Amyloidogenic immunoglobulin light chain kinetic stabilizers comprising a simple urea linker module reveal a novel binding sub-site

Abstract

In immunoglobulin light chain (LC) amyloidosis, the misfolding, or misfolding and misassembly of LC proteins or fragments thereof resulting from aberrant endoproteolysis, causes organ damage to patients. A small molecule “kinetic stabilizer” drug could slow or stop these processes and improve prognosis. We previously identified coumarin-based kinetic stabilizers of LCs that can be divided into four components, including a “linker module” and “distal substructure”. Our prior studies focused on characterizing carbamate, hydantoin, and spirocyclic urea linker modules, which bind in a solvent-exposed site at the V_L-V_L domain interface of the LC dimer. Here, we report structure-activity relationship data on 7-diethylamino coumarin-based kinetic stabilizers. This substructure occupies the previously characterized “anchor cavity” and the “aromatic slit”. The potencies of amide and urea linker modules terminating in a variety of distal substructures attached at the 3-position of this coumarin ring were assessed. Surprisingly, crystallographic data on a 7-diethylamino coumarin-based kinetic stabilizer reveals that the urea linker module and distal substructure attached at the 3-position bind a solvent-exposed region of the full-length LC dimer distinct from previously characterized sites. Our results further elaborate the small-molecule binding surface of LCs that could be occupied by potent and selective LC kinetic stabilizers.

Graphical Abstract

In immunoglobulin light chain (LC) amyloidosis (AL), free full-length immunoglobulin LCs secreted by a plasma cell clone undergo some combination of misfolding, aberrant proteolysis, and misassembly to afford a variety of non-native structures, including amyloid fibrils^1–6. These processes are associated with post-mitotic tissue degeneration and organ dysfunction in AL patients. These diseases are eventually fatal if untreated^{2–3, 7}. Existing therapies for AL rely on eradicating the plasma cell clones, which stops the secretion of amyloidogenic LCs⁸. Many of these treatments utilize cytotoxic chemotherapy regimens that are often poorly tolerated by AL patients, especially in those with severe cardiac pathology. Daratumumab, the first FDA-approved drug for AL, is an anti-CD38 monoclonal antibody that induces apoptosis in plasma cell clones and is a safe and effective emerging treatment for treating a subset of AL patients⁹. However, there is still a need for therapies that can suppress LC misfolding, aberrant proteolysis and aggregation, as it is very difficult fully eradicate the slowly expanding plasma cells.

We envision the use of small molecule “kinetic stabilizers” that bind to the native state of full-length (FL) LC dimers to prevent misfolding, misassembly, and aberrant endoproteolysis of LCs and ameliorate AL^10–12. By selectively binding to the native state of FL LCs over the misfolding transition state, these kinetic stabilizers increase the activation free energy for misfolding (or misfolding and misassembly, with or without endoproteolysis) and lower the population of proteotoxic species^13–14. Kinetic stabilizers are a category of pharmacological chaperones, which are drugs that correct the cellular folding of misfolding-prone proteins by native state binding^15–19. The kinetic stabilizer strategy should be complementary to plasma cell-eradication approaches for treating AL in most scenarios. For example, FL LC kinetic stabilizers could be employed to lower cardiotoxicity prior to plasma cell-eradication therapy, or be used to prevent relapse by stabilizing residual amyloidogenic LCs that are secreted by slowly reemerging plasma cell clones²⁰. In similar amyloid diseases, the small molecule kinetic stabilizer tafamidis has proven to be an effective strategy to ameliorate the transthyretin amyloidoses when used alone^{7, 21–23}.

Immunoglobulin light chains are a particularly challenging target for the design of kinetic stabilizers. FL LCs consist of an N-terminal variable (V_L) domain connected to a C-terminal constant (C_L) domain. FL LCs primarily exist as homodimers, often stabilized by a disulfide bond between the C_L domains⁵. Each AL patient has a unique FL LC sequence^24–25, with most of the sequence variation concentrated in the V_L domains. A viable kinetic stabilizer must be able to stabilize the majority of FL LCs, which have no known natural ligands and lack obvious ligand-binding site(s) based on most apo FL LC structures^{13, 26}. Thus, it was initially unclear whether a FL LC kinetic stabilizer exhibiting high affinity and selectivity over other blood proteins could be discovered. Early reports from Edmundson et al. showed that the V_L-V_L domain interface of a specific FL LC dimer can bind a variety of aromatic small molecules and peptides^27–28. However, this binding site largely comprises non-conserved residues of the complementary-determining region (CDR) loops in the V_L domains. Later, Brumshtein et al. showed that the V_L-V_L dimer interface can bind the small molecules, sulfasalazine and methylene blue, which also inhibit aggregation of an isolated V_L-V_L domain^{27, 29}. However, these small molecules do not stabilize FL LC dimers^{13, 30}.

We recently reported coumarin-based FL LC kinetic stabilizers, hereafter called “stabilizers”, exhibiting single-digit nanomolar potency in a protease-coupled fluorescence polarization (PCFP) assay ^{13, 31}. The PCFP assay reports on the ability of a stabilizer to stabilize an amyloidogenic FL LC dimer, called WIL-FL, against endoproteolysis by proteinase K^{13, 31}. These small molecule stabilizers do not inhibit proteinase K. Crystallographic studies of these stabilizers revealed that they can be partitioned into four components. Each stabilizer component binds a separate region of a FL LC dimer at the V_L-V_L domain interface. The “anchor substructure” binds the hydrophobic “anchor cavity”, the “aromatic core” binds the “aromatic slit”, the “linker module” binds the “solvent-exposed linker binding site”, and the “distal substructure” binds the “distal aromatic pocket” (Figure S1). These binding regions are distinct from the small molecule binding sites in the V_L-V_L domain interface that were reported in earlier studies^27–29.

Fortunately, the FL LC amino acid residues comprising the stabilizer binding site are highly conserved among AL-associated FL LCs, increasing the likelihood that a single, optimized stabilizer can potently stabilize a majority of the amyloidogenic FL LCs. Using the high-throughput screening hit 1 (Figure 1) as a template¹³, which only comprises a 7-diethylamino anchor substructure covalently attached to a coumarin aromatic core component, we previously reported that adding a carbamate (e.g., as found in 2), hydantoin, or spirocyclic urea (e.g., 3) linker module at the 3-position significantly increased stabilizer potency (up to 3,000-fold) compared to 1 (Figure 1)¹⁴. Herein, we synthesized and evaluated a series of 7-diethylamino coumarin-based kinetic stabilizers with novel amide and urea linker modules attached at the 3-position.

Figure 1.

Examples of coumarin-based LC stabilizers and their associated activities. For stabilizer 1, the 3-and 7-position carbons are labeled. The EC₅₀ is defined as the concentration of stabilizer that affords 50% maximal activity (i.e., proteinase K protection) for fluorescein-labeled WIL-FL (WIL-FL*, 5 to 20 nM dimer) using the protease-coupled fluorescence polarization assay. The T46L/F49Y variant replaces two residues in the stabilizer binding site to the residues most commonly found in AL LC sequences^34–35. Note that stabilizers 2 and 3 were referred to as 26 and (R)-83, respectively, in our previous study¹⁴.

The syntheses of coumarin amides 9 - 12 are described in Scheme 1. Briefly, commercially available 3-diethylamino phenol 4 was subjected to a titanium-mediated Pechmann condensation³² with diethyl 2-acetylsuccinate or diethyl 2-acetylpentanedioate to afford 3-position ethyl esters 5 and 6, which were hydrolyzed to carboxylic acids 7 and 8. These acids were coupled to a variety of amines to afford amides 9 - 12.

Scheme 1.

Synthesis of coumarin stabilizers 9 to 12 comprising an amide linker module^a; the R₁ substructures are defined in Table 1.

^a Reagents and conditions: (a) diethyl 2-acetylsuccinate or diethyl 2-acetylpentanedioate, ClTi(OiPr)₃, toluene, 110 °C, 24 h; (b) LiOH, H₂O, MeOH, THF, 23 °C, 16 h; (c) i. carbonyldiimidazole (CDI), DMF, 23 °C, 1 h, ii. R₁NH₂, DMF, 23 °C, 1 h.

Reversal of the orientation of the 3-position amide bond could be achieved by synthesizing 7-diethylamino coumarin-3-ethylamine 15 (Scheme 2). Briefly, 3-bromocoumarin 13 was converted to the Boc-protected ethylamine 14 using a Suzuki coupling³³. The Boc protecting group was removed to give the free amine 15. Primary amine 15 was reacted with activated carboxylic acids to yield amides 16 and 17, wherein the orientation of the amide bond is reversed relative to amides 11 and 12.

Scheme 2.

Synthesis of coumarin stabilizers 16 to 23 comprising amide or urea linker modules^a; the R₂ substructures are defined in Table 1.

^a Reagents and conditions: (a) potassium 2-(Boc-aminoethyl)trifluoroborate, Pd(dba)₂, RuPhos, Cs₂CO₃, toluene, H₂O, 90 °C, 16 h; (b) TFA, DCM, 23 °C, 1 h; (c) i. R₂CO₂H, carbonyldiimidazole (CDI), DMF, 23 °C, 1 h, ii. 15, DMF, 23 °C, 1 h; (d) i. CDI, DMF, 0 °C, 1 h, ii. R₂NH₂, DMF, 23 °C, 16 h.

Intermediate 15 also made it possible to synthesize ureas 16 - 23 (Scheme 2) via reaction with carbonyldiimidazole followed by displacement of the remaining imidazole employing primary amines.

In total, we synthesized and evaluated forty-eight 7-diethylamino coumarins substituted at the 3-position with linker module-distal substructure pairs that had not yet been explored. The data for representative analogues (1, 2, 9 – 12, 16 – 23) are shown in Table 1, and data for all analogues are summarized in Table S1. All data reported in Table 1 are the mean of at least three independent replicates ± one standard deviation. In the assay common to Table 1 and Table S1 (column 3 in both), unlabeled WIL-FL LC dimer (2.5 μM) was treated with kinetic stabilizer candidate (10 μM) and proteinase K (50 nM) for 2 h at 37 °C. This assay reports on the ability of stabilizer candidates to protect FL LC dimers from conformational excursions into protease-sensitive conformations^{13–14, 34}. The remaining FL LC dimer after proteolysis was quantified using size-exclusion chromatography, and kinetic stabilizer activity is reported as “fold protection”, which is the fraction of remaining FL LC dimer in the presence of stabilizer corrected for the residual FL LC dimer in the vehicle (1% DMSO) control¹⁴. This proteinase K sensitivity assay was performed on all 48 analogues synthesized. In the fourth column of Table 1, results from a “protease-coupled fluorescence polarization” (PCFP) assay are reported. Briefly, a K79C WIL-FL LC dimer with one fluorescein molecule attached to Cys79 per dimer (“WIL-FL*”, 10 nM) was treated with kinetic stabilizer candidate (5 nM to 100 μM) and proteinase K (200 nM) for 24 h at 23 °C. Fluorescence polarization (FP) was then measured to report on the presence of remaining WIL-FL* dimer, which has a higher FP signal compared to fluorescein-labeled proteolytic fragments that tumble faster in solution. This assay readily allows for determination of an EC₅₀ for each compound, which is defined as the concentration of kinetic stabilizer that affords 50% of the maximal fluorescence polarization signal. In the fifth column of Table 1, we also performed the PCFP assay on fluorescein-labeled T46L/F49Y LC dimer WIL-FL* (n.b., concentrations used are identical to the native WIL-FL* PCFP assay reported in column 4 of Table 1). These mutations replace two non-conserved residues in the aromatic slit and distal aromatic pocket, respectively, with the consensus residues present in most AL-associated FL LC dimer sequences^{14, 35}. We constructed dose-response curves using PCFP assay data from WIL-FL* and WIL-FL* T46L/F49Y, which were used to derive the EC₅₀ of each stabilizer for each fluorescein-labeled WIL variant (Figure S2).

Table 1.

SAR data on representative coumarin-linker module-distal substructure stabilizers 9 to 23.

Compound	R =	WIL-FL fold protection at 10 μM small molecule	WIL-FL* EC50 (nM)	WIL-FL* T46L/F49Y EC50 (nM)
1	H	0.28 ± 0.02	1630 ± 150	4130 ± 470
2		0.91 ± 0.04	69.4 ± 12	26.2 ± 2.1
9		0.28 ± 0.02	1830 ± 350	853 ± 140
10		0.39 ± 0.01	1750 ± 93	2660 ± 59
11		0.61 ± 0.01	634 ± 89	2110 ± 230
12		0.13 ± 0.02	6800 ± 870	> 10,000
16		0.17 ± 0.02	490 ± 16	513 ± 160
17		0.55 ± 0.06	1180 ± 160	1610 ± 50
18		0.69 ± 0.05	348 ± 70	381 ± 33
19		0.66 ± 0.05	462 ± 120	810 ± 6.5
20		0.50 ± 0.03	1040 ± 180	2050 ± 110
21		0.81 ± 0.05	140 ± 14	74.1 ± 5.6
22		0.74 ± 0.04	233 ± 29	377 ± 29
23		0.66 ± 0.03	111 ± 24	27.4 ± 7.2

For comparison, data for compounds 1 and 2 from our previous studies^13–14 are shown. Each data value is the mean of at least three independent replicates ± 1 standard deviation. In column 3, “fold protection” data from the proteinase K sensitivity assay using unlabeled WIL-FL dimer (2.5 μM) are reported. In columns 4 and 5, EC₅₀ data from the PCFP assay employing fluorescein-labeled (“*”) WIL-FL or WIL-FL T46L/F49Y dimer (10 nM), respectively, are reported. For full assay details, see the main text.

Our initial efforts focused on 7-diethylamino coumarins substituted at the 3-position with amide-based linker modules (e.g., 9 - 12 and reversed amide analogues 16 and 17, Table 1; for a complete list see analogues S1 to S16 in Table S1). We tested distal substructures including phenyl, pyridyl, or dimethylamine groups in anticipation of capturing stabilizing π-π or cation-π interactions with the conserved Y49’ residue, as we observed in our work on carbamate linker-based stabilizers¹⁴. Several of the analogues (9 to 12), exhibit modest increases in potency compared to screening hit 1 (Table 1). Stabilizer 11 was the most potent in this sub-series across two of the three assays, exhibiting a 0.33 increase in fold protection for unlabeled WIL-FL (0.61), a 2.6-fold potency increase for WIL-FL* (EC₅₀ = 634 nM) and two-fold potency increase for WIL-FL* T46L/F49Y (EC₅₀ = 2110 nM) compared to stabilizer 1 (WIL-FL fold protection = 0.28, WIL-FL* EC₅₀ = 1630 nM, WIL-FL* T46L/F49Y EC₅₀ = 4130 nM). We next evaluated two “reversed” amide analogues, comprising a phenyl or 3-pyridyl distal substructure (i.e., 16 and 17). Interestingly, stabilizer 16 exhibited noticeably higher potencies in the PCFP assays (WIL-FL* EC₅₀ = 490 nM, WIL-FL* T46L/F49Y EC₅₀ = 513 nM) than would be expected given its relatively poor activity in the single-dose (10 μM stabilizer) proteinase K sensitivity assay with unlabeled WIL-FL (fold protection = 0.17). However, stabilizer 16 achieved a lower maximal fluorescence polarization compared to compound 1, consistent with its activity in the single-dose assay at 10 μM stabilizer (Figure S3). Stabilizer 17 exhibited unremarkable activity considering all 3 assays.

We next focused on stabilizer candidates comprising 3-position simple urea linker modules and an aromatic distal substructure, exemplified by stabilizers 18 to 23 (Table 1; for a comprehensive list see stabilizers S17 to S33 in Table S1). These compounds are differentiated from our previous work on spirocyclic urea linker modules.¹⁴ Of this urea linker module series, stabilizer 21 comprising a meta-phenyl-imidazole distal substructure exhibited the highest activity in the single-dose proteinase K sensitivity assay (fold protection = 0.81) and was up to 56-fold more potent than stabilizer 1 in the PCFP assays (WIL-FL* EC₅₀ = 140 nM, WIL-FL* T46L/F49Y EC₅₀ = 74.1 nM). Stabilizer 21 was slightly less potent than the analogous carbamate stabilizer 2 (Figure 1; WIL-FL* EC₅₀ = 69.4 nM, WIL-FL* T46L/F49Y EC₅₀ = 26.2 nM) across all three assays. Stabilizer 23, which contains a 2-napthyl distal substructure, exhibited the highest potency among “simple” ureas in the PCFP assays (WIL-FL* EC₅₀ = 111 nM, WIL-FL* T46L/F49Y EC₅₀ = 27.4 nM), which is greater than would be expected by its activity in the single-dose assay (fold protection = 0.66).

Stabilizer 19 (Figure 2A), unlike the other urea stabilizers studied, could be crystallized in complex with JTO-FL. We were able to solve an x-ray structure at 2.1 Å resolution (Table 2, Figure S4). JTO-FL is a non-amyloidogenic, but crystallization-amenable FL LC dimer that is structurally very similar to amyloidogenic WIL-FL^{13, 31}. Stabilizer 19 comprises a 3-pyridyl distal substructure and exhibits moderate potency (WIL-FL* T46L/F49Y EC₅₀ = 810 nM). The diethylamino anchor substructure and coumarin aromatic core occupy the consensus anchor substructure cavity and aromatic slit, respectively, at the FL LC V_L-V_L domain interface, i.e., the same binding sites occupied in the previous six JTO-FL•coumarin crystal structures that we reported (Figure 2B) ^13–14. Even though 19 harboring a urea linker is structurally similar to the carbamate linker-based stabilizers previously reported, e.g., 2 (Table 1), the urea linker module and 3-pyridyl distal substructure of 19 do not occupy the same linker-binding site and distal aromatic pocket that were reported in our previous publications on carbamate, hydantoin, and spirocyclic urea linker module-based stabilizers^{14, 34}. Instead, the 3-substituted urea linker module–pyridine distal substructure pair bind a novel, solvent-exposed region nearby, but oriented opposite to the sites occupied by the carbamate linker–distal substructure pair (Figures 2C and 2D). This region contains a shallow cavity formed by a loop comprising residues Q37’ to I45’ (Figure 2E). Antibody residues are numbered according to the Kabat system, and the prime notation indicates the second polypeptide chain of the FL LC dimer. The urea N-H proximal to the coumarin of 19 may hydrogen-bond with the backbone carbonyl of P44’, as the distance between the amide proton and acceptor carbonyl oxygen is 1.9 Å. The pyridine distal substructure partially occupies a shallow cavity in this region, and it is not possible to conclusively model the position of the pyridine nitrogen at this resolution. While there are multiple possible hydrogen-bonding interactions between the FL LC and two ordered water molecules in this cavity, the pyridine nitrogen is unlikely to participate due to being at least 3.7 Å away from the closest ordered water, and any hydrogen bonds would likely contribute little to binding affinity due to the solvent-exposed nature of this region (Figure 2F). We examined the conservation of residues comprising this binding site among 347 AL-associated LC sequences (Figure 2G) ^34–35. Most residues are poorly conserved, appearing in fewer than 25% of LC sequences. However, occupancy of this site by a kinetic stabilizer substructure has the potential to increase stabilizer potency and fine-tune kinetic stabilizer physicochemical properties.

Figure 2.

Crystal structure of JTO-FL (gray) in complex with stabilizer 19 (blue, PDB: 7RTP). A) Line drawing of 19. B) Close-up of the binding site of 19 (in blue) emphasizing the diethylamino anchor substructure and coumarin aromatic core. The binding of carbamate stabilizer 2 (red, PDB: 7LMN) in complex with JTO-FL is overlaid. The prime label denotes residues of the second monomer comprising the LC dimer. C) Zoomed-out view of the binding site for 19, which binds at the V_L-V_L domain interface of the JTO-FL LC dimer. D) Zoomed-out comparison of the binding modes of 19 and carbamate stabilizer 2. E) Close-up of the binding site for 19 emphasizing the urea linker module and 3-pyridyl distal substructure. F) Potential hydrogen-bonding interactions (dashed lines) present in the binding site for 19. A possible hydrogen bond between a urea N-H of 19 and P44’ is indicated with a red dashed line and is approximately 1.9 Å in length. G) Conservation of the binding site utilized by 19. “Conservation” is defined as the percentage of 347 AL-associated LC sequences that the designated JTO-FL residues are found in those sequences.

Table 2.

Data collection and refinement statistics for the crystal structure of JTO-FL•19^a.

PDB code	7RTP
Data collection
Space group	P212121
(a, b, c) (Å)	63.64, 80.58, 94.64
(α, β, γ) (°)	90, 90, 90
Resolution range (Å)	47.32 – 2.09 (2.20 – 2.09)
Unique reflections	29,655 (4,230)
Completeness (%)	99.8 (99.0)
Rsym	0.12 (0.72)
Rpim	0.034 (0.21)
CC(1/2)	1.00 (0.92)
I/σ (I)	14.9 (3.7)
Redundancy	12.9 (12.4)
Wilson B factor (Å2)	24
Refinement
No. atoms
total	3681
protein	3284
water	357
buffer	10
ligand	30
Resolution range (Å)	47.37 – 2.09
No. reflections - work	28,112
No. reflections - free	1,409
Rwork	0.179
Rfree	0.238
RMS bond length (Å)	0.012
RMS bond angle (°)	1.68
Mean B value (Å2)
overall	34
protein	33
water	34
buffer	60
ligand	41
Ramachandran favored (%)	95.8
Ramachandran allowed (%)	100.0
Clash score	3.39

^aThe structure was determined from one crystal. For all statistics, values in parentheses apply to reflections in the highest-resolution shell.

In summary, we have described a series of coumarin-based FL LC dimer kinetic stabilizers comprising previously unreported amide and urea linker modules. The most potent stabilizers, 21 and 23, comprising urea linker modules, exhibit up to 56-fold and 150-fold increased potency, respectively, compared to 1. It is of note that some stabilizers (16 and 23) exhibit noticeably higher activity in the PCFP assays than would be expected by their weaker performance in the single-dose proteinase K sensitivity assay with unlabeled WIL-FL. We observed a similar apparent discrepancy with the stabilizer 2-(7-(diethylamino)-4-methyl-2-oxo-2H-chromen-3-yl)ethyl benzylcarbamate in our previous study¹⁴. This stabilizer exhibited modest activity in the single-dose proteinase K sensitivity assay (WIL-FL fold protection = 0.34) slightly higher than that of stabilizer 1 (fold protection = 0.28), but exhibited more than ten-fold improved potency than 1 in the PCFP assays. We are currently trying to better understand the origin of these differences.

Stabilizer 19 partially occupies a novel, solvent-exposed binding site, although it is unclear if other urea-based stabilizers preferably bind this region instead of the FL LC binding site occupied by the carbamate linker-distal substructure-based stabilizers. Occupying the newly observed binding site while concurrently occupying the carbamate linker-distal substructure consensus binding sites could be a strategy to increase binding affinity and selectivity¹⁴. However, the urea linker-binding site is poorly conserved among FL LC AL-associated sequences. Overall, our data further elaborate a potential small molecule binding site in amyloidogenic FL LC dimers which may be targeted to increase kinetic stabilizer potency.