Structural basis for the pro-amyloidogenic action and ligand binding of a novel W72R variant of human apolipoprotein A-I

Abstract

Apolipoprotein A-I (apoA-I), the major structural and functional protein of plasma high-density lipoprotein (HDL), mediates cholesterol removal from the body and has other beneficial properties. APOA1 gene mutations can cause diseases including human hereditary AApoAI amyloidosis, a life-threatening multiorgan disorder caused by apoA-I misfolding and amyloid deposition. We report esophageal AApoAI amyloidosis in a 66-year-old patient. Amyloid fibrils are comprised of a novel p.Trp96Arg (W72R) apoA-I variant identified by protein mass spectrometry and gene sequencing. Both variant and wild-type apoA-I circulate in plasma at low HDL levels typical of this dominant negative disorder. To understand why W72R substitution promotes amyloid formation, we generated recombinant full-length W72R and wild-type proteins and compared their structural, stability and ligand-binding properties using spectroscopic (circular dichroism, fluorescence, absorption) and biochemical approaches. High-resolution structures of apoA-I in native and in amyloid states were harnessed to propose structural underpinning for the pro-amyloidogenic effects of W72R substitution. W72R substitution slightly destabilized native lipid-free protein; decreased structural protection of amyloid-promoting regions; accelerated lipid solubilization by apoA-I and destabilized model lipoproteins; facilitated apoA-I binding to heparin (a proxy for heparan sulfate) and to type-I collagen; accelerated amyloid formation and potentially stabilized the amyloid structure. All these effects are pro-amyloidogenic. The results help explain the pro-amyloidogenic nature of this mutation, suggest apoA-I binding to extracellular matrix components such as heparan sulfate and collagen as therapeutic targets, and expand our knowledge of the clinical presentation and molecular basis of this complex disease.

Graphical Abstract

Introduction

Apolipoprotein A-I (apoA-I), a water-soluble lipid surface-binding protein of 243 a.a., is the major structural scaffold and functional ligand on plasma high-density lipoprotein (HDL) [1]. Synthesized from the APOA1 gene, ApoA-I in the form of HDL removes excess cholesterol from peripheral cells and has anti-inflammatory, antioxidant and other beneficial properties [2–4]. Plasma levels of HDL cholesterol correlate inversely with the risk of atherosclerosis and are regulated by genetic factors [5, 6]. Over 50 APOA1 gene mutations have been reported that can cause aberrant plasma lipid metabolism, cardiovascular disease and/or systemic AApoAI amyloidosis, a life-threatening disease that is difficult to diagnose and treat [7–13]. No effective pharmacotherapies are available for AApoAI amyloidosis. Current work aims to expand our understanding of the clinical presentation and molecular origins of this incurable disease to help guide its therapeutic targeting.

ApoA-I is secreted by the liver and the small intestine and is catabolized by the liver and kidney [14, 15]. Although the majority of circulating apoA-I is bound to HDL surface in a stable highly α-helical conformation, a small fraction (<5%) is released from HDL in a transient lipid-poor/free state, henceforth termed “free apoA-I” [1]. This structurally labile species either re-associates with lipids or is rapidly catabolized; alternatively, free apoA-I forms the protein precursor of amyloid [16–18] that deposits as fibrils in two forms of systemic human AApoAI amyloidoses, acquired and hereditary. In acquired amyloidosis, which is linked to aging, wild type (WT) full-length Met-oxidized apoA-I deposits as amyloid fibrils in the cardiovascular system, which can contribute to the development of atherosclerosis and calcific aortic valve disease [12, 19–21].

Hereditary AApoAI amyloidosis is an autosomal dominant disorder caused by APOA1 gene mutations. In this rare life-threatening disease, 9–11 kDa N-terminal fragments of the variant apoA-I deposit as amyloid fibrils in various organs (kidney, liver, heart, nerves, etc.) leading to organ damage [7, 9, 13, 22–27]. Over 20 AApoAI variants have been identified to-date; most variants contain single point substitutions, and most substitution sites fall in two residue segments, 26–60 (often associated with amyloid deposition in the kidney, liver and other organs) and 170–178 (often associated with cardiac amyloidosis) [24, 28,29]. APOA1 mutations outside these segments include D20Y variant associated with esophageal AApoAI [27].

Hereditary AApoAI amyloidosis is often a slow-developing disease with variable penetration and a heterogeneous clinical presentation that can vary from mild to severe and involve various organs [7, 9, 11, 23, 30]. This overlaps the clinical presentation of amyloidoses caused by other fibril-forming proteins [31]. An additional complication is that apoA-I often co-deposits with other amyloids, thus providing an “amyloid signature” protein [31]. Correct identification of the fibril-forming proteins is paramount for diagnosis and treatment; current amyloid typing approaches include immunohistochemistry, immunogold electron microscopy and mass spectrometry analysis of biopsied tissues [11, 32, 33].

The treatment options for AApoAI are limited to supportive strategies. Additionally, transplantation of the affected organ combined with liver transplantation to remove the source of the aberrant apoA-I are available for eligible patients [11, 30]. Triglyceride-lowering therapies have been proposed to diminish the release of free apoA-I from HDL [34]. To develop much-needed pharmacotherapies for AApoAI, one needs to improve our understanding of the clinical presentation of the disease and its molecular underpinnings, and to identify potential therapeutic targets.

Biophysical studies have shown that AApoAI-associated mutations can interfere with the protein-lipid interactions; perturb the α-helical structure in free apoA-I and decrease the structural protection in its amyloid-promoting regions (APRs); and promote apoA-I proteolysis in specific regions to release N-terminal fragments with enhanced amyloid-forming vs. degradation propensity [10, 13, 35, 36–38]. It is unclear whether apoA-I proteolysis releasing the aggregation-prone N-terminal fragments is an obligatory step preceding amyloid deposition in hereditary amyloidosis.

New insights have emerged from the atomic structures of human tissue-derived AApoAI amyloids determined recently using cryogenic electron microscopy (cryo-EM) for L90P and R173P variants [39]. Both variants had similar structures with well-ordered amyloid core spanning residues 14–73 flanked by disordered segments. This verified the central role of APRs in amyloid core formation and suggested that AApoAI-associated variants containing mutation sites outside this core form similar amyloid structure in vivo [39]. Despite these advances, the molecular underpinnings for the pro-amyloidogenic effects of many AApoAI-associated variants remain unclear and their interactions with amyloid-modulating factors other than lipids have not been explored.

Interactions of amyloidogenic proteins with the extracellular matrix (ECM) components modulate amyloid deposition in the ECM [29, 40]. Heparan sulfate (HS) is a major ECM component and an “amyloid signature” factor that co-deposits with diverse amyloids in vivo [31]. HS can act as a scaffold to promote amyloid formation in vitro [41]. Heparin, a highly sulfated mimetic of HS, provides a useful model for exploring protein-HS interactions. Such interactions with another HDL protein, serum amyloid A (SAA), are synergistic with amyloid formation and represent a therapeutic target for AA amyloidosis [42, 43]. Prior studies reported that unmodified WT apoA-I shows minimal heparin binding and does not readily form amyloid at pH 7.4, 150 mM NaCl [8, 41], though acidic pH and lower salt concentration promote amyloid formation by apoA-I and its fragments, especially in the presence of heparin [8, 44, 45]. Notably, full-length oxidized WT apoA-I binds heparin/HS and forms amyloid in vitro and in vivo in age-related amyloidosis [20, 41].

Besides HS, other ECM biopolymers interact with amyloidogenic proteins and potentially influence amyloid deposition. The major ECM protein, collagen type-I (henceforth termed collagen for brevity), is intimately associated with amyloid deposits in vivo; furthermore, collagen type-VI, a minor ECM component, has been implicated in longitudinal binding to patient-derived amyloid fibrils in other systemic amyloidoses [21, 46–49]. This begs the question: do HS and collagen influence amyloid deposition in hereditary AApoAI amyloidosis?

Recently we reported a new D20Y apoA-I variant featuring rare esophageal amyloidosis [27]. The current study describes a case of esophageal amyloidosis caused by a novel W72R apoA-I mutation. To determine the molecular basis for the pro-amyloidogenic effects of this mutation, we explore recombinant variant and WT proteins and harness the available atomic structures of apoA-I in native and in amyloid states. Additionally, we explore apoA-I interactions with heparin and collagen and test whether W72R substitution alters these interactions. The results suggest these interactions as potential therapeutic targets for AApoAI amyloidosis.

Results

Case report

A 66-year-old white male proband of Polish descent presented for evaluation at the Amyloidosis Clinic with gastrointestinal amyloidosis diagnosed by esophageal biopsy performed 6 years earlier. He had a long history of generalized abdominal pain, intermittent constipation and diarrhea, external hemorrhoids, rectal bleeding, hypertension, a cerebral vascular accident with mild facial droop and right lid ptosis at age 50, three-vessel coronary artery disease, myxomatous mitral valve disease, and a complete heart block with dual pacemaker placement at age 63. His medications included aspirin, propranolol, omeprazole, cholecalciferol, lisinopril, and rosuvastatin.

The medical history of proband’s maternal and paternal grandparents is unknown. His mother passed away at age 85 from unknown cause, and his father passed away at age 72 from complications following a stroke at age 67. His parental uncle and three parental aunts, all deceased, as well as his living parental first cousins had no known signs of AApoAI amyloidosis; one maternal uncle passed away in his 70s from unknown cause. Proband’s eldest brother passed away at age 67 from unknown cause and second oldest brother passed away at age 67 from pancreatic cancer. His sister, age 71, was speaking very scratchily and hoarsely over the last 1–2 years and intermittently lost her voice; she also had high blood pressure. The proband had no nieces, nephews, or children.

At age 54, twelve years prior to the initial evaluation, proband’s colonoscopy demonstrated a polyp and diverticulosis in the sigmoid colon; upper endoscopy was not performed. At age 59 the radiographic and ultrasound evaluation for chronic abdominal pain identified umbilical defect and rectus diastasis that did not correlate with clinical presentation. At age 60 the proband was diagnosed with esophageal amyloidosis during endoscopy ordered for abdominal pain and weight loss. The findings included many diminutive nodules with a diffuse distribution in the middle third of the esophagus (Figure 1A); scattered minimal inflammation characterized by erosions and erythema in the gastric fundus; diffuse mild inflammation and a single diminutive sessile polyp in the duodenal bulb; three diminutive polyps in the transverse colon, in the ascending colon and in the cecum; a few small-mouthed diverticula in the descending, transverse and ascending parts of the colon; and external and internal hemorrhoids. Esophageal biopsy demonstrated Congo Red-positive staining for amyloid. Gastric, duodenal and colonic tissue biopsies, as well as four biopsied polyps stained negative for amyloid. Over the next several years, multiple attempts were made to have the proband evaluated at the Amyloidosis Clinic.

Figure 1. Upper gastrointestinal endoscopy findings, and histological and ultrastructural analyses in the proband.

(A) Esophagoscopic view showing multiple diminutive nodules in the middle third of the esophagus (circled).

(B) Light microscopy image of Congo Red-stained esophageal biopsy showing amyloid deposits. Original magnification ×10.

(C) Areas in panel B viewed by polarized microscopy showing nodular amyloid deposits with characteristic apple-green birefringence. Original magnification ×10.

(D) Immunogold labeling of formalin-fixed paraffin embedded esophageal tissue demonstrating electron dense deposits with anti-apoA-I antibody. Original magnification, ×40,000.

(E) High-magnification view of the boxed area in panel D demonstrating the accumulation of numerous immunogold particles indicative of AApoAI amyloid deposition.

(F) Electron micrograph of esophageal biopsy showing haystack-like amyloid fibrils in the extracellular space. Original magnification ×80,000.

(G) Enlarged view of panel D showing dense extracellular accumulation of randomly oriented, non-branching amyloid fibrils with an average diameter of approximately 13.5 nm (n = 20 measurements). Original magnification, ×120,000.

At evaluation, the proband presented with no vital organ involvement. His blood pressure was 163/103 mmHg on lisinopril without orthostatic changes. His heart was normal with regular rhythm. The electrocardiogram was paced with a rhythm at a rate of 74 with occasional premature ventricular contractions, and echocardiogram demonstrated normal measurements with interventricular septum of 10 mm and left ventricular ejection fraction of 52.7%. Serum cardiac biomarkers were normal at 117 pg/mL for NT-proBNP, 52 pg/mL for BNP and <3 ng/L for high-sensitivity troponin-I, indicating no evidence of cardiac amyloidosis. His renal and hepatic functions were normal with creatinine at 0.8 mg/dL, estimated glomerular filtration rate at 90 mL/min/1.73 m², 24-hour urine partial collection with unmeasurable proteinuria, and alkaline phosphatase at 86 U/L. His neurologic evaluation demonstrated mild ptosis of the right eye, right sided facial weakness, and hyperreflexia resulting from the stroke. Cardio C-reactive protein was normal at 1.2 mg/L. The body mass index (29.4 kg/m²), the levels of non-fasting glucose (101 mg/dL) and uric acid (7.1 mg/dL) were elevated; hemoglobin A1C (5.7%) was in the pre-diabetic range; triglycerides (94 mg/dL) and cholesterol (152 mg/dL) were normal on 10 mg rosuvastatin that was started one year prior to the initial evaluation. The HDL levels were consistently low (39–45 mg/dL) during the nine-year period prior to the initial evaluation, which is typical of AApoAI amyloidosis [50]. The proband had no evidence of systemic plasma cell dyscrasia based on negative results of serum and urine immunofixation electrophoresis and normal results of free light chain assay, strongly arguing against light chain amyloidosis. Abdominal fat pad aspirate showed negative Congo Red staining.

Histological characterization of amyloid deposits

Histological section of the biopsy obtained in the middle third of the esophagus showed dense uniform nodular amyloid deposits replacing the submucosa (Figure 1B). Congo Red staining showed birefringence under polarized light, characteristic of amyloid (Figure 1C). Immunogold electron microscopy showed immunoreactivity with apoA-I antibody (Figure 1D–E). Electron microscopy of the congophilic material in esophageal tissue demonstrated extracellular non-branching fibrils ~13.5 nm in width, which is typical of amyloid fibrils (Figure 1F, G).

Genetic analysis

Direct DNA sequencing of exon 4 in the APOA1 gene of the proband revealed heterozygous T to C transition at the first base position of TGG codon 72, c.286T>C (Figure 2A), which encodes for p.Trp96Arg (W72R) variant protein. No mutations were noted in exon 2 that encodes residues 1–14 of the pre-peptide; exon 3 that encodes the rest of the pre-peptide, the pro-peptide, and apoA-I residues 1–43; or the rest of exon 4 that encodes residues 44–243. The genomic DNA was not available for testing proband’s family members. Bioinformatic analysis using PolyPhen-2 tool [51] predicted W72R substitution to have damaging effects on the apoA-I structure or function (Figure 2B).

Figure 2. Genetic analysis of APOA1 gene in proband and predicted damaging effects of the W72R replacement in apoA-I protein.

(A) Partial sequence chromatogram of the APOA1 gene in proband shows a single base nucleotide transition c.286T>C (arrow) at the first base position of the TGG codon in exon 4, leading to p. Trp96Arg (W72R) amino acid replacement. This replacement has not been previously reported in major genetic databases, including single nucleotide polymorphism database (https://www.ncbi.nlm.nih.gov/snp/?term=APOA1), Exome Aggregation Consortium (https://gnomad.broadinstitute.org/gene/ENSG00000118137?dataset=gnomad_r2_1), LOVD v.3.0 (https://databases.lovd.nl/shared/variants/APOA1?search_var_status=%3D%22Marked%22%7C%3D%22Public%22) or ClinVar (https://www.ncbi.nlm.nih.gov/clinvar) accessed in November 2025. (B) In silico analysis using PolyPhen-2 [51] predicts functional consequences of W72R substitution in apoA-I protein as “probably damaging” with a score range of 0.93 to 0.99 suggesting high probability of impaired protein function.

Top Down Mass Spectrometry Characterization of Circulating apoA-I Proteoforms

Analysis of proband’s plasma by individual ion mass spectrometry (I²MS) identified ten apoA-I proteoforms. Five major species corresponded to five pairs of proteoforms derived from the W72R variant and WT apoA-I (Figure 3A; Table S1). All W72R variant apoA-I proteoforms showed masses approximately 30 Da lower than their WT counterparts, consistent with tryptophan to arginine substitution (Table S1). The four principal proteoforms comprised full-length apoA-I, C-terminally truncated apoA-I, apoA-I glycated at Lys133, and pro-apoA-I containing a 7-residue portion of the N-terminal pre- and propeptide, ARHFWQQ (Figure 3A). W72R to WT apoA-I ratios were quantitatively reproducible across all proteoforms, with variation below 5%. Except for the glycated species, apoA-I variant proteoforms were consistently slightly more abundant in plasma compared to their WT counterparts suggesting higher levels of circulating variant apoA-I (Table S1). Low-level oxidation was detected predominantly in full-length proteoforms of both W72R and WT apoA-I, accounting for a minor fraction of the total apoA-I.

Figure 3. Proteomic identification of apoA-I in circulation and in amyloid deposits.

(A) I²MS spectrum of apoA-I proteoforms from the soluble human plasma proteoform fraction of proband’s plasma sample. The landscape is dominated by W72R followed by WT proteoforms. Distinct proteoforms are displayed in different colors and annotated as shown in Table S1.

(B) Proteomic detection of W72R apoA-I in proband’s amyloid deposits. Two Congo Red-positive areas of an esophageal biopsy were laser microdissected and analyzed using LC-MS/MS. The Scaffold readout lists top 10 proteins. Green boxes highlight the protein abundance based on total MS/MS spectral counts. Blue star marks the amyloid type-deterministic protein, apoA-I. Abundant apoA-I peptides were detected in both tested samples (538 and 748 spectra) with >95% probability. Yellow stars indicate “amyloid signature” proteins. No other known amyloid precursor proteins were detected.

(C) MS/MS spectrum shows variant apoA-I peptide, EQLGPVTQEFW-30, in amyloid deposits. Red asterisk marks the W72R substitution in the y-ion series. W-30 indicates the mass shift corresponding to the W to R replacement. The WT apoA-I peptide containing W72 was not detected in amyloid.

(D) Combined amino acid sequence coverage of apoA-I W72R variant protein (shown with 24-residue signal peptide and pro-peptide) in amyloid deposits. Peptides identified by MS (bold letters) accounted for 78.3% of the entire apoA-I sequence and for 88.9% of mature apoA-I sequence. R72 is in red. Last residue of the pre-peptide (A, black underline) and six residues of pre-peptide (RHFWQQ, red underline) were identified. The remainder of the signal peptide is underlined (dash line). In the deposits, the N-terminal peptide of mature variant protein (DEPPQS PWDRVKDLAT VYVDVLKDSG RDYVSQFEGS ALGK) was present along with the N-terminally extended peptide of variant protein (ARH FWQQDEPPQS PWDRVKDLAT VYVDVLKDSG RDYVSQFEGS ALGK) (15 and 2 spectra, respectively).

Abbreviations: I²MS, individual-ion mass spectrometry; LC-MS/MS, liquid chromatography-assisted tandem mass spectrometry; WT, wild-type.

Proteomic characterization of apoA-I amyloid fibrils

Liquid chromatography-assisted tandem mass spectrometry (LC-MS/MS) of esophageal amyloid deposits showed W72R variant but no WT apoA-I (Figure 3C–D), which is typical of hereditary AApoAI amyloidosis [11, 30]. N-terminal fragments of mature variant apoA-I were present along with those containing a seven-residue N-terminal extension, ARHFWQQ, from the signal peptide (Figure 3D). C-terminal peptides missed the last four residues. The sequence coverage of mature variant apoA-I in deposits was 88.9%, suggesting the deposition of full-length protein. Similarly, we found full-length D20Y apoA-I in esophageal amyloid deposits in hereditary AApoAI amyloidosis [27], though other studies reported the deposition of N-terminal 9–11 kDa apoA-I fragments in hereditary AApoAI amyloidosis [7, 23]. No other known amyloidogenic proteins were detected in deposits, except for the amyloid signature proteins apoE, SAP and apoA-IV (Figure 3B).

W72R substitution can decrease structural protection of amyloid-promoting regions in native apoA-I and stabilize the amyloid core

To understand the structural basis for the pro-amyloidogenic effects of W72R substitution, we first considered its location in lipid-bound apoA-I, which is the major protein form in circulation. Amino acid sequences of apolipoproteins contain 11/22-mer tandem repeats with high propensity to form amphipathic class-A α-helices. Large apolar faces of these α-helices can insert into the lipid surface while polar faces are exposed to solvent. In human apoA-I, W72 is located at the apolar-polar interface of such a helix; basic residues typically occupy such interfacial locations where they can interact favorably with the lipoprotein phospholipids [52]. Moreover, W72 is strictly conserved in mammals but is replaced with R72 in birds [53] and some fishes [54]. This helps explain why W72R substitution has been tolerated for over six decades in our patient.

Next, we considered W72 location in the native structure of lipid-free apoA-I, which is the protein precursor of amyloid. As a model, we used the X-ray crystal structure of human apoA-I residue fragment 1–185 [55] (Figure 4A). This fragment forms a crystallographic dimer containing two 4-helix bundles, which represents a structural intermediate between the lipid-free apoA-I monomer and the lipid-bound dimer. W72 is in a short α-helix at the bottom of the helix bundle (Figure 4A, B) and its solvent-accessible sidechain forms ring stacking interactions with W8 and F71 in the aromatic cluster. In W72R apoA-I (Figure 4B shows AlphaFold model), R72 sidechain can potentially form either cation-π interactions with W8 and/or F71, or an intrahelical salt bridge with E70. These stabilizing interactions can partially compensate for the destabilization of the aromatic cluster upon W72 removal.

Figure 4. W72 location in the native structure of apoA-I and in AApoAI amyloid fibrils.

(A) X-ray crystal structure of native lipid-free apoA-I (C-terminally truncated residue fragments 1–184, PDB ID: 3R2P, [55]. W72 side chains are shown. Dimer-forming molecules 1 and 2 are related by a 2-fold symmetry (vertical arrow). Amyloid-promoting regions (APRs) in residues 14–22 (major), 53–58 and 69–72 are color-coded [35]. K59 and R61 locations in the short helix 55–65 at the bottom of the 4-helix bundle are marked (deep blue).

(B) Zoomed-in view of the boxed area in panel A shows W72 in the aromatic cluster (left). Alphafold model shows the same area in W72R apoA-I (right). Circular arrows depict W/R72 sidechain rotation from the “closed” to the “open” conformation described in [38].

(C) CryoEM structure of AApoAI amyloid core (PDB ID:9PVY https://stamp.switchlab.org/database/Apolipoprotein%20A-I) [39]. Cross-sectional view down the fibril axis shows one molecular layer comprised of residues 14–73 from two dimer-forming apoA-I molecules. APRs are color-coded. W72 sidechains are shown with the nearby acidic residues, E70 and D73 (in red); in W72R variant, E70 may interact favorably with R72. Arrows mark basic sidechains, K59 and R61 (yellow), and aromatic sidechains, Y18 (green) and W72 (orange), exposed on the fibril surface. Insert: Cryo-EM map of cardiac L90P AApoAI amyloid shows unidentified densities (arrows) coordinated by the outward-facing sidechains of K59, R61 (yellow), Y18 (green), and W72 (orange) [39], which may represent amyloid cofactors bound laterally along the fibril sides, as illustrated for heparin/HS and for collagen triple helix (Figure 8D, F below).

Prior computational studies showed that in WT apoA-I, W72 sidechain adopts the predominantly “closed” conformation seen in the crystal structure. However, in several AApoAI variants including E34K, F71Y and L159R, W72 sidechain shows increased population of the “open” rotamers (shown by arrow in Figure 4B). This opening provides solvent access into the hydrophobic protein interior, including two amyloid-promoting regions (APRs) at the bottom of the helix bundle in residues 14–22 and 69–72, which harbor W72 (Figure 4A) [38]. Decreased structural protection of APRs is expected to promote amyloid formation [35, 37]. Since Arg sidechain is much more hydrophilic than Trp, R72 is expected to shift the population towards highly hydrated “open” conformations (Figure 4B). Therefore, W72R substitution is expected to decrease the structural protection in APRs in native lipid-free apoA-I.

Next, we considered W72 location in the structure of AApoAI amyloid (Figure 4C). Three similar structures have been determined recently by cryo-EM using fibrils extracted from the heart and kidney of AApoAI patients harboring L90P and R173P mutations [39]. The authors suggest that similar amyloid structures are formed by other apoA-I variants with mutations located outside the ordered amyloid core. AApoAI fibrils typically contain 9–11 kDa N-terminal fragments such as 1–93, yet the well-ordered amyloid core encompasses residues 14–73 that contain three APRs. Notably, two major APRs in residues 14–22 and 53–58 pack antiparallel against each other both in the native and in the amyloid structures (blue and teal in Figure 4A, C), facilitating a transition from the native structure to amyloid. W72, which is at the edge of the third APR (residues 69–72), faces solvent at the periphery of the amyloid core (Figure 4C). R72 is not only compatible with this location but can also form a salt bridge with the nearby acidic sidechain E70 to balance its uncompensated negative charge and thereby stabilize amyloid. This suggests that W72R can promote amyloidosis by stabilizing the amyloid core. In summary, W72R can potentially promote amyloidogenesis via a dual mechanism: decreased protection of APRs in native lipid-free apoA-I and stabilization of the amyloid core.

W72R substitution increases solvent access to the hydrophobic interior in native apoA-I

To experimentally probe the pro-amyloidogenic effects of W72R substitution, we generated recombinant full-length W72R and WT apoA-I as described in Methods. Each protein was generated in two forms, either mature or with a 7-residue N-terminal extension, ARHFWQQ, to model similar forms of W72R apoA-I found in amyloid deposits. In vitro studies of recombinant proteins detected no significant effects of the N-terminal extension on the conformation, stability or lipid interactions of either variant or WT apoA-I (Figure S1). Studies of recombinant mature proteins testing the effects of W72R substitution are described below.

First, we compared the conformation and stability of mature native lipid-free W72R and WT apoA-I. Far-UV circular dichroism (CD) spectra of these proteins largely superimposed, indicating similar secondary structures (Figure 5A insert). Near-UV CD spectra were also generally similar, with a negative peak at 280 nm indicating native-like aromatic packing in the helix bundle of the variant protein (Figure 5B). This conjecture is supported by the Trp emission spectra showing a similar wavelength of maximal fluorescence for the two proteins (Figure 5C insert). Decreased fluorescence intensity observed in W72R protein is consistent with the removal of one of four Trp in apoA-I. Limited tryptic digestion showed no detectable differences in the proteolytic susceptibility of the two proteins (Figure S2A). However, thermal unfolding monitored by far-UV CD (which indicates a decrease in the melting temperature from 60±1 °C for WT to 55±1 °C for W72R protein) and chemical unfolding monitored by the intrinsic Trp fluorescence (which shows a large shift in the unfolding curve to lower denaturant concentrations) consistently showed significant protein destabilization upon W72R substitution (Figure 5A, C).

Figure 5. Conformation and stability of recombinant mature apoA-I proteins in native lipid-free state.

W72R (blue) and WT (black) proteins are color-coded similarly in Figs. 5–7.

(A) Far-UV CD spectra (insert) and the heating and cooling data at 222 nm report on the protein secondary structure and stability.

(B) Near-UV CD spectra report on the aromatic packing in the 4-helix bundle.

(C) Intrinsic Trp fluorescence spectra (295 excitation, insert) and chemical unfolding monitored by Trp fluorescence.

(D) ANS binding monitored by fluorescence emission spectra (λ_ex = 360 nm).

(E) Förster resonance energy transfer from Trp to ANS (λ_ex = 295 nm).

(F) DMPC liposome clearance by apoA-I monitored by turbidity. The protein was added at t=0 min and the time course of liposome remodeling into smaller apoA-I:DMPC particles was monitored at 24 °C.

Fluorescent dye anilinonaphthalene sulphonate (ANS), which binds at solvent-accessible hydrophobic surfaces, showed significantly increased emission intensity and a 6 nm blue shift in the emission maximum, from 470±2.5 nm in WT to 464±2.5 nm in W72R, indicating deeper penetration of ANS into the hydrophobic interior of the helix bundle in variant vs. WT apoA-I (Figure 5D). Consistent with this conjecture, variant protein showed a fivefold increase in the Förster resonance energy transfer (FRET) from Trp to ANS (Figure 5E). Collectively, the results in Figure 5A–E show that, while lipid-free variant and WT proteins have very similar overall conformations, W72R substitution causes a small but significant destabilization of free protein and increases solvent accessibility of its hydrophobic interior.

ApoA-I binding to lipids involves helix bundle opening to expose the hydrophobic α-helical faces to lipid [55]. Therefore, increased solvent accessibility in the hydrophobic interior of the variant protein is expected to promote its binding to phospholipid surface. This conjecture is supported by experiments. The time course of clearance of model liposomes containing a phospholipid dimyristoyl phosphatidylcholine (DMPC) was monitored by turbidity. Spontaneous remodeling of large multilamellar vesicles into smaller lipoprotein particles was faster for W72R vs. WT apoA-I (Figure 5F). This finding is relevant to the major function of apoA-I that interacts with the lipids in the plasma membrane and remodels them into nascent HDL [1]. We conclude that W72R substitution does not impair the lipid binding function of apoA-I.

W72R apoA-I forms lipoproteins with reduced stability

Next, we explored how W72R substitution influences the properties of the recombinant lipoproteins, which provide models of nascent “discoidal” HDL. Complexes of DMPC and mature apoA-I, either W72R or WT, were prepared using thin film evaporation and were isolated as described in Supplemental Methods. Non-denaturing gel electrophoresis and size exclusion chromatography (SEC) showed different migration patterns for lipid-free W72R and WT proteins (Figure 6A, B), which probably reflected differences in their charge and solubility. Importantly, both proteins formed similar-size complexes (~12 nm) with DMPC (Figure 6A, B), which is commensurate with the HDL size.

Figure 6. Structure and stability of apoA-I:DMPC lipoproteins containing either W72R or WT mature apoA-I.

(A) Non-denaturing polyacrylamide gel electrophoresis (PAGE) of lipid-free proteins and their complexes with DMPC.

(B) SEC of lipid-free proteins (bottom) and of their complexes with DMPC (top).

(C) Conformation and stability of apoA-I:DMPC complexes monitored by CD. Far-UV CD spectra of WT and W72R apoA-I:DMPC complexes fully overlap (insert). CD melting data recorded during heating and cooling at a rate of 70 deg/hour. Arrows indicate the direction of temperature changes.

Far-UV CD spectra of these apoA-I:DMPC complexes containing either W72R or WT apoA-I fully overlapped, indicating similar conformations of these proteins on model lipoproteins (Figure 6C insert). Limited proteolysis suggested similar proteolytic susceptibility of these complexes (Figure S2B). The heating data recorded by CD shifted to lower temperatures by approximately 4 °C for the W72R-containing complexes, indicating their reduced stability (Figure 6C). This implies that, despite the relative ease of the helix bundle opening in W72R apoA-I, which is a rate-limiting step in model lipoprotein formation (Fig. 5F), enhanced release of W72R apoA-I from lipoproteins during thermal denaturation causes overall lipoprotein destabilization upon W72R substitution. Trp sidechains often assume interfacial positions anchoring the protein into lipid monolayers; this may contribute to the destabilizing effect of the W72R substitution on apoA-I:DMPC complexes. Together, the results in Figure 6 show that W72R substitution causes no detectable changes in the overall assembly and protein conformation of model lipoproteins but reduces their stability. These findings are consistent with low levels of plasma HDL found in our patient.

W72R substitution facilitates heparin and collagen binding to apoA-I

Next, we explored apoA-I–heparin interactions. Heparin affinity chromatography and dose-dependent enzyme-linked immunosorbent assay (ELISA) revealed that, unlike native lipid-free WT apoA-I that shows minimal binding, W72R variant binds heparin with high capacity and near-nanomolar affinity (Figure 7A, B). Similarly, apoA-I:DMPC complexes showed heparin binding for W72R but not for WT protein. (Figure 7C, D). Both methods suggest similar heparin binding affinity for lipid-free and DMPC-bound W72R apoA-I, but lipid-free protein showed greater binding capacity, suggesting that apolipoprotein-lipid binding interferes with the heparin binding.

Figure 7. Heparin and collagen binding by apoA-I in the native free state and in apoA-I:DMPC complexes.

W72R variant (blue) or WT (black) proteins in the form of lipid-free apoA-I or apoA-I:DMPC complexes (as indicated in the panels) were used in the binding studies to heparin (A-D) and to collagen (E, F). Heparin binding was assessed by affinity chromatography (A, C) and ELISA (B, D). Collagen binding was assessed by ELISA (E, F). ELISA data show mean ± SD of technical triplicates from three independent experiments (n=3). Solid lines show data fitting using a one-site specific binding model (R² ≥ 0.99 for heparin; R²≥0.97 for collagen). Dose-dependent ELISA showed saturable binding for heparin. For collagen, the binding saturation was not reached under conditions explored limiting the fitting accuracy.

This conjecture is consistent with studies of other plasma and model lipoproteins showing that heparin/HS binding promotes apolipoprotein release from the lipoprotein surface and is synergistic with protein misfolding in SAA [42, 43]. To test whether such binding promotes apoA-I release from model lipoproteins in the physiological pH range of 5.5–7.4, we used SEC to explore apoA-I:DMPC complexes with or without heparin. W72R apoA-I:DMPC complexes released a protein fraction that increased upon addition of heparin and/or acidification; in contrast, WT apoA-I:DMPC complexes showed no protein release in the absence of heparin and some release in the presence of heparin at pH 5.5 (Figure S3). These results confirm that heparin binding to model lipoproteins containing an amyloidogenic apoA-I variant, along with acidic pH, promotes protein release to generate free protein precursor of amyloid. We conclude that W72R substitution promotes heparin binding by lipid-bound and especially by lipid-free apoA-I; moreover, heparin binding shifts the population distribution from lipid-bound to lipid-free protein, which is pro-amyloidogenic.

How does W72R substitution in apoA-I facilitate heparin binding? The binding is generally driven by Coulombic interactions between the basic protein residues and heparin/HS sulfates and carboxylates [41, 56]. For example, fish apoA-I, which contains R72 and several basic residues nearby, binds heparin with micromolar affinity via this basic patch, which was proposed to promote apoA-I accumulation in the ECM [54]. We hypothesize that local positive charge on the surface of human W72R apoA-I contributes to heparin binding. Moreover, destabilization of the short α-helix 55–65 at the bottom of the helix bundle upon W72R substitution may release K59 and R61 sidechains for heparin binding (Figure 4A). Computational studies supporting this idea are described below.

Next, we explored apoA-I–collagen interactions. Dose-dependent ELISA showed minimal binding by native lipid-free WT apoA-I, while lipid-free W72R binds collagen with much greater capacity and apparent near-nanomolar affinity (Figure 7E). Free apoA-I and apoA-I:DMPC complexes showed similar dose-dependent binding (Figure 7E, F) suggesting that association with lipids only slightly decreases the collagen binding capacity of apoA-I.

Taken together, our results suggest that W72R substitution, which decreases the protein and lipoprotein stability (Figure 5A, C, Figure 6C) and increases solvent accessibility of the hydrophobic protein interior (Figure 5D), enhances the exposure of cryptic binding sites in apoA-I thereby promoting the binding of diverse ligands including lipids (Figure 5F), heparin (Figure 7A–D), and collagen (Figure 7E, F).

W72R substitution facilitates amyloid formation that enhances heparin and collagen binding

To test how W72R substitution influences amyloid formation in vitro, free proteins were incubated for two weeks under amyloid-promoting conditions (37 °C, pH 6.0 with stirring). Fluorescence of thioflavin T (ThT), a diagnostic dye whose emission increases upon binding to amyloid-like structure, indicated formation of such structure in W72R but not in WT apoA-I (Figure 8A). Additionally, far-UV CD spectra showed an α-helix to β-sheet conversion in variant protein upon incubation, while WT protein showed a decreased helical content (Figure 8B). These results indicate that W72R substitution promotes amyloid formation by recombinant full-length apoA-I in vitro, consistent with the finding of full-length W72R apoA-I in proband’s amyloid deposits.

Figure 8. Amyloid formation by apoA-I bolsters heparin and collagen binding.

The proteins, W72R (blue) or WT (black) apoA-I, were incubated under amyloid-promoting conditions for two weeks. (A) Formation of amyloid-like structure was assessed by ThT binding/fluorescence. Spectrum of ThT alone is shown as negative control.

(B) Far-UV CD spectra of W72R and WT apoA-I before (0 hours, solid lines) and after incubation (2 weeks, dashed lines).

(C) Single-point ELISA (using 20 μg apoA-I) shows apoA-I binding to heparin-coated plates measured before and after protein incubation.

(D) Heparin binding to the AApoAI fibril structure predicted using ClusPro. Fibril side view shows a representative docking pose of heparin tetrasaccharide (yellow) coordinated by sidechains of K59 and R61 (blue). ApoA-I molecules in amyloid layers are numbered. Fibril z-axis is marked by arrow.

(E) Single-point ELISA using 20 μg apoA-I shows protein binding to collagen-coated plates measured before and after protein incubation. Data in panels C and E represent mean ± SD of technical triplicates from each of three independent experiments (n = 3); *p < 0.05; ***p < 0.001.

(F) Collagen binding to the AApoAI fibril structure. A 30-residue collagen peptide in a triple-helical conformation (PDB ID: 1CAG) was used. Side view shows a representative docking pose coordinated by the Y18 aromatic array. Amyloid layers are numbered; arrow marks fibril z-axis.

Next, we compared heparin binding by apoA-I before and after incubation under amyloid-promoting conditions. Single-point ELISA showed a nearly 2-fold increase in heparin binding upon incubation of each protein; notably, the binding was much greater for W72R vs. WT apoA-I (Figure 8C). This suggests that heparin binding increases upon formation of amyloid-like structure by W72R apoA-I. A similar trend was observed for apoA-I–collagen binding, with even greater difference between W72R and WT proteins (Figure 8E). Taken together, these results suggest that formation of amyloid-like structure in apoA-I bolsters its binding to heparin, collagen and, potentially, other biopolymers.

Lastly, we performed in silico binding studies of ECM biopolymers to mature amyloid fibrils using the ClusPro server. The cryo-EM structure of cardiac L90P AApoAI fibrils [39] was used as a model. First, we docked heparin tetrasaccharide, a heparin unit from the ClusPro library. All predicted docking poses were oriented along the fibril side and were coordinated by the outward-facing K59 and R61 sidechains from multiple fibril layers (Figure 8D). This was not surprising since K59 and R61 comprise the sole positively charged array on the fibril surface; such basic arrays form the preferred binding sites for polyanions including heparin/HS [41, 56]. Since uncompensated charge arrays are highly energetically unfavored due to intra-and interlayer repulsion in amyloid, HS binding at this site is expected to stabilize amyloid.

Docking of a model 30-residue triple-helical collagen peptide to the amyloid structure predicted longitudinal binding at a surface site coordinated by Y18 (Figure 8F); another potential binding site near W72, which was absent from the W72R variant, was not considered. Our prediction is consistent with recent reports suggesting longitudinal binding of collagen to the aromatic arrays on the surface of other tissue-derived amyloids [47, 48].

Notably, in the three available structures of AApoAI amyloids (including cardiac L90P AApoAI amyloid and cardiac and renal R173P AApoAI), both dimer-forming molecules consistently showed unidentified additional densities in the cryoEM maps near the outward-facing K59 and R61 sidechains [39] (Figure 4C insert, yellow arrows). Our docking results suggest that these densities potentially represent bound polyanions such as HS. Even larger extra densities are seen near W72 and, particularly, Y18 (Figure 4C insert, orange and green arrows, respectively). Collagen binding at these sites may potentially contribute to these densities, though binding of other ligands is also possible.

In summary, affinity chromatography and ELISA show that W72R substitution facilitates heparin binding by lipid-bound and, particularly, by free apoA-I (Figure 7A–D), and that amyloid formation by apoA-I promotes heparin binding (Figure 8C) and may be synergistic with it. We propose that heparin binding can promote amyloid deposition via several mechanisms. Firstly, lipoprotein binding to heparin/HS induces apolipoprotein release from the lipid surface (Figure S3), generating the lipid-free protein precursor of amyloid, like that observed in SAA [42, 43]. Secondly, the binding is expected to bolster local accumulation of W72R apoA-I favoring its aggregation in the ECM. Moreover, heparin/HS is predicted to bind longitudinally to amyloid fibrils (Figure 8D) acting as a polyanionic scaffold that accelerates the formation of amyloid and stabilizes its basic arrays [41].

Furthermore, ELISA showed that W72R substitution facilitates collagen binding by lipid-bound and, particularly, free apoA-I (Figure 7E, F), while docking results suggest longitudinal binding of the collagen triple helix to aromatic arrays on the amyloid surface (Figure 8F). A similar binding mode was suggested for other tissue-derived amyloids [47, 48].

3. Discussion

W72R is a novel AApoAI variant associated with unusual esophageal amyloid deposition

We report a W72R apoA-I variant linked to esophageal AApoAI amyloidosis in a 66-year-old male proband. This mutation is novel, as the heterozygous APOA1 single base c.286T>C nucleotide transition was absent from the major genetic databases. The mutation site is unusual, as most known AApoAI mutations fall into residue segments 26–60 and 170–178 [7, 36]. The origin of the mutation in our proband is unknown as no parental DNA was available for the analysis. Many reported cases of AApoAI amyloidosis have negative family history suggesting spontaneous de novo APOA1 mutations or incomplete penetrance [30]. Consistent with low HDL levels reported in other AApoAI cases [7, 57, 58], the proband had low HDL (~40 mg/dL) for at least nine years prior to assessment. Like in other reported AApoAI cases [27], amyloid deposits contained only variant but no WT protein, while circulating apoA-I showed both W72R and WT apoA-I. These findings are consistent with the dominant negative inheritance typical of hereditary AApoAI amyloidosis. Interestingly, amyloid deposits contained W72R apoA-I both in mature form and with a 7-residue extension from the signal peptide. No significant effects of this N-terminal extension on the conformation, stability and lipid binding of the recombinant proteins were detected (Figure S1). Another interesting finding was the presence of full-length apoA-I in amyloid deposits (Figure 3). In contrast, other studies of hereditary AApoAI amyloidosis report the deposition of N-terminal 9–11 kDa apoA-I fragments [23]. The only other finding of full-length apoA-I in hereditary AApoAI amyloidosis was our recent study of a novel D20Y variant involved in esophageal amyloidosis [27]. These findings, taken together with prior studies [20, 35], suggest that, similar to non-hereditary amyloidosis, proteolytic cleavage of apoA-I to release the misfolding-prone 9–11 kDa N-terminal fragment may not be necessary for amyloid deposition in hereditary AApoAI. amyloidosis.

Similar to prior studies of esophageal amyloidosis [30], this patient had clinically asymptomatic disease until the sixth decade when amyloid deposits were incidentally found only in the esophagus, but not in any other parts of the gastrointestinal tract. Over the next six years, no new organ involvements were noted, indicating a slowly progressing disease. In vitro and in silico studies of W72R and WT apoA-I help explain some of these clinical findings and provide a structural basis for amyloid deposition by this novel apoA-I variant.

Molecular origins of the pro-amyloidogenic effect of W72R substitution

Our results suggest that W72R substitution can exert its pro-amyloidogenic effects through several mechanisms. First, the substitution slightly destabilizes full-length native lipid-free apoA-I (Figure 5), particularly in sensitive APRs (Figure 4). This generally favors protein misfolding and is consistent with amyloid formation by free full-length variant vs. WT apoA-I in vitro (Figure 8A, B). Second, W72R substitution destabilizes model lipoproteins (Figure 6C), promoting protein release in a labile free state that is the precursor of amyloid [16]. Third, W72R substitution facilitates the binding of apoA-I to heparin (Figure 7A–D) and to collagen (Figure 7E, F), perhaps due to greater access to the cryptic binding sites in native apoA-I; the binding of both biopolymers further increases upon amyloid formation by W72R apoA-I (Figure 8C, E). Fourth, W72R substitution is expected to stabilize the amyloid core in apoA-I by improving local charge balance on the fibril surface (Figure 4C). These combined effects help explain why this novel mutation is associated with amyloidosis. Furthermore, decreased stability of model lipoproteins containing W72R vs. WT apoA-I is consistent with low plasma HDL levels found in our patient, which may promote atherogenesis.

Some of these properties differ from those of another variant associated with esophageal amyloidosis, D20Y apoA-I [27]. Unlike W72R, D20Y substitution at pH 7.4 causes minimal or no changes in stability of lipid-free and lipid-bound apoA-I, no significant changes in lipid clearance kinetics, and very slow loss of protein secondary structure upon incubation under amyloid-promoting conditions. In fact, all major structural and stability changes in D20Y apoA-I occur at acidic pH. This is consistent with the deposition of D20Y apoA-I in Barrett’s esophagus that is subject to acidic bursts, unlike esophagus of the current study. Furthermore, both W72R and D20Y substitutions are located in the amyloid core but are potentially compatible with its known structure (Fig. 4C). Unlike W72R, D20Y is not expected to stabilize amyloid since D20 replacement with Y is expected to replace D20-K23 surface salt bridge with a Y20-K23 π-cation interaction. Both substitutions decrease the negative charge on apoA-I and both promote its binding to heparin.

Extracellular matrix interactions in amyloid deposition

Our analysis of apoA-I, taken together with prior studies of heparin interactions with SAA [42, 43], suggest that heparin/HS binding by these and other apolipoproteins is synergistic with amyloid formation. In fact, pro-amyloidogenic modifications such as oxidation [41] or W72R substitution (Figure 7A–D) facilitate heparin binding to apoA-I, while heparin binding promotes protein release from the lipoproteins (Figure S3) and favors amyloid formation perhaps via the scaffolding mechanism (Figure 8D). A similar mechanism has been proposed for SAA-heparin interactions [56]. Therefore, like SAA, apoA-I interactions with heparin may provide a potential therapeutic target for AApoAI amyloidosis if such interactions are experimentally confirmed for other amyloidogenic apoA-I variants. This mechanism may perhaps be extended to apoA-I interactions with other ECM biopolymers such as collagen, as suggested by our limited collagen binding studies to variant and WT apoA-I (Figure 7E, F; Figure 8E, F). If so, apoA-I interactions with various biopolymers may promote amyloid deposition in the ECM, while blocking these interactions may delay or halt this pathogenic process.

Direct role of ECM, particularly HS and collagen components, in amyloid deposition is consistent with endoscopic and histological findings in the proband. Dense nodular amyloid deposits replacing the submucosa of the middle third of the esophagus (Figure 1A–C) strongly suggest ECM as the site of amyloid deposition in this patient. Fibrillar collagen is a dominant structural component of the esophageal submucosal ECM, as observed by imaging of squamous esophageal tissue where collagen is concentrated in the connective tissues of the lamina propria and submucosa [59]. Similar to other submucosal connective tissue compartments, the esophageal submucosa contains HS proteoglycans [60], which actively contribute to amyloid deposition by various proteins [41]. Therefore, the submucosal ECM presents an environment rich in both fibrillar collagen and HS. W72R substitution in apoA-I greatly increased the binding to these ECM components (Figures 7, 8). This binding may contribute to the retention of the variant protein within the HS-rich submucosal ECM, with HS providing a scaffold for amyloid fibril formation. Collagen binding adds a second layer of ECM anchoring, consistent with the established role of fibrillar collagen in stabilizing amyloid deposits and protecting them from clearance [46, 47, 61]. Together, these interactions suggest a molecular basis for amyloid localization to the submucosal compartment of the esophagus in this patient, while the underlying cause of esophageal apoA-I deposition in this case remains unknown.

Conclusions

This study reports a novel amyloidogenic W72R variant with an unusual location of the substitution site in the apoA-I structure. Other unique features include only esophageal involvement and the deposition of variant apoA-I in full-length mature and immature forms, as opposed to 80- to 100-residue N-terminal fragments reported in most other cases of hereditary AApoAI amyloidosis. The molecular basis for the misfolding of W72R apoA-I is proposed. Binding by variant apoA-I to ECM components, such as HS and collagen, is suggested as a potential therapeutic target in AApoAI amyloidosis.

Methods

Clinical data and sample collection

The study was conducted in accordance with the Declaration of Helsinki. Informed consents for data and sample collections were obtained upon permission from the Institutional Review Board at Boston Medical Center and Mayo Clinic. Additional details are provided in the Supplementary Appendix.

Histological, immunoelectron microscopic, genetic, and proteomic analyses

Esophageal biopsy was evaluated for amyloid by Congo Red staining imaged by light microscopy in bright and polarized light. Amyloid typing was performed by immunogold electron microscopy and LC-MS/MS using the formalin-fixed paraffin-embedded esophageal tissue biopsy as previously reported [62]. The APOA1 gene was screened for mutations by direct DNA sequencing. ApoA-I proteoforms were analyzed in soluble human plasma proteoforms fraction via acetonitrile precipitation followed by I²MS [63]. Details are provided in the Supplementary Appendix.

Preparation and characterization of recombinant proteins and lipoproteins

Recombinant full-length W72R and WT apoA-I proteins with and without the 7-residue N-terminal extension, ARHFWQQ (four proteins in total), were generated using an E. coli-based expression system and purified to >95 % purity as previously described [27]. The N-terminal 6xHis tag, which improves protein solubility but has no significant effect on the structure, stability or lipid binding properties of apoA-I, was retained. Each protein migrated as a single band on SDS PAGE; correct protein folding was verified by CD spectroscopy.

Model apoA-I:DMPC lipoproteins were prepared by thin film evaporation. Biophysical studies of recombinant proteins and lipoproteins using far- and near-UV CD, fluorescence and absorption spectroscopy, gel electrophoresis, SEC, affinity chromatography, ELISA, and limited proteolysis, as well as molecular docking of heparin and collagen to the atomic structure of AApoAI amyloid are described in the Supplemental Appendix.

RESEARCH HIGHLIGHTS.

• A novel W72R variant of human apoA-I is linked to unusual esophageal amyloidosis

• The mutation site location and full-length apoA-I deposition in amyloid are unusual

• W72R substitution destabilizes free and lipid-bound apoA-I promoting the misfolding

• W72R apoA-I binding to heparan sulfate and collagen type-I promotes misfolding

• ApoA-I binding to extracellular matrix may represent a potential therapeutic target

Abbreviations:

AApoAI

apoA-I amyloidosis

ANS

8-anilino-1-naphthalenesulfonic acid

apoA-I

apolipoprotein A-I

CD

circular dichroism

cryo-EM

cryogenic electron microscopy

DMPC

dimyristoyl phosphatidylcholine

ECM

extracellular matrix

ELISA

enzyme-linked immunosorbent assay

FRET

Förster resonance energy transfer

HDL

high-density lipoprotein

LC-MS/MS

liquid chromatography-assisted tandem mass spectrometry

ThT

thioflavin T

WT

wild type

Supplementary Appendix

Provided are details of clinical data and sample collections; histological, immunoelectron microscopic, genetic, and proteomic analyses; preparation of recombinant proteins and lipoproteins, and their biochemical, biophysical and structural analyses including their interactions with heparin and collagen.

Data availability

The gene sequence of the W72R apoA-I variant has been deposited in the ClinVar database (accession No. SCV007495699.1;

https://www.ncbi.nlm.nih.gov/clinvar/variation/4812879/?term=SCV007495699). Additional information is available from the corresponding authors upon request.