Pre-clinical evaluation of an enhanced-function factor VIII variant for durable hemophilia A gene therapy in male mice

Abstract

Durable factor VIII expression that normalizes hemostasis is an unrealized goal of hemophilia A adeno-associated virus-mediated gene therapy. Trials with initially normal factor VIII activity observed unexplained year-over-year declines in expression while others reported low-level, stable expression inadequate to restore normal hemostasis. Here we demonstrate that male mice recapitulate expression-level-dependent loss of factor VIII levels due to declines in vector copy number. We show that an enhanced function factor VIII variant (factor VIII-R336Q/R562Q), resistant to activated protein C-mediated inactivation, normalizes hemostasis at below-normal expression without evidence of prothrombotic risk in male hemophilia A mice. These data support that factor VIII-R336Q/R562Q may restore normal factor VIII function at low levels of expression to permit durability using low vector doses to minimize dose-dependent adeno-associated virus toxicities. This work informs the mechanism of factor VIII durability after gene transfer and supports that factor VIII-R336Q/R562Q may safely overcome current hemophilia A gene therapy limitations.

A major limitation in hemophilia A gene therapy is the progressive loss of factor VIII expression. Here, the authors demonstrate that low-level expression of a gain-of-function factor VIII variant may safely restore normal hemostasis and permit durability of expression.

Introduction

Hemophilia A (HA) is an X-linked congenital bleeding disorder due to a deficiency in (f)actor VIII cofactor function that is essential for enhancing FIX catalytic activity to amplify coagulation and prevent blood loss after vascular injury¹. Clinical manifestations include life-threatening hemorrhage and recurrent joint bleeding with disabling arthropathy² that are predicted by circulating plasma FVIII activity. Patients with severe HA (FVIII activity <1% of normal) have frequent and spontaneous hemorrhage that occurs less commonly in moderate HA patients (1- < 5% of normal). The mild HA (5- < 40% of normal) phenotype is heterogenous, but FVIII activity around 20% protects against spontaneous joint bleeds³. The current standard-of-care is recurrent intravenous FVIII infusion or subcutaneous administration of a FVIIIa-mimetic antibody, emicizumab^4,5. While effective, these therapies still require recurrent administration. Factor VIII gene therapy holds the promise of a one-time therapy to durably correct phenotype.

Current HA gene therapy efforts predominantly use adeno-associated virus (AAV)-mediated gene addition to exogenously express B-domain deleted FVIII in hepatocytes^6–9. First-generation clinical trial data outline a series of limitations that include, among others, an AAV dose-dependent anti-AAV cellular immune response resulting in loss of transgene expression in at least 1 trial⁶, inadequate FVIII expression to correct phenotype, and unexplained multi-year declines in FVIII expression^6,7,10–13. While the only licensed HA vector (valoctocogene roxaparvovec) initially achieved normal/near-normal FVIII activity, FVIII expression declined by almost half year-over-year such that median FVIII activity fell into the range of low mild to moderate HA within 5 years post-vector^7,10,12. In contrast, stable FVIII expression has been observed in the range of moderate or low mild HA^6,9,14, which is at or below the relative FVIII equivalency imparted by emicizumab prophylaxis¹⁵. These data demonstrate that AAV gene addition can impart durable FVIII expression, albeit possibly only at low levels of expression.

To overcome existing limitations, multiple novel approaches are in development for HA gene therapy¹⁶. One possibility is to express an enhanced function FVIII variant to normalize hemostasis at low plasma FVIII concentration to permit expression durability. In addition to overcoming HA gene therapy-specific limitations, the general advantage of an enhanced function protein variant is exemplified by the success of second-generation hemophilia B (HB) gene therapies that universally employ a gain-of-function FIX variant, FIX-Padua^17–19. In the absence of a naturally occurring enhanced function FVIII variant, rationally engineering FVIII to bypass mechanisms of activated FVIII (FVIIIa) regulation is an attractive approach^20–22. Factor VIIIa is inactivated by spontaneous A2-domain dissociation or by proteolytic cleavage by activated protein C (APC). We previously generated a FVIII variant with Arg to Gln mutations at the two established APC cleavage sites R336 and R562 (FVIII-R336Q/R562Q or FVIII-QQ)^20,23,24. Factor VIII-QQ in vitro specific-activity is analogous to wild-type FVIII and has normal A2-domain dissociation kinetics but is resistant to APC-mediated proteolytic inactivation. As a result, in recombinant protein studies, FVIII-QQ demonstrated 4–5-fold greater in vivo hemostatic function relative to FVIII-WT across multiple injury models in HA mice²⁰. Here we investigate the durability, efficacy and safety of using AAV-mediated expression of FVIII-QQ in mice as a second-generation approach to HA gene transfer to overcome current limitations.

Results

Durable FVIII expression is dependent on plasma FVIII antigen levels

Clinical trial data thus far suggest that sustained FVIII expression above the range of mild HA may not be feasible^6,7,10,12,14. To investigate the mechanism of durable FVIII expression post-AAV-mediated gene transfer, HA/C57BL/6 CD4-knockout (HA/CD4KO) mice were treated with a codon optimized AAV8 vector (CO1) to express B-domain deleted FVIII (FVIII-SQ²⁵, herein FVIII-WT) or FVIII-QQ at variable plasma FVIII concentrations. These experiments were also designed to assess if there were any differences in longitudinal pharmacokinetics of FVIII-WT and FVIII-QQ expression. HA/CD4KO mice were used because immune-competent C57BL/6 mice mount a robust antibody response to human FVIII expression²⁶. Animals were assigned to the mild HA (0.05 to <0.4 nM), normal (0.4 to <1.5 nM) and elevated FVIII ($\ge$1.5 to 3 nM) cohorts based on their steady-state week 8 plasma FVIII antigen (Fig. 1A, B and Supplementary Table 1). Factor VIII expression pharmacokinetics were determined up to 72 weeks. Plasma FVIII antigen at week 8 versus week 72 did not significantly differ in the mild HA and normal FVIII cohorts, but was significantly reduced at week 72 for the elevated FVIII-WT and FVIII-QQ cohorts (Fig. 1A, B). Combining FVIII-WT and FVIII-QQ data again demonstrated a significant reduction in week 72 versus week 8 antigen among the elevated FVIII cohort only (Fig. 1C, D). The data demonstrated that elevated FVIII negatively impacts durability for both FVIII-WT or FVIII-QQ expression. These findings are the first to document plasma FVIII antigen-dependent expression durability in mice and are qualitatively analogous to human clinical trial data^6,7,10,13,14.

Fig. 1. Durable FVIII expression is dependent on plasma FVIII antigen levels.

A, B HA/CD4KO mice were treated with AAV vector to express FVIII-WT (A) or FVIII-QQ (B) and assigned to the mild HA (0.05 to <0.4 nM FVIII, n = 5–14), normal FVIII (≥0.4 to 1.5 nM FVIII, n = 4–12), and elevated FVIII (≥1.5 to 3 nM FVIII, n = 4–10) cohorts according to plasma antigen values measured 8 weeks post-vector. Differences in plasma FVIII levels measured at 8- and 72-weeks post-vector are shown for each antigen cohort. Significance was determined by one-way mixed-effects analysis with Šídák’s multiple comparisons test. Individual data points with mean ± SEM are shown. (C) Plasma antigen at 8 versus 72 weeks post-vector for combined FVIII-WT and FVIII-QQ antigen cohorts (mild HA n = 11–21, normal FVIII n = 17–20, elevated FVIII n = 11–16). Differences in FVIII antigen at week 72 versus week 8 were determined by one-way mixed-effects analysis with Šídák’s multiple comparisons test. Individual data points with mean ± SEM are shown. D Plasma FVIII antigen (mean ± SEM) over time for mild HA, normal, and elevated FVIII cohorts (n = 21, 19, and 16, respectively). E Plasma FVIII antigen (mean ± SEM) measured 8 weeks post-vector did not differ between 8-week and 72-week study mice (mild HA n = 8–26, normal FVIII n = 11–27, elevated FVIII 6–19; multiple two-tailed unpaired t-tests with Welch correction for variance and Holm–Šídák correction for multiple comparisons). F Ratios of relative F8 mRNA expression versus VCN for 72-week livers did not differ among the treatment groups (mild HA n = 10, normal FVIII n = 16, elevated FVIII n = 11; Kruskal–Wallis test with Dunn’s multiple comparisons test). Individual data points with median ± 95% CI are shown. G–J Isolated liver VCN (G, I) and F8 mRNA expression (H, J) for 8-week mice (G, H) and 72-week mice (I, J). VCN (G) and F8 mRNA (H) significantly increased with cohort FVIII antigen level for 8-week mice (mild HA n = 8, normal FVIII n = 19, elevated FVIII n = 6). VCN (I) and F8 mRNA (J) did not differ between the normal and elevated FVIII cohorts for 72-week mice (mild HA n = 10, normal FVIII n = 16, elevated FVIII n = 11). Adjusted p-values were obtained via Kruskal–Wallis tests with Dunn’s multiple comparisons testing. Individual data points with median ± 95% CI are shown. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Source data are provided as a Source Data file.

Leading mechanistic hypotheses to explain declines in FVIII expression include gene silencing or vector properties that may preclude stable episome formation (e.g. cassette sizes and manufacturing platform^27–30). Single AAV8 vector preparations to express FVIII-QQ or FVIII-WT were used for all expression cohorts. The oversized cassette size (5 kb) demonstrated expected and comparable genomic heterogeneity between the FVIII-QQ and FVIII-WT vectors (Supplementary 1a, b). Significant reductions in plasma FVIII antigen were isolated to the elevated FVIII cohorts, suggesting that virion genomic heterogeneity or manufacturing platform^29,30 do not mechanistically account for the loss of FVIII expression. Additionally, declines in FVIII plasma antigen occurred >6 months post vector, which is beyond the window of stable episome formation demonstrated with similarly sized cassettes²⁸. Overall, these data do not implicate vector properties or unstable episome formation as the cause of concentration-dependent loss of FVIII expression.

To further investigate the mechanism of FVIII expression loss, livers were harvested from mice at week 72 and analyzed for vector copy number (VCN) and F8 mRNA. Additional HA/CD4KO mice were treated with the same AAV8 vector constructs to express FVIII-WT and FVIII-QQ in antigen-matched cohorts and livers were harvested at week 8 post vector. Week 8 FVIII antigen values for each FVIII cohort did not significantly differ among mice with livers harvested at 8 versus 72 weeks post vector (Fig. 1E). Inconsistent with a hypothesized mechanism of gene silencing^25,30,31, the ratio of F8 mRNA to VCN did not significantly differ among the FVIII antigen groups for week 72 (Fig. 1F). As expected^28,32, VCN and F8 mRNA in week 8 harvested livers significantly increased by plasma FVIII antigen cohort level (Fig. 1G, H), which is generally consistent with small cohort observations of HA human liver biopsies post AAV vector³³. In contrast with week 8 data, analysis of week 72 harvested livers demonstrated no significant difference in VCN and F8 mRNA between the normal and elevated FVIII cohorts (Fig. 1I, J), which is consistent with no significant difference in FVIII antigen between the normal and elevated FVIII cohorts at week 72 (Fig. 1C). These data support that decreases in FVIII expression in the elevated FVIII cohort are due to loss of VCN over time.

Given the established relationship between FVIII expression and induction of an endoplasmic reticulum (ER) stress response^32,34–36 that can lead to apoptosis of transduced cells and thereby loss of VCN, we investigated whether the reduction of FVIII plasma antigen in the elevated cohort was due to a cellular stress response. Markers of cellular stress, binding immunoglobulin protein (BiP) and C/EBP (CHOP) mRNA, were quantified from week 8 and 72 harvested livers. Among week 8 livers, BiP and CHOP values did not significantly differ from time-matched WT/CD4KO mice (Supplementary Fig. 1c), but CHOP expression was significantly higher in the elevated FVIII cohort relative to the mild HA and normal FVIII cohorts (Supplementary Fig. 1d). For week 72 analyzed livers, BiP was significantly elevated among all cohorts (Supplementary Fig. 1e) and CHOP was significantly elevated in the normal and elevated FVIII cohorts (Supplementary Fig. 1f) relative to time-matched WT/CD4KO livers. These studies established a positive association between plasma FVIII antigen and markers of cellular stress that suggest ER stress may be responsible for the loss of transduced cells and, thus, VCN. A competent immune system, not represented in these studies, modulates ER stress³⁷ and will be important to model in future investigations. Overall, we conclude that FVIII expression-level dependent decreases in plasma FVIII are due to decreased VCN over time. These studies provide a strong rationale for using a gain-of-function FVIII variant to restore FVIII function at low plasma FVIII antigen levels to permit durable expression.

AAV gene addition of FVIII-QQ has enhanced function without impacting expression

We previously demonstrated that bypassing APC regulation with FVIII-QQ had improved hemostatic potency in recombinant protein studies²⁰. To investigate if FVIII-QQ retained a hemostatic advantage over FVIII-WT in the setting of AAV-mediated gene transfer, HA/CD4KO mice were treated with AAV8 vectors to express FVIII-WT or FVIII-QQ. Animals were treated to achieve plasma FVIII-WT and FVIII-QQ antigen values of 0.05 nM (analogous to chromogenic-assay determined 5% of normal FVIII activity) that did not significantly differ (Fig. 2A). This expression-level was targeted because it is subtherapeutic for FVIII-WT in a tail clip assay²⁰ and would permit sensitivity to discriminate between enhanced FVIII-QQ function over FVIII-WT. Blood loss in FVIII-QQ expressing mice normalized to that of WT mice while mice expressing FVIII-WT did not. Additionally, blood loss in FVIII-QQ expressing mice was significantly lower than FVIII-WT mice (one-tailed Mann–Whitney test, p = 0.009). Thus, FVIII-QQ demonstrated greater hemostatic function than FVIII-WT after gene transfer (Fig. 2b).

Fig. 2. AAV gene addition of FVIII-QQ has improved in vivo hemostatic function relative to FVIII-WT in HA mice without impacting expression.

A Week 8 plasma FVIII antigen for AAV-treated HA/CD4KO mice expressing FVIII-QQ (n = 5) and FVIII-WT (n = 8) did not significantly differ by unpaired t test. B At 9 weeks post-vector, animals underwent tail clip assay. FVIII-QQ (n = 5) blood loss did not significantly differ from WT/CD4KO (n = 8) mice while both HA (n = 7) and FVIII-WT (n = 8) expressing mice had significantly greater blood loss than WT/CD4KO mice. Significance of blood loss versus WT/CD4KO was determined by Brown–Forsythe and Welch ANOVA tests with Dunnett’s T3 multiple comparisons test. C Adult HA/CD4KO mice were treated with AAV8 codon-optimized expression cassettes (CO1 and CO2) at 7 × 10¹¹ vg/kg to express FVIII-QQ (n = 4–9) or FVIII-WT (n = 4–8). Steady-state plasma FVIII antigen (D), liver VCN (E), and F8 mRNA (F) were analyzed for significant differences between FVIII-WT and FVIII-QQ expressing mice by unpaired t-tests. Individual data points with cohort mean ± SEM are shown. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. Source data are provided as a Source Data file.

To investigate if the introduction of the two amino acid changes of FVIII-QQ impacted expression, HA/CD4KO mice were treated with two AAV8 codon optimized constructs (CO1 or CO2) with the same promoter at a dose of 7 × 10¹¹ vg/kg to express FVIII-WT or FVIII-QQ (Fig. 2C) in the range of mild HA (0.05 to <0.4 nM), which would be targeted in clinical translation. There were no differences in week 8 steady-state FVIII-WT versus FVIII-QQ plasma concentration from either vector construct (Fig. 2D). Similarly, analysis of week 8-12 harvested livers demonstrated no significant difference in FVIII-WT and FVIII-QQ VCN (Fig. 2E) and generally no appreciable difference in F8 mRNA (Fig. 2F). Lastly, consistent with no apparent differences in induction of an ER stress response, but limited by a single time point analysis, week 8 BiP and CHOP mRNA values did not differ between FVIII-QQ and FVIII-WT expressing mice (Supplementary Fig. 2a, b). Overall, the data do not suggest differences in FVIII-QQ or FVIII-WT expression post-AAV-mediated gene transfer with codon optimized vectors.

APC contributes to in vivo physiologic FVIIIa regulation

While acute injury models identify differences in hemostatic efficacy, they do not comprehensively recapitulate spontaneous bleeding or thrombosis³⁸. To address this issue and further interrogate the role of APC in FVIIIa in vivo regulation, we backcrossed HA/CD4KO mice with FV-Leiden mice (FV^Q/Q) to generate HA/FV^Q/Q mice that were treated with AAV vector to express FVIII-WT or FVIII-QQ (Fig. 3A and Supplementary Table 1). FV-Leiden (FV-R506Q) imparts APC resistance and is the most common genetic risk factor for venous thrombosis^39,40. Mice homozygous for FV-Leiden exhibit a prothrombotic phenotype without diminished survival⁴¹. Since protein C deficiency is perinatally lethal in mice^42,43, we hypothesized that expression of FVIII-QQ in HA/FV^Q/Q mice would have a major impact on the APC pathway and diminish survival.

Fig. 3. Normal to elevated plasma FVIII-QQ expression in HA/FV^Q/Q mice reduces survival.

A HA/FV^Q/Q/CD4KO mice were treated with AAV vector to express FVIII-WT or FVIII-QQ. Animals were assigned to the mild HA (0.05 to <0.4 nM FVIII, n = 9–10) or combined normal FVIII (≥0.4 to 1.5 nM FVIII) and elevated FVIII (≥1.5–3 nM FVIII) cohort (n = 4–9) according to plasma FVIII antigen 8 weeks post-vector (mean ± SEM). B, C Survival outcomes for mock-injected HA/FV^Q/Q/CD4KO mice and AAV-treated HA/FV^Q/Q/CD4KO mice expressing FVIII-WT or FVIII-QQ in the ranges of normal to elevated FVIII (B) and mild HA (C) were compared to mock-injected WT/CD4KO mice. Adjusted p-values were obtained using pairwise log-rank (Mantel–Cox) tests to determine differences in individual cohort survival versus WT/CD4KO controls and corrected for multiple comparisons using the Holm–Šídák method. **p < 0.01; ***p < 0.001. Source data are provided as a Source Data file.

Confirming the physiologic relevance of APC in FVIIIa regulation, expression of normal or elevated plasma FVIII-QQ antigen in HA/FV^Q/Q mice resulted in significantly reduced survival relative to WT/CD4KO mice (Fig. 3B). In contrast, similar levels of FVIII-WT expression were tolerated. Commensurate with FVIII prothrombotic risk being concentration-dependent, expression of FVIII-QQ below normal FVIII plasma concentrations (<0.4 nM) in HA/FV^Q/Q mice had no survival disadvantage (Fig. 3C). These data demonstrate clear evidence that APC regulation of FVIIIa has physiologic relevance and further support that FVIII-QQ has enhanced procoagulant function.

FVIII-QQ is safe and does not exhibit a prothrombotic risk

To investigate FVIII-QQ prothrombotic risk in the absence of an underlying thrombophilia like FV-Leiden, a survival analysis was concurrently conducted on the same cohorts of mice analyzed for FVIII expression durability in Fig. 1. Specifically, survival of HA/CD4KO mice expressing FVIII-WT or FVIII-QQ in the range of mild HA, normal FVIII, or elevated FVIII concentrations were compared to WT/CD4KO mice (Fig. 4A and Supplementary Table 1) because HA/CD4KO mice do not survive to 72 weeks⁴⁴. Among all cohorts of FVIII-WT or FVIII-QQ expressing mice, there was no survival disadvantage relative to WT/CD4KO controls (Fig. 4B–D). HA/CD4KO mice had significantly reduced survival commensurate with that previously observed for immune-competent HA/C57BL/6 mice⁴⁴, supporting that the CD4KO background did not impact survival. To investigate further, week 12 and 32 d-dimer values of HA/CD4KO AAV treated mice and week 12 d-dimer values of HA/FVL^Q/Q treated mice were analyzed relative to thrombin-injected positive control mice. Week 12 and 32 d-dimer values for all FVIII expression cohorts in HA/CD4KO and HA/FVL^Q/Q mice, except for HA/FVL^Q/Q mice expressing normal or elevated FVIII-QQ, significantly differed from thrombin-injected positive d-dimer controls (Supplementary Fig. 3). These data correlate with survival study observations (Figs. 3, 4) and support that FVIII-WT and FVIII-QQ expression were tolerated at all ranges of expression in HA/CD4KO mice.

Fig. 4. FVIII-QQ expression is tolerated in all ranges in HA mice.

A HA/CD4KO mice were treated with AAV vector to express FVIII-WT or FVIII-QQ. Animals were assigned to mild HA (0.05 to <0.4 nM FVIII, n = 8–14), normal FVIII (≥0.4 to 1.5 nM FVIII, n = 8–12), and elevated FVIII (≥1.5 to 2 nM FVIII, n = 5–10) cohorts according to FVIII antigen (mean ± SEM) measured 8 weeks post-vector. B–D Survival outcomes for mock-injected HA/CD4KO mice and AAV-treated HA/CD4KO mice expressing FVIII-WT or FVIII-QQ in the ranges of mild HA (B), normal FVIII (C), and elevated FVIII (D) were compared to mock-injected WT/CD4KO mice. Adjusted p-values were obtained using pairwise log-rank (Mantel–Cox) tests to determine differences in individual cohort survival versus WT/CD4KO controls and corrected for multiple comparisons using the Holm–Šídák method. *p < 0.05. Source data are provided as a Source Data file.

To further assess prothrombotic risk in vivo, an Arg to Gln mutation at residue 562 was introduced by CRISPR/Cas9 to generate FVIII-QQ mice (FVIII^QQ); wild-type mice have a Gln at position 336⁴⁵. Homozygous female and hemizygous male FVIII^QQ mice were viable and fertile with normal Mendelian inheritance (Supplementary Table 2). Together, FVIII^QQ mice observations and that FVIII-QQ expression up to 2-fold normal FVIII concentrations was well-tolerated in HA/CD4KO mice demonstrates the safety of FVIII-QQ expression across two different mouse models.

Immune responses to FVIII-QQ and FVIII-WT do not differ in immune-competent mice

To assess the immunological risk, we investigated whether FVIII-QQ may differentially break tolerance in a mouse model of relative (h)uman FVIII-WT tolerance or induce an immune response in a hFVIII naïve model. Given first-in-human HA gene therapy efforts enroll patients with established FVIII-WT tolerance^6,46, the risk of FVIII-QQ versus FVIII-WT to break hFVIII-WT tolerance was investigated first. These studies were conducted in an immune-competent HA mouse model with ectopic hFVIII-WT expression in platelets (HA/pF8 mice)⁴⁷ that permits relative hFVIII tolerance⁴⁸. Mice were challenged with weekly 1μg of FVIII-WT or FVIII-QQ protein for 6 weeks followed by a larger dose of 5μg (Fig. 5A). FVIII-QQ and FVIII-WT challenged mice demonstrated no significant difference in the incidence or magnitude of functional FVIII inhibition determined by Bethesda titer (Fig. 5B) or total anti-FVIII IgG (Fig. 5C). Next, HA/pF8 mice without inhibitory antibodies following 4 weekly 1μg recombinant hFVIII-WT protein challenges were treated with AAV vector at a dose that conferred 1 nM steady-state FVIII expression in HA/CD4KO mice. As with recombinant protein challenge, there was no significant difference in the incidence or magnitude of Bethesda inhibitory antibody titer between FVIII-QQ and FVIII-WT expressing mice (Fig. 5D). Commensurate with the ability of liver-directed gene transfer to induce tolerance^49,50, all inhibitor-positive AAV-treated HA/pF8 mice tolerized (Fig. 5D). Consistent with prior work^26,51,52, studies in immune-competent HA/C57BL/6 mice challenged with recombinant FVIII (Supplementary Fig. 4a–c) or AAV-mediated FVIII expression (Supplementary Fig. 4d) demonstrated a robust immune response. However, like HA/pF8 observations, there were no differences in the incidence or magnitude of inhibitor development in FVIII-QQ versus FVIII-WT treated HA/C57BL/6J mice (Supplementary Fig 4). While limited to measuring an anti-FVIII humoral immune response only, together, these studies support that FVIII-QQ does not have increased risk of breaking FVIII tolerance or differential antibody formation relative to FVIII-WT.

Fig. 5. FVIII-QQ did not differentially break human FVIII tolerance relative to FVIII-WT.

A Schematic of recombinant FVIII immunology study design. Bethesda (B) and total anti-human FVIII IgG (C) titers for HA/pF8 mice immunized with recombinant human FVIII-WT (n = 5) or FVIII-QQ (n = 6). Magnitudes of FVIII-QQ versus FVIII-WT Bethesda and IgG titers did not significantly differ (multiple Mann–Whitney tests with Holm–Šídák correction for multiple comparisons). Cumulative incidence of inhibitor development (inset in B) did not significantly differ between FVIII-WT and FVIII-QQ immunized mice (Fisher’s exact test). Positive Bethesda titers were defined as ≥0.6 BU/mL (B, dotted line). Positive IgG titers were defined as those above untreated HA/pF8 background (2 µg/mL; C, dotted line). D Inhibitor-negative HA/pF8 mice following recombinant FVIII-WT protein challenge were infused with AAV vector to express human FVIII-WT (n = 7) or FVIII-QQ (n = 8) at normal FVIII antigen levels. Inhibitor development was determined by Bethesda assay with positive titers defined as ≥0.6 BU/mL (dotted line). Bethesda titers and cumulative incidence of inhibitor development (inset) did not differ between FVIII-WT and FVIII-QQ expressing mice (multiple Mann–Whitney tests with Holm–Šídák correction for multiple comparisons and Fisher’s exact test, respectively). Source data are provided as a Source Data file. Figure (D) were created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

Functional assessment of FVIII-QQ

Activation of protein C requires endothelial transmembrane-bound proteins (thrombomodulin and endothelial protein C receptor), which are not present in plasma⁵³. Thus, no appreciable APC is generated in traditional clotting assays. As a result, the specific activities of FVIII-QQ and FVIII-WT do not differ²⁰ in the absence of the PC pathway. To that end, to assess the hemostatic potential of FVIII-QQ more closely, modification of the traditional clotting assay with the incorporation of APC is required. Ultimately, this could prove beneficial or may be necessary for clinical translation of FVIII-QQ to both inform and assess clinical outcomes and safety. To investigate this, FVIII procoagulant function was assayed by activated partial thromboplastin time (aPTT) with or without APC to generate an APC resistance ratio (APCR). This approach is analogous to assays previously established in clinical practice to diagnose heterozygous FV-Leiden^54,55. An APCR is the ratio of the clot times in the presence versus absence of APC such that a lower APCR indicates APC resistance. Plasma from AAV-treated HA (Fig. 6A) or HA/FV^Q/Q (Fig. 6B) survival study mice was pooled and assayed across a range of plasma FVIII concentrations. As expected, FVIII APCR values were lower for FVIII-QQ versus FVIII-WT, and APCR values for mice on the FV^Q/Q background (Fig. 6B) were lower than HA mice (Fig. 6A). Similarly, APCR values for a humanized system of recombinant FVIII-QQ reconstituted in human HA plasma were lower than those for equivalent concentrations of FVIII-WT (Fig. 6C). Consistent with prior studies of the impact of FVIII concentration on APC resistance^56,57, APCR inversely correlated with FVIII concentration for all conditions tested. Importantly, survival outcomes for AAV-treated HA and HA/FV^Q/Q mice (Figs. 3, 4) correlated with APCR when stratified by FVIII concentration (Fig. 6A, B), which demonstrates that the safety of FVIII-QQ expression is APCR-dependent.

Fig. 6. Functional assessment of FVIII in vitro activity in the presence of APC.

A, B Pooled plasma from HA/CD4KO (A) and HA/FV^Q/Q/CD4KO (B) mice expressing FVIII-WT or FVIII-QQ was assayed at varying plasma FVIII concentrations for APC resistance (APCR). The blue-shaded regions represent APCR ± 10% for 1 nM FVIII-WT in the associated mouse plasma. Horizontal dotted lines represent APCR of mouse plasma without FVIII. Observed survival outcomes for varying levels of FVIII expression in HA (A and Fig. 4) and HA/FV^Q/Q (B and Fig. 3) mice are noted compared to their corresponding APCR values. C APCR for recombinant human FVIII-WT or FVIII-QQ reconstituted in human HA plasma at varying FVIII concentrations. The blue-shaded region represents APCR ± 10% for pooled normal human plasma (NHP; dotted line). D The concentration of FVIII-WT necessary to achieve the same clot time or procoagulant activity as FVIII-QQ in the presence of APC was calculated for FVIII-QQ concentrations in the range of mild HA. The dotted line represents the line of unity. E Fold-improved procoagulant function of FVIII-QQ versus FVIII-WT in the presence of APC was calculated. Data represent mean ± SEM of 3 independent measurements performed in duplicate. Source data are provided as a Source Data file.

Within the translationally targeted range of FVIII-QQ expression (0.05 to ≤0.4 nM), 3-5-fold higher FVIII-WT concentrations were required to achieve the same procoagulant activity as FVIII-QQ in the presence of APC (Fig. 6D, E). This observation was maintained in plasma from AAV-treated mice and in a fully humanized system of human HA plasma reconstituted with recombinant FVIII-WT or FVIII-QQ protein. These data are consistent with the previously demonstrated 4-5-fold improved in vivo hemostatic function of FVIII-QQ over FVIII-WT in multiple hemostatic injury models²⁰. Thus, aPTT-based FVIII activity measured in the presence of APC (APC FVIII activity) appears to be predictive of FVIII-QQ in vivo hemostatic function and it could be used as a simple clinical assay to assess FVIII-QQ function post gene transfer.

Discussion

Despite significant progress in HA gene therapy, the goal of sustained FVIII expression that normalizes hemostasis has not been achieved. Specifically, one major limitation in the field is durability of FVIII expression, a problem that has blunted licensed HA gene therapy uptake by patients. Data in mice herein support that plasma FVIII level-dependent declines in expression are due, at least in part, to a reduction in VCN. These important findings address a major mechanistic gap in the field and suggest that there may be intractable barriers to expressing FVIII in hepatocytes above a certain threshold. To circumvent this, we show that FVIII-QQ is safe and has superior hemostatic function relative to FVIII-WT in the context of gene therapy. These data are the first adaptation of a gain-of-function FVIII variant to demonstrate a more potent in vivo hemostatic response relative to FVIII-WT at equal plasma antigen expression values using a gene therapy approach. Our data suggest that FVIII-QQ would permit the use of lower AAV vector doses to minimize dose-dependent AAV immune toxicities, including those that limit efficacy^6,13,58, while normalizing FVIII function at below normal plasma FVIII concentrations to permit durable expression^6,9 and overcome current limitations of HA gene therapy.

Unexplained year-over-year loss of initially normal FVIII-WT expression^7,11,14 versus durable low-level FVIII-WT expression^6,9,14 have each been demonstrated. Our study is the first to demonstrate longitudinal pharmacokinetics of expression-level dependent loss of FVIII post-AAV gene therapy in an animal model that is qualitatively consistent with human observations. In mice with initially elevated steady-state FVIII levels, plasma FVIII and VCN were significantly reduced 72 weeks post-vector, supporting that VCN loss is an important contributor to declining FVIII levels. Our data do not support a mechanism involving gene silencing, which contrasts recent data^25,30,31,59. Though these mechanisms are not mutually exclusive, we note that the same AAV8 mammalian-produced vector preparations of an oversized cassette (5 kb) were used for all animals followed for 72 weeks. This suggests against capsid-specific epigenetic reduction in transcription observed with AAV-LK03 in mice²⁵, which was not maintained in non-human primates⁵⁹ or, as supported by clinical trial data, humans⁶. Additionally, prior work demonstrated epigenetic transcriptional reduction in human α1-antitrypsin expression with baculovirus, but not mammalian-produced vector in mice³⁰; while the absence of gene silencing observed with mammalian manufactured vector is consistent with our data, the impact of the observed transcriptional changes with baculovirus-produced vector are unclear because both vector preparations lost half of peak transgene expression³⁰. Further, vector properties, including incompletely packaged cassettes, exacerbated by the baculovirus manufacturing platform and oversized cassettes^29,60, are suggested to impact hepatocyte transgene expression and durability³⁰. Inconsistent with these hypotheses, we observed FVIII-dependent declines in expression with an oversized cassette (5 kb), independent of the baculovirus system. Additionally, adult mice were treated and declining FVIII plasma antigen in the elevated FVIII cohort began >40 weeks post-vector. This is outside the window of expected stable episome formation²⁸ and demonstrated mouse hepatocyte proliferation⁶¹, suggesting against a mechanism involving either processes. Consistent with prior studies of FVIII expression in mammalian expression systems^32,34–36, we observed a significant positive correlation between ER stress markers and FVIII expression. These data do not establish a definitive mechanistic link between ER stress and VCN loss, which may be limited by insensitive timing of liver sampling, which is also a concern for ongoing human studies³¹, analysis methods or other factors. Nonetheless, we hypothesize that loss of VCN is due to induction of an unfolded protein response driven by the level of FVIII expression resulting in loss of transduced cells. An ongoing study of human liver biopsies post valoctocogene roxaparvovec may determine if our studies in mice translate to humans³¹. Nonetheless, human observations expressing FVIII, including clinical trial data of an enhanced secretion FVIII variant⁹, have not achieved sustained normal FVIII expression suggesting this may not be possible with current approaches. Alternatively, the goal of sustained normalized FVIII hemostatic function may best be achieved using a gain-of-function variant, such as FVIII-QQ.

To this end, we demonstrate that FVIII-QQ has greater function relative to FVIII-WT in the setting of gene therapy in multiple model systems. Interestingly, kinetic studies of FVIII/FVIIIa regulation have reported APC cleavage is up to an order of magnitude slower than A2-domain dissociation^62–64, which calls into question the physiologic relevance of APC in FVIIIa regulation. Importantly, however, there are limitations of modeling FVIIIa inactivation in vitro due to the difficulty of purified systems to accurately replicate physiologic inter- and intramolecular interactions that are known to impact A2-domain dissociation^65–67 and APC cleavage^68,69. For the clinical translation of FVIII-QQ, this highlights the importance of in vivo assessment of APC on FVIIIa regulation. We demonstrated that inhibiting APC regulation of FVIIIa via FVIII-QQ improves in vivo hemostatic function in recombinant protein studies²⁰ and, herein, post gene transfer. Furthermore, induction of APC anticoagulant deficiency by expressing FVIII-QQ in HA/FVL^Q/Q mice recapitulated the lethal phenotype observed in protein C-deficient mice and humans^42,43,70. This unambiguously demonstrates that APC regulation of FVIIIa has physiologic significance and extends mechanistic understanding of in vivo FVIIIa regulation.

The prohemostatic effect of bypassing APC regulation of FVIIIa must be balanced and thoughtfully considered with established correlations between elevated FVIII levels and venous thrombosis risk^56,71,72. That variable FVIII expression has been observed in multiple HA gene therapy trials^8,46,73 and excess FVIII and von Willebrand factor function are the most prothrombotic of all procoagulant proteins⁷¹ highlights the importance of ensuring FVIII expression is maintained within a safe therapeutic window. Indeed, the range of tolerated FVIII expression will narrow inversely with the degree of enhanced FVIII hemostatic effect, raising the possibility of diminishing therapeutic returns of efforts to improve FVIII function. Analogous to our approach, preliminary clinical trial data of a recombinant protein partially inhibiting APC anticoagulant function⁷⁴ demonstrated marked improvement in bleeding without prothrombotic events⁷⁵. Our studies show that all ranges of FVIII-QQ expression, including up to 2-fold normal values, were well tolerated in HA/CD4KO mice. While encouraging, we acknowledge that mice do not comprehensively recapitulate human venous thrombosis^38,76. To address this, we also evaluated FVIII-QQ prothrombotic risk in a FV-Leiden mouse model, a provocative positive control that probes the same regulatory pathway and has a known venous thrombosis risk in humans and mice^39–41. We found that FVIII-QQ expression in the range of mild HA (0.05 to <0.4 nM) was tolerated on the homozygous FV-Leiden background. These data provide compelling evidence that support FVIII-QQ safety in this targeted range of therapeutic expression, even in the setting of a major thrombophilia. Further, in vitro assessment of transgene-derived or recombinant FVIII-QQ demonstrated FVIII-QQ has 3 to 5-fold enhanced activity over FVIII-WT in the presence of APC; this is remarkably consistent with established in vivo enhanced potency of FVIII-QQ over FVIII-WT²⁰. Thus, data support in vitro determination of FVIII activity in the presence of APC may accurately reflect in vivo hemostatic function and predict clinical outcome. Overall, data support that targeting FVIII-QQ expression in the range of mild HA may balance enhanced hemostatic function with thrombotic risk thereby permitting a range of safe therapeutic expression.

In conclusion, our data in mice demonstrate that plasma FVIII level-dependent declines in expression are due, at least in part, to loss of VCN and, thus, begin to address a major outstanding question in HA gene therapy. Data support that an enhanced function FVIII variant may be necessary for second-generation HA gene therapy to overcome current durability and efficacy challenges. Herein reported data support that FVIII-QQ expression below normal FVIII concentrations may safely restore hemostasis while permitting durability of expression and are the first to demonstrate that a gain-of-function FVIII variant has improved hemostatic function in the setting of gene therapy. These results address the molecular basis of a major limitation of HA AAV gene therapy and provide a rationally bioengineered solution.

Methods

Proteins and reagents

Recombinant human B-domain deleted FVIII proteins (FVIII-WT and FVIII-QQ) were purified as previously described²⁰. FVIII-WT is the FVIII-SQ variant of B-domain deleted FVIII⁷⁷. FVIII protein concentration was determined using a molecular weight (M_r) of 165,000 and extinction coefficient (E^0.1%) of 1.6. Pooled normal human plasma and FVIII-deficient plasma were purchased from George King Biomedical. TriniCLOT Automated aPTT reagent (Tcoag) was purchased from Diagnostica Stago. Human plasma-derived activated protein C (APC) was purchased from Haematologic Technologies/Prolytix and concentration was determined using a M_r of 45,000 and E^0.1% of 1.45²⁰.

Recombinant AAV vectors

Human B-domain deleted F8 cDNA (FVIII-WT, FVIII-QQ) was cloned into a previously described pAAV plasmid (gift from D. Sabatino, Children’s Hospital of Philadelphia) containing a modified transthyretin (TTRm) promoter, synthetic intron, and AAV2 inverted terminal repeats (ITRs) flanking the expression cassette⁷⁸. Two sets of FVIII-WT and FVIII-QQ expression cassettes differing only by F8 cDNA (CO1, CO2) were used to evaluate FVIII expression. Two separate preparations of FVIII-QQ and FVIII-WT CO1 vectors were used for experiments and demonstrated comprable expression profiles. Recombinant AAV8 vectors were generated via HEK293 triple transfection by the Research Vector Core at the Children’s Hospital of Philadelphia and titerd by digital droplet PCR per previously described methods^78–80. Vectors were stored at -80 °C until use.

Mouse studies

Unless otherwise noted, all mice used were on the C57BL/6 background and 8-12 weeks of age at the time of treatment. Given that FVIII is X-linked and hemophilia A (HA) predominantly affects the male sex, corresponding male sex mice were used for all studies. Immune-deficient CD4-knockout (WT/CD4KO) mice were purchased from Jackson Laboratory and used as healthy experimental controls. HA/CD4KO mice were crossed with homozygous FV-Leiden mice (FV^Q/Q) to generate HA/FV^Q/Q/CD4KO mice. To generate a mouse model of endogenous FVIII-QQ expression (FVIII^QQ), an Arg to Gln mutation was introduced at residue 562 in C57BL/6 embryos using CRISPR/Cas9 by the Transgenic Core at the Children’s Hospital of Philadelphia per previously described methods⁸¹. Wild-type mice are naturally Gln at residue 336⁴⁵. FVIII^QQ mice underwent ≥5 backcrosses to C57BL/6 mice prior to study. For (r)ecombinant FVIII and AAV-FVIII immunology studies, two mouse models were used: immune-competent HA (HA/C57BL/6) mice and transgenic HA C57BL6/129 mice that ectopically express human FVIII-WT under a megakaryocyte-specific promoter in platelets (HA/pF8)⁴⁷. HA/pF8 mice were 13-17 weeks of age and HA/CD4KO mice were 8–12 weeks at the time of treatment. Recombinant FVIII protein (in PBS) and AAV-FVIII vectors (in sterile-filted PBS + 0.001% Pluronic F-68 [Gibco]) were diluted to 200 µL final volumes and infused by tail vein injection. Mock-injected animals for AAV studies received 200 µL PBS/0.001% Pluronic F-68. Peripheral blood was collected from the retroorbital plexus into 3.8% sodium citrate (9:1 vol/vol). Plasma was isolated by centrifuging citrated blood for 10 min at 10,000g at 4 °C and stored at −80 °C until use. All animals were fed PicoLab® 5015 diet and experimental procedures were approved by the Institutional Animal Care and Use Committee at the Children’s Hospital of Philadelphia.

Quantification of human FVIII antigen in mice

Plasma FVIII antigen circulating in AAV-treated mice was determined by enzyme-linked immunosorbent assay (ELISA) using matched-pair antibodies against human FVIII (Affinity Biologicals). Two modifications were made to the manufacturer’s protocol: wells were coated with primary capture antibody overnight at 4 °C and well volumes were reduced by half (e.g. 50 µL sample volume). Standard curves were prepared with respective recombinant FVIII protein (FVIII-WT or FVIII-QQ) reconstituted in plasma collected and pooled from untreated mice of the appropriate strain. Standards and mouse plasma samples were diluted 10-fold in green dilution buffer prior to plating in duplicate. Sample FVIII antigen was determined by converting absorbance at 490 nm using the sample’s respective standard curve. Measured values below background were assigned a value of 0.

Assessment of AAV vector genome integrity

Heterogenous vector genome packaging of FVIII-WT and FVIII-QQ AAV8 vectors was compared to that of a control vector, AAV2-CAG-eGFP (3 kb expression cassette; prepared in-house by the Research Vector Core at the Children’s Hospital of Philadelphia), using genomic DNA ScreenTape (Agilent) on a TapeStation 4150 system (Agilent). Vector DNA was incubated at 95 °C for 10 min and cooled down to room temperature gradually, allowing single-stranded AAV genomes (positive and negative sense) to form double-stranded DNA. Heat-denatured double-stranded DNA samples were then diluted in sample buffer and loaded into the TapeStation system according to the manufacturer’s protocol. Simulation gel images and electropherograms were automatically generated by the Agilent TapeStation Controller software.

Isolation of nucleic acid from mouse livers

At endpoint collections (8, 12, or 72 weeks post-vector), the left lobe of the liver was harvested and stored at −80 °C for future analysis. RNA and DNA was isolated from frozen liver using the AllPrep DNA/RNA/Protein Mini Kit (QIAGEN) following the manufacturer’s protocol. Briefly, liver samples (20-25 mg) were added to Buffer RLT Plus with ß-ME and homogenized using stainless steel beads (QIAGEN) in a TissueLyser LT (QIAGEN). Eluted RNA and genomic DNA (gDNA) was quantified and stored at -80 °C until further use.

Determination of vector copy number (VCN)

Droplet digital PCR (ddPCR) was used to quantify VCN for livers harvested from AAV-treated mice as previously described⁸². Isolated gDNA (200 ng) was digested with restriction enzymes that cut at the 5’ (SacII) and 3’ (NotI) ends to excise the vector expression cassette (excluding ITRs). Digested gDNA ( ~ 20 ng) was added in duplicate to 96-well PCR plates containing a master mix of ddPCR Supermix for Probes (No dUTP) (Bio-Rad). FAM-conjugated Assays for human F8 were designed and ordered from Integrated DNA Technologies to specifically detect either CO1 or CO2 F8, and a HEX-conjugated ddPCR Copy Number Assay for mouse diploid reference gene Ap3b1 (dMmuCNS801070401, Bio-Rad) was used. Nanoliter-sized droplets were generated with a QX200 Droplet Generator (Bio-Rad), transferred to a new 96-well PCR plate, and run in a T100 thermal cycler (Bio-Rad) following the manufacturer’s standard protocol. Droplets from each well were read and individually analyzed on a QX200 Droplet Reader (Bio-Rad) with PCR-positive and PCR-negative droplets identified using QX Manager Software (Bio-Rad). VCN was then calculated as vector copies per diploid genome.

Determination of relative mRNA expression

To measure relative mRNA expression, cDNA was synthesized from 2 µg of total RNA isolated from AAV-treated mouse livers using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) according to the manufacturer’s protocol. Synthesized cDNA was used to quantify relative expression of mRNA by real-time reverse transcription quantitative PCR (RT-qPCR) using mouse GAPDH (TaqMan Gene Expression Assay Mm99999915_g1, Thermo Fisher) as an endogenous control. Gene Expression Assays for human F8 were used to evaluate FVIII-WT and FVIII-QQ mRNA expression. To assess induction of ER stress, mouse BiP (or HSPA5; TaqMan Gene Expression Assay Mm00517691_m1, Thermo Fisher) and CHOP (or DDIT3; TaqMan Gene Expression Assay Mm00492097_m1, Thermo Fisher) probes were used. RT-qPCR was performed using TaqMan Universal Master Mix II, no UNG (Applied Biosystems) according to the manufacturer’s protocol on a CFX384 Touch Real-Time PCR System (Bio-Rad). Relative mRNA expression was calculated using the comparative C_T method (2^-∆∆Ct method)⁸³ and GAPDH was used as an internal control gene.

Tail clip assay

Tail clip injury was performed 9- or 10-weeks post-vector as previously described^20,84 and animals were euthanized after the assay. Briefly, mice were anesthetized with 2% isoflurane and tail was prewarmed to 37 °C for 3 min. The tail was transected at a 3-mm diameter. After tail transection, blood was collected for 2 min and then an additional 10 min into normal saline at 37 °C. Ten minute samples were hemolyzed. Total hemoglobin was determined by measuring absorbance at 575 nm and converted to blood loss (µL) using an established standard curve of known amounts of hemolyzed murine whole blood as previously described⁸⁴.

Survival analysis

Animals were monitored for survival over a 72-week period with blood collections every 4 weeks. Deaths were reported on the week of the first blood collection for which animals were confirmed dead. Kaplan-Meier curves were plotted to determine survival of AAV-treated mice compared to mock-treated WT/CD4KO mice.

Genotyping of FVIII^QQ progeny

Ear snips from FVIII^QQ progeny were collected when mice were 5-9 weeks of age. Genomic DNA was extracted from ear snips using the KAPA HotStart Mouse Genotyping Kit (Roche) according to the manufacturer’s protocol. The region surrounding the mouse FVIII-R562Q mutation was amplified with forward and reverse primers 5’-GGATTTGCCAATTCATCCAGGAG-3’ and 5’-CTAAGTGGAACTCACCTTAATCAC-3’, respectively. The following modifications were made to the cycling protocol: 35 cycles consisting of 30 s at 95 °C, 15 s at 48 °C, and 10 s at 66 °C followed by 1 min at 66 °C. PCR products were then column purified using the QIAquick PCR Purification Kit (QIAGEN) and Sanger sequenced by Genewiz (Azenta Life Sciences) with primer 5’-GTGGACAGTTACAGTAGAAGATGG-3’.

D-dimer quantification

D-dimer levels were measured by the human-specific Asserachrom D-Di ELISA (Diagnostica Stago) according to the manufacturer’s protocol and per prior description⁸⁴. Mouse plasma was diluted 5-fold in dilution buffer prior to plating and D-dimer concentration was interpolated from the standard curve with adjustment for differences in plating dilutions. Positive D-dimer mouse samples were collected in 30 min after injection of 12.5 units of thrombin (Haemtech) into WT/CD4KO, analogous to prior description of methods used to induce thromboembolism in mice⁸⁵.

Quantification of FVIII activity in mice

FVIII activity by chromogenic assay (COAMATIC, Chromogenix) was measured according to the manufacturer’s protocol with some modifications. Standard curves were prepared with rFVIII-WT or rFVIII-QQ reconstituted in plasma collected and pooled from untreated mice of the appropriate strain. Standards and mouse plasma samples were diluted 80-fold in dilution buffer prior to plating in duplicate and incubating 4 min at 37 °C. Pre-warmed factor reagent (FIXa, FX, IIa, CaCl₂, phospholipids) was added to the wells and incubated 4 min at 37 °C. Pre-warmed substrate (S-2765 + I-2581) was added to the wells immediately before measuring the change in absorbance at 405 nm and 37 °C over 2 min (high standard curve, t = 0–2) or 8 min (low standard curve, t = 6–14). Sample FVIII activity was determined first using the high standard curve. Samples with FVIII activity measuring <0.5 nM with the high standard curve were then determined using the low standard curve and the two FVIII activity values were averaged. If the result was below background, it was assigned a value of 0.

FVIII-specific APCR assay

Adapting a previously established clinical APC resistance ratio (APCR) assay to diagnose FV-Leiden⁵⁴, FVIII APCR was determined by the ratio of time to clot formation (CT) in the presence of APC versus the absence of APC using a STart4 analyzer (Diagnostica Stago).

$$\mathrm{APCR}=\frac{{\mathrm{CT}}_{\mathrm{APC}}}{\mathrm{CT}}$$

For untreated mice, plasma was diluted 5-fold in assay buffer (HBS, 0.1% BSA) and kept on ice until assayed. Samples were then incubated with equal volumes of aPTT reagent and human FVIII-deficient (HA) plasma for 3 min at 37 °C. Coagulation was initiated with the addition of an equal volume of CaCl₂ ± APC (15 nM final APC concentration). For AAV-treated mice, plasma was pooled from at least 3 different mice and chromogenic FVIII activity of the pool was determined. Serial dilutions of the pooled sample were made with plasma from untreated mice of the respective strain and assayed for APC resistance as described above. For human FVIII-deficient plasma reconstituted with rFVIII, samples were diluted 5-fold in human FVIII-deficient plasma and incubated with an equal volume of aPTT reagent for 3 min at 37 °C. Coagulation was initiated with the addition of an equal volume of CaCl₂ ± APC (7.5 nM final APC concentration).

Bethesda assay

To measure anti-FVIII inhibitor titers, the Bethesda assay was performed per prior description⁸⁶ with modifications. Mouse plasma samples were heat inactivated at 56 °C for 30 min to remove residual FVIII and stored at −80 °C if not used immediately⁸⁷. 1 U/mL rFVIII-WT reconstituted in imidazole-buffered human FVIII-deficient plasma was used as a control for normal FVIII activity. Equal parts sample mouse plasma (diluted into HBS/0.05% Tween-80) and control plasma were incubated at 37 °C for 2 h. Control plasma diluted 2-fold with HBS/0.05% Tween-80 was used as a negative assay control. A standard curve of rFVIII-WT reconstituted in imidazole-buffered human FVIII-deficient plasma was generated to quantify residual FVIII activity. Standards were diluted 2-fold with HBS/0.05% Tween-80 prior to activity measurement. FVIII activity was determined via one-stage aPTT using a STart4 analyzer (Diagnostica Stago). 5 µL of standard, control, or sample mouse plasma mixture was incubated with 50 µL human FVIII-deficient plasma and 50 µL aPTT reagent for 3 min at 37 °C. Coagulation was initiated with the addition of 50 µL of 25 mM CaCl₂. One Bethesda unit per milliliter (BU/mL) was defined as the dilution at which 50% of FVIII activity was inhibited. A sample was considered inhibitor-positive if it had ≥0.6 BU/mL.

IgG ELISA

Total anti-FVIII IgG titers were measured as previously described⁴⁸ with modifications. Briefly, wells were coated with either a standard curve prepared from mouse IgG (Sigma-Aldrich; 0.98-2000 ng/mL) or 1 µg/mL rFVIII-WT to capture FVIII-specific IgG from sample mouse plasma. Pooled plasma from untreated HA/C57BL/6 or HA/pF8 mice was used as background controls for treated mice of their respective strains. Plasma samples from HA/pF8 mice were heat inactivated prior to assaying. These mice ectopically express human FVIII in platelets, some of which may be in circulation and bound to hFVIII-specific antibodies. Bound IgG was detected with 1 µg/mL polyclonal rabbit anti-mouse IgG-HRP (Dako).

Data analysis

Analysis were performed with Graphpad Prism software with specific statistical testing outlined in figure legends and/or in the text.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

A.R.S., C.M.R., R.J.D., and X.L. performed experiments and analyzed the data. A.R.S. coordinated experiments. A.R.S. and L.A.G. designed experiments, analyzed data, and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Edward Tuddenham who co-reviewed with Giulia Simini and the other anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data generated in this study have been deposited in the figshare database (10.6084/m9.figshare.26308642) and cannot be used for regulatory filings and/or commercial drug development. Source data are provided with this paper.

Competing interests

The authors declare the following competing interests, Dr. George is on the Scientific Advisory Board of Form Bio and STRM.bio, a consultant for Pfizer, Regeneron, Spark and Tome Biosciences and received licensing fees and research support from Asklepios BioPharmaceutical. All other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-51296-8.