlights.

• TFPIα-deficient mice crossed with Factor V Leiden (FVL) exhibit mid-gestation embryonic lethality.

• Platelet-specific TFPIα deletion does not cause lethality in FVL mice.

• Placental labyrinth defects and impaired angiogenesis are observed in lethal genotypes.

• Thrombin inhibition via dabigatran partially rescues both lethality and placental defects.

Non-standard Abbreviations and Acronyms

APA

Alternative Polyadenylation

ATG

Start Codon

CD34

Cluster of Differentiation 34 (fetal endothelial marker

DAPI

4′,6-diamidino-2-phenylindole

E

Embryonic day

FB

Fetal Blood

FBV

Fetal Blood Vessels

FVL

Factor V Leiden

FV

Factor V

FVIIa

Activated Factor VII

FXa

Activated Factor X

GSP

Gene-Specific Primer

H&E

Hematoxylin and Eosin

IHC

Immunohistochemistry

IF

Immunofluorescence

KRT8

Keratin 8

Lab

Labyrinth zone (placenta)

L

DNA ladder (1Kb+)

M

Maternal blood space or Male (context-dependent)

MBS

Maternal Blood Space

NTC

No Template Control

p-TGC

Parietal Trophoblast Giant Cell

q-value

FDR-adjusted p-value

RT-PCR

Reverse Transcriptase Polymerase Chain Reaction

S-TGC

Sinusoidal Trophoblast Giant Cell

SynT

Syncytiotrophoblast

TF

Tissue Factor

TFPI

Tissue Factor Pathway Inhibitor

TFPIα

TFPI alpha isoform

TFPIβ

TFPI beta isoform

TFPIγ

TFPI gamma isoform

Tfpi

Mouse gene encoding TFPI

UTR

Untranslated Region

VEGF

Vascular Endothelial Growth Factor