Characterising diflunisal as a transthyretin kinetic stabilizer at relevant concentrations in human plasma using subunit exchange

Abstract Transthyretin (TTR) dissociation is the rate limiting step for both aggregation and subunit exchange. Kinetic stabilisers, small molecules that bind to the native tetrameric structure of TTR, slow TTR dissociation and inhibit aggregation. One such stabiliser is the non-steroidal anti-inflammatory drug (NSAID), diflunisal, which has been repurposed to treat TTR polyneuropathy. Previously, we compared the efficacy of diflunisal, tafamidis, tolcapone, and AG10 as kinetic stabilisers for transthyretin. However, we could not meaningfully compare diflunisal because we were unsure of its plasma concentration after long-term oral dosing. Herein, we report the diflunisal plasma concentrations measured by extraction, reversed phase HPLC separation, and fluorescence detection after long-term 250 mg BID oral dosing in two groups: a placebo-controlled diflunisal clinical trial group and an open-label Japanese polyneuropathy treatment cohort. The measured mean diflunisal plasma concentration from both groups was 282.2 M 143.7 M (mean standard deviation). Thus, quantification of TTR kinetic stabilisation using subunit exchange was carried out at 100, 200, 300, and 400  concentrations, all observed in patients after 250 mg BID oral dosing. A 250 M diflunisal plasma concentration reduced the wild-type TTR dissociation rate in plasma by 95%, which is sufficient to stop transthyretin aggregation, consistent with the clinical efficacy of diflunisal for ameliorating transthyretin polyneuropathy.

Diflunisal (250 mg BID) is established to reduce the rate of progression of neurologic impairment in patients with transthyretin familial amyloid polyneuropathy (FAP, now known as ATTRv-PN) based on results from a placebo-controlled randomised clinical trial [21]. Here we report the quantification of the diflunisal concentrations in the plasma samples from participants in the placebo-controlled diflunisal clinical trial using methodology previously utilised for tafamidis [12,21,22]. We also quantified diflunisal levels in the plasma of Japanese polyneuropathy patients treated with oral diflunisal (250 mg BID) in an open-label study [16]. Knowing the diflunisal plasma concentration ranges in the clinical trial samples (250 mg BID) and in the Japanese open-label study (250 mg BID) allowed us to add diflunisal back to healthy plasma to determine the extent of wild type (WT) TTR kinetic stabilisation over this range of diflunisal plasma concentrations.

Materials and methods
Detailed experimental procedures can be found in the Supplementary Material for diflunisal extraction from human plasma, for the high-performance liquid chromatography quantification of plasma diflunisal concentration, and for performing the subunit exchange assay to assess kinetic stabilisation of transthyretin.

Diflunisal plasma concentrations achieved by 250 mg BID oral dosing
Diflunisal plasma levels were measured in 50 plasma samples from the placebo-controlled clinical trial (Figure 1(A,B)) [21].
These polyneuropathy patients were either on 250 mg diflunisal BID or placebo. Twenty-three samples exhibited a peak at $14.3 min in the HPLC chromatogram ( Figure 1(C, red filled circles)). We confirmed that this peak (Figure 1(A)) was diflunisal using liquid chromatography/negative ion mass spectrometry (m/z 248.9). The remainder of the patient samples showed no signal, consistent with these patients being in the placebo group. The majority of samples (21/23) exhibited diflunisal plasma concentrations ranging from % 100 to 400 mM (Figure 1(C, red filled circles)), approximating the trough and average plasma concentrations upon chronic 250 mg BID dosing of hereditary ATTR amyloidosis patients [19,20,23]. The FDA package insert cites a mean plasma concentration of 56 mg/mL, or 224 mM, for the 250 BID diflunisal dose. Among the placebo-controlled clinical trial samples, the mean diflunisal plasma concentration was 275.0 mM. Higher diflunisal concentrations were exhibited by two samples (% 600 mM and % 760 mM), likely approximating the peak concentrations realised by oral dosing, which is achieved 2-3 h after oral dosing [19,20,23].
Diflunisal concentrations were also measured in 23 plasma samples from Japanese individuals participating in a diflunisal (250 mg BID) open-label clinical study at the Shinshu University School of Medicine [16]. The mean diflunisal plasma concentration in this cohort was 289.4 mM (Figure 1(C, blue filled circles)). The majority of samples exhibited diflunisal plasma concentrations comparable with the clinical trial, ranging from % 100 to 500 mM; however, one sample possessed a higher concentration outside this range (% 670 mM; Figure 1(C, blue filled circles)).
We observed two additional peaks in the HPLC chromatograms at $10.1 and $11.1 min that were identified as the diflunisal metabolites, diflunisal acyl glucuronide and diflunisal phenolic glucuronide, both exhibiting a molecular weight of 426.32 (Figure 1(B)). Both metabolites gave [M-H] À peaks of 425.0, confirming their identity [24]. While the plasma diflunisal peak at $14.3 min reflected the majority of the total HPLC chromatographic area (69.8 À 85.8%) in treated patients, the metabolite peak at $10.1 min ranged from 6.5% to 18.9% of the total chromatographic peak area and the metabolite peak at $11.1 min constituted 5.8 À 15.4% of the total chromatographic peak area.
We found no significant difference when we compared the mean diflunisal plasma concentration in the combined treatment groups by gender (male: 272.06126.6 mM; female: 298.06169.5 mM) (Figure 1(D)). However, we did see a  (Figure 1(D)). The mean diflunisal plasma concentration of 282.26143.7 lM (mean 6 SD) was derived from quantifying the concentration of diflunisal from the plasma samples of the combined treatment groups (Figure 1(C, red and blue filled circles)). Patient demographics were similar amongst the two groups. In the Clinical Trial cohort, 10/23 patients were under 65 and 13/ 23 were male. Among the Japanese group, 9/23 patients were under 65 and 15/23 were male. The most prevalent mutation in both groups was V30M (13/23 in the Clinical trial group and 14/23 in the Japanese cohort). Other mutations harboured by the patients were: E42G, T60A, Y114C, H90N, T58H, T60A, D38A, S50R, I107V, S50I, and E54K.

Subunit exchange assay to assess TTR kinetic stability
Subunit exchange between endogenous plasma WT TTR ($3.7 mM) and recombinant dual-FLAG-tagged Cys10Ala TTR (FT 2 -C10A TTR) added at a final concentration of 1 mM to pooled healthy human plasma samples was monitored over seven days, at final diflunisal concentrations of 100, 200, 300, and 400 mM versus vehicle control (DMSO) (Figure 2). In order to determine the concentration of endogenous tetrameric TTR in our pooled healthy plasma sample (3 donors, see Supplementary Material), we compared the integrated area of Peak 1 (endogenous native TTR tetramer) with that of Peak 5 (FT 2 -C10A-TTR) before subunit exchange occurred. The Peak 1 Peak 5 ratio afforded a value of 3.7 mM for the endogenous TTR tetramer concentration. We had planned to quantify diflunisal-mediated TTR stabilisation in the plasma of the diflunisal familial polyneuropathy clinical trial participants by subunit exchange, however the high viscosity of the plasma samples, likely due to plasma thawing and the plasma heating during shipping, precluded this. The apparent rate constant for TTR subunit exchange (k ex ) determined in this study in the absence of any kinetic stabiliser is 0.0151 h À1 , consistent with the rate constant from prior subunit exchange data (k ex ¼ 0.0149 h À1 ) [13]. The plasma TTR subunit exchange rate constant (k ex ) at a final diflunisal plasma concentration of 100 lM decreased to 0.0027 h À1 , at 200 mM diflunisal it decreased to k ex ¼ 0.0010 h À1 , at 300 mM diflunisal it decreased to k ex ¼ 0.0006 h À1 , and at a 400 mM diflunisal concentration the subunit exchange rate decreased to 0.0003 h À1 . In our previous paper [13], we assessed subunit exchange between plasma WT TTR and dual-FLAG-tagged WT TTR (FT 2 -WT TTR; employing a final concentration ¼ 1 mM) at 1, 5, 10, 20, and 30 mM final diflunisal plasma concentrations, all concentrations below those that are pharmacologically relevant at the standard 250 mg BID dose, but were germane to the prior publication in that they enabled direct comparisons with the plasma concentrations of tolcapone, tafamidis, and AG10 achieved by oral dosing [13]. From the data reported by Nelson et al. [13] we were able to conclude that the primary determinant of a kinetic stabiliser's efficacy at a given concentration for inhibiting TTR tetramer dissociation in plasma is the spread between its dissociation constant from the first TTR thyroxine binding site engaged (K D1 ) [10] and its dissociation constant from albumin (K D,Alb ¼ 1.2 ± 0.1 mM for diflunisal) [13]. The K D,Alb/ K D1 ratio decreases in the order of AG10 (K D,Alb/ K D1 ¼1980), tafamidis (K D,Alb/ K D1 ¼580) and diflunisal (K D,Alb/ K D1 ¼16) [13].
Fitting the data obtained herein from the subunit exchange between WT TTR in plasma and FT 2 -C10A TTR added to a final concentration of 1 lM in the presence of 100, 200, 300 and 400 mM final diflunisal plasma concentrations (data summarised in Figure 2) [13], allows us to more accurately model TTR subunit exchange at pharmacologically relevant diflunisal concentrations. Given the K D1 ¼ 75 nM and K D2 ¼ 1.1 mM TTRdiflunisal dissociation equilibrium constants, we find that the best fit value of K D, Alb is 1.4 ± 0.1 mM. This value was determined using a previously published method which can be found in the Supplemental Materials section of Nelson et al. [13]. Thus, by using a pharmacologically relevant range of diflunisal concentrations, our K D,Alb/ K D1 becomes 18.7, similar to the value 16 we calculated with the 1 À 30 mM diflunisal concentration range [13]. Based on this value of K D,Alb and the known K D1 and K D2 for diflunisal dissociation from TTR, we calculate that dissociation of TTR can be limited to 10% of its normal rate at a diflunisal concentration of 172 mM (188 mM was the value extracted from the 1 to 30 mM diflunisal subunit exchange data [13]). For comparison, 10% of the normal subunit exchange rate is achieved with 12.0 mM tafamidis and 5.7 mM AG10 plasma concentrations. Dissociation of WT TTR can be limited to 5% of its normal rate at diflunisal, tafamidis and AG10 concentrations of 250 mM, 20.7 mM, and 8.5 mM, respectively.

Discussion and conclusion
These data and prior data [13] demonstrate that any desired level of TTR stabilisation can be attained in human plasma by adjusting the kinetic stabiliser oral dose, which determines the kinetic stabiliser plasma concentration, safety attributes permitting. Collectively, our results herein and previously reported [13] clearly demonstrate that oral dosing of 500 mg diflunisal/day, 61 mg tafamidis/day, and 1600 mg of AG10/day achieves comparable WT TTR kinetic stabilisation, i.e. these daily doses limit the dissociation of WT TTR to 5% of its normal rate at diflunisal, tafamidis and AG10 plasma concentrations of 250 mM, 20.7 mM, and 8.5 mM, respectively [13]. Owing to this, differences in safety, pharmacokinetics, absorption, metabolism, tissue distribution, and tissue pharmacodynamics become important characteristics to consider when deciding which kinetic stabiliser to use. As the diflunisal dose used for TTR kinetic stabilisation is below doses typically required to treat arthritis, diflunisal has proven to be remarkably well tolerated in patients with ATTR amyloid polyneuropathy [16,21,25]. Nonetheless, potential NSAID-related decreases in renal blood flow and increased gastrointestinal bleeding risk warrant monitoring. While it remains unclear what minimal reduction in tetramer dissociation rate correlates with a maximum clinical response, published literature is starting to provide some guidance [16,17,21,[26][27][28]. We previously found that the average tafamidis concentration in patients with TTR polyneuropathy dosed with the Vyndaqel 20 mg once daily formulation of tafamidis was 8.2 mM, and that this concentration was sufficient to achieve a clinical response in about two-thirds of these patients [12]. This concentration of tafamidis would suppress the WT TTR dissociation rate to about 16% of normal (less dissociation rate suppression would be observed in heterozygous polyneuropathy patients having predominantly heterotetramers in solution that bind the kinetic stabilisers less well, e.g. V30M heterozygotes; this would almost certainly be the case for diflunisal as well). Suppressing the WT TTR dissociation rate to about 16% of normal can be achieved at a plasma diflunisal concentration of 124 mM, which is achieved in the vast majority of the patients evaluated in this study.

Disclosure statement
JWK and ETP discovered tafamidis, receive sales royalties and sales milestone payments.

Funding
The repurposing of diflunisal and the discovery of tafamidis for ATTR amyloidosis would not have been possible without sustained NIH funding [DK 046335] for the Kelly lab. Figures were made with BioRender.com.