D25V apolipoprotein C-III variant causes dominant hereditary systemic amyloidosis and confers cardiovascular protective lipoprotein profile

Apolipoprotein C-III deficiency provides cardiovascular protection, but apolipoprotein C-III is not known to be associated with human amyloidosis. Here we report a form of amyloidosis characterized by renal insufficiency caused by a new apolipoprotein C-III variant, D25V. Despite their uremic state, the D25V-carriers exhibit low triglyceride (TG) and apolipoprotein C-III levels, and low very-low-density lipoprotein (VLDL)/high high-density lipoprotein (HDL) profile. Amyloid fibrils comprise the D25V-variant only, showing that wild-type apolipoprotein C-III does not contribute to amyloid deposition in vivo. The mutation profoundly impacts helical structure stability of D25V-variant, which is remarkably fibrillogenic under physiological conditions in vitro producing typical amyloid fibrils in its lipid-free form. D25V apolipoprotein C-III is a new human amyloidogenic protein and the first conferring cardioprotection even in the unfavourable context of renal failure, extending the evidence for an important cardiovascular protective role of apolipoprotein C-III deficiency. Thus, fibrate therapy, which reduces hepatic APOC3 transcription, may delay amyloid deposition in affected patients.

T here is growing evidence that apolipoprotein C-III (apoC-III), a small exchangeable apolipoprotein carried in the circulation by VLDL and HDL 1 , is a master regulator of plasma TG homeostasis and downregulation of its expression has a cardiovasculo-protective effect 2 . ApoC-III inhibits the activity of lipoprotein lipase (LPL), the enzyme catalysing the first and rate-limiting step of TG-rich lipoprotein catabolism 3 , and also impairs uptake of TG-rich lipoproteinremnants by the low-density lipoprotein (LDL)-receptor (LDL-R), increasing remnant residence time in the circulation 4 . ApoC-III is a glycosylated apolipoprotein composed of 79 amino acids, synthesized mainly in the liver, and existing in three isoforms, apoC-III 0 , apoC-III 1 and apoC-III 2 with, respectively, 0, 1 and 2 residues of sialic acid per molecule, but little is known about sialylation of apoC-III in disease states. Heterozygosity for the rare null allele of APOC3, R19X, has been associated with high levels of plasma HDL cholesterol (HDL-C), low levels of plasma TG and reduced incidence of ischaemic cardiovascular disease (CVD) in the Amish population 5 . Subsequently, genome-wide association studies also identified the Arg19Stop variant in other population isolates with a strong atheroprotective phenotype 6 . More importantly, a recent general population-based study confirmed the association of APOC3 deficiency with an atheroprotective lipid profile and clinical protection against ischaemic CVD 7,8 .
In its lipid-bound state, apoC-III is composed of six amphipathic a-helices, the structural motif shared by all exchangeable apolipoproteins that confers their structural stability 9 . Conversely, apolipoproteins in lipid-free form have lower conformational stability and some, such as native and/or variants of apoA-I (ref. 10), apoA-II (ref. 11) and apoA-IV (ref. 12), are prone to self-assembly in vitro and in vivo as amyloid fibrils. Recombinant wild-type apoC-III forms insoluble aggregates with cross-b structure in vitro, reminiscent but not typical of amyloid fibrils 13 . Amyloid fibril formation is associated with a wide variety of diseases, and the hereditary monogenic forms of human amyloidoses, although rare, provide unique insight into mechanisms of protein misfolding and fibrillogenesis as recently highlighted by the discovery of the amyloidogenic variant of b2-microglobulin 14 .
Here we report a French family with severe renal amyloidosis and hypotriglyceridemia, both caused by a novel D25V apoC-III variant. Remarkably, this variant induced a favourable cardiovascular lipoprotein profile, even in the context of chronic kidney disease (CKD). The D25V protein is highly fibrillogenic in its lipid-free state, and forms amyloid fibrils under physiological conditions in vitro. This study provides new insights into the genetics of apoC-III and into VLDL and HDL metabolism in health and disease, illustrating the increasing importance of apoC-III in diverse biological processes.

Results
Patients. The proband (III.3 in Fig. 1a) presented at age 51 years with sicca syndrome and hypertension. Two years later, he was diagnosed with CKD; his serum creatinine level was 1.9 mg dl À 1 with modest proteinuria (40 mg dl À 1 ). Salivary gland and renal biopsies showed amyloid deposits. A search for an underlying monoclonal gammopathy was negative, and plasma concentrations of serum amyloid A protein (SAA) and C-reactive protein were in the normal range. The proband reached end-stage renal disease (ESRD) aged 56 years and received a cadaveric kidney transplant uneventfully. 123 I-labelled serum amyloid P component (SAP) scintigraphy, undertaken just before renal transplantation in the proband III.3 (Fig. 1b) and in affected member IV.3 (Fig. 1c), showed a small total body amyloid load in the spleen and kidneys, at which point a fat aspirate demonstrated extensive amyloid. Four years after renal transplantation, graft function was normal (serum creatinine 0.88 mg dl À 1 ), without detectable proteinuria. There was a strong family history of amyloidosis ( Fig. 1a) with all affected members suffering first from sicca syndrome and later developing mildly proteinuric CKD. The proband's grandmother (patient I.2) and aunt (patient II.1) had suffered from sicca syndrome for many years before developing systemic amyloidosis and had died aged 74 and 80 years, respectively. The proband's father (patient II.3) had presented with sicca syndrome aged 63 years, Raynaud phenomenon, high blood pressure and diffuse pruritus. Salivary gland, rectal and bronchial biopsies had all shown amyloid. He reached ESRD aged 64 years. The proband's younger sister (patient III.5) had died aged 49 years of staphylococcal infection, shortly after kidney transplantation. She had presented with cardiac amyloidosis proven by biopsy 10 years earlier, associated with CKD and mild proteinuria. She had reached ESRD and started haemodialysis aged 40 years. She had also suffered from sicca syndrome, Raynaud phenomenon, oesophageal reflux with oesophagitis, diffuse pruritus and severe hypertension. Her two children (patients IV.3 and IV.4) presented with sicca syndrome at age 23 and 20 years, respectively, and both of them showed amyloid deposits on salivary gland biopsies ( Supplementary  Fig. 1). Both of them subsequently developed hypertension and stage II CKD. The proband's son (IV.1) suffered from sicca syndrome aged 30 years and showed amyloid deposition on salivary gland biopsy ( Supplementary Fig. 1), but he had no hypertension and his renal function was normal with a serum creatinine level at 0.73 mg dl À 1 and no proteinuria. The proband's daughter (IV.2) was clinically asymptomatic, and had no amyloid deposits on salivary gland biopsy.
Amyloid and hypotriglyceridemia in D25V apoC-III carriers. An initial genetic screening for mutations in all genes recognized until now to be linked to hereditary amyloidosis was negative (see Methods for details). Based on the fact that affected patients with amyloidosis also displayed hypotriglyceridemia and increased plasma levels of HDL-C as compared with non-carriers in the family (Table 1), we next screened several candidate genes including the APOC3, APOC1, APOC2, APOC4, APOA4, APOA5 and APOE genes. The unique variant identified was a single-base substitution (c.134A4T) in exon 3 of APOC3, encoding replacement of the negatively charged aspartate residue at position 25 of the mature apoC-III by a valine residue (D25V; Fig. 1d). All D25V-carriers displayed a 30-50% decrease in plasma apoC-III concentration (Table 1), and there was no correlation between APOC3 promoter polymorphism genotypes and plasma TG levels among family members (Fig. 1a). Therefore, D25V apoC-III variant was uniquely found in family members affected with amyloidosis, hypotriglyceridemia and reduced apoC-III levels, and not in healthy normotriglycemic family members, demonstrating the co-inheritance of this variant with both the lipidemic and amyloidogenic phenotypic traits.
Histology and LMD/MS-proteomic analysis of amyloid deposits. To establish that apoC-III D25V variant is the causative amyloidogenic protein, we analysed amyloid deposits by immunohistochemical staining, immunogold technology and proteomic methodologies. Amyloid deposits were discovered in a total of 17 diverse disease tissues from 6 different affected family members through three generations (kidney, salivary glands, heart, bowel, fat, and bronchia from III.3; kidney, salivary glands, skin, heart, liver from III.5; salivary glands, kidney, skin, kidney, bowel from II.3; salivary glands from IV.1, IV.3 and IV.4).
All patients with sicca syndrome, the first clinical symptom, were carriers of D25V (II.3, III.3, III.5, IV.1, IV.3 and IV.4) and had amyloid deposits in salivary gland biopsies. Conversely, clinically non-affected individuals and non-carriers of D25V (III.1 and IV.2) demonstrated no amyloid deposits on salivary gland biospsies. In the proband III.3, all amyloid deposits observed in salivary gland, digestive tract and kidney were Congo red positive and immunoreactive with the anti-apoC-III antibody ( Fig. 1e-g). Moreover, extensive amyloid deposits were found around salivary acini, in the interstitial matrix and in the walls of blood vessels, and were associated with areas of fibrosis and massive acinar atrophy (Fig. 1e). Similar pathological findings in salivary glands from younger affected family D25V-carriers (individuals IV.1, IV.3 and IV.4) were found and are shown in Supplementary  Fig. 1. Therefore, this extensive amyloid accumulation associated with gland atrophy could explain the sicca symptoms of mouth reported by the patients. Renal histology from affected proband III.3 at an advanced stage of renal insufficiency showed diffuse amyloid deposition in the renal cortex that mostly affected the vascular compartment with abundant amyloid deposits in the walls of renal arterioles leading to lumen obliteration (Fig. 1g,h). Deposits were also observed in the glomeruli with prominent mesangial distribution, within the renal interstitium, and in the peritubular basement membranes with tubular atrophy. Amyloid   APOC3 mutation present; clinical syndrome; ApoC-III amyloid confirmed histologically; decreased plasma triglyceride levels; no clinical syndrome, no amyloid on histology, absence of APOC3 mutation, no hypotriglyceridemia. APOC3 haplotypes distribution of C-482T and T-455C polymorphisms is shown. D25V þ/ À , heterozygous for the apoC-III amyloidogenic variant; D25V À/ À , absence of apoC-III mutation. (b,c) Serial anterior and posterior planar whole-body 123 I-SAP scan. The scans from III.3 in b and IV.3 in c show a congested liver and amyloid in the spleen and kidneys. (d) Partial sequence chromatograms of APOC3 exon 3. (e) Amyloid deposits in salivary glands from proband III.3. (f) Amyloid deposits in the duodenum from proband III.3 predominate in the submucosa around the vessels and around Brenner's glands. There are also amyloid deposits grouped in small clods within the intestinal villi. (g) Abundant vascular, moderate glomerular and interstitial amyloid deposits in kidney from proband III.3. For each tissue in e-g, left panels show the Congo red staining, scale bar, 100 mm; middle panels show the Congo red-stained section viewed under polarized light, scale bar, 100 mm; right panels show the immunofluorescence with anti-apoC-III antibody; scale bar, 100 mm. (h) Left panel: amyloid deposits in the kidney are observed in the mesangium and the glomerular arteriole (arrow) (Jones staining, scale bar, 100 mm); middle and right panels: electron micrograph of amyloid fibrils on immunogold staining with anti-apoC-III antibody, scale bar, 500 and 200 nm, respectively). P component co-localized with the above-mentioned deposits, and vascular, glomerular, tubular and interstitial amyloid deposits were positively and specifically immunoreactive with apoC-III antibody (Fig. 1g). Electron micrographs of the kidney specimen from III.3 showed numerous non-branched and randomly arranged deposits with a fibrillar appearance, measuring 7-10 nm in diameter, typical of amyloid (Fig. 1h). Immunoelectron microscopic analysis subsequently showed that these renal fibrillar aggregates were specifically labelled by gold-conjugated anti-apoC-III antibody (Fig. 1h). Therefore, our histologic analysis, in conjunction with 123 I-SAP scintigraphy, clearly established the correlation between the clinical symptoms of D25V-carriers (sicca syndrome and renal failure) and the presence of amyloid in the respective tissues. In all 11 affected tissues from 6 D25V-carriers over three generations, the anti-apoC-III antibody consistently and specifically bound to all Congo red-staining amyloid areas, whereas there was no binding of antibodies against the other known amyloid fibril proteins. Moreover, the fact that anti-apoC-III antibody directly bound to the fibrillar aggregates confirms at the ultrastructural level the diagnosis of apoC-III amyloidosis. In addition to immunofluorescence and immunogold staining, amyloid subtyping was also confirmed by laser microdissection/ mass spectrometry (LMD/MS)-based proteomics on three different tissues: digestive tract and heart from proband III.3, and salivary glands from individual IV.3, with several amyloid deposits microdissected per tissue (Fig. 2a,b and Supplementary Table 1). In all proteomes, only the variant apoC-III containing Valine 25 was invariably found in the diverse amyloid deposits along with SAP and apoE, with no trace of the corresponding wild-type apoC-III peptide (Fig. 2c). In the amyloid proteomes of the digestive tract and salivary glands, wild-type apoA-IV and apoA-I proteins were variably present, whereas both proteins were absent in heart proteome ( Fig. 2a,b and Supplementary Table 1). These findings are in agreement with a number of studies showing the presence of apoA-I and A-IV in various amyloid deposits, irrespective of the nature of the amyloidogenic protein 15 . The concordance between the amyloid proteomes from different tissues of three D25V-carriers of this family, and the fact that no other amyloid-associated proteins were found in the heart except the mutated apoC-III clearly demonstrated that the mutated apoC-III was the amyloid fibril protein. Importantly, neither peptides representing the D25V or the wild-type apoC-III have been identified in the amyloid proteomes from over 5,000 patients with different types of systemic amyloidosis 15 , clearly indicating that apoC-III is not a protein commonly co-deposited in amyloid aggregates. Therefore, the D25V variant was specifically found in this kindred for the first time,  and histologic and proteomic analyses were consistent with the genetic findings.
MS analysis of apoC-III from D25V-HDLs. To determine the relative amounts of wild-type and mutant apoC-III as well as the proportions of apoC-III isoforms in HDL2 and HDL3, MS was carried out in these lipoproteins from hypotriglyceridemic D25V carriers III.3, IV.3 and IV.4 and normotriglyceridemic non-D25V-carrier III.4. The resulting MS profiles showed that D25V-carriers exhibited a double peak for apoC-III 1 and apoC-III 2 isoforms (only the profile of individual IV.3 is presented), whereas the non-carrier of D25V had a normal profile (Fig. 3a). The measured mass difference of 15.5 Da between the two signals is in agreement with the difference between the  wild-type aspartate residue and the mutated valine residue (15.9 Da). A higher magnification of MS profile is presented in Fig. 3b. To confirm that the additional peak observed in D25V-carriers corresponds to the apoC-III variant, LC/MS-MS sequencing was performed. As shown in Fig. 3c, both the tryptic peptides containing the mutant Val25 and the wild-type Asp25 of apoC-III isoforms were identified in heterozygous D25V-carriers, whereas individual III.4 exhibited the wild-type apoC-III tryptic peptide on HDL2/HDL3 subfractions. Relative quantification of each apoC-III isoforms was performed in the three D25V-carriers (III.3, IV.3 and IV.4) and compared with the non-carrier individual III.4. Data, summarized in Table 2, show a twofold reduction of D25V-apoC-III compared with its wild-type counterpart in all three affected patients with amyloidosis. The levels of di-sialylated apoC-III were higher in D25Vcarriers compared with the non-D25V-carrier, with a corresponding decrease in mono-and non-sialylated apoC-III.
Lipoprotein studies. The lipoprotein profiles of III.3 and a normolipidemic subject were substantially different. Interestingly, VLDL was nearly absent, LDL was reduced and HDL was massively increased in the proband's plasma (Fig. 4a). A detailed analysis of all lipoprotein constituents shows a fivefold increase in HDL2 (Table 3) but a lower apoC-III content in the proband's HDL2/HDL3 fractions ( Supplementary Fig. 2). In addition, HDL particles from III.3 were enriched in esterified and free cholesterol, and in phospholipid, but depleted in TG and protein, a composition typical of hyperalphalipoproteinemic HDL ( Table 4). The profile of several lipid classes in the proband's HDL was altered (Table 5). Although the proband had CKD, his HDL maintained an elevated cholesterol efflux capacity compared with HDL of the normolipidemic control (Fig. 4b), whereas the antioxidative activities of total HDL and small, dense HDL3c were similar between the proband and the normolipidemic control subject (Fig. 4c). The hypotriglyceridemia and The antioxidative capacity of HDL3c and total HDL from III.3 and one normolipidemic control subject to inhibit LDL oxidation was assayed at a physiological HDL to LDL ratio. HDL particles were compared on the basis of their total mass. Lipoprotein oxidation was induced by the azo-initiator 2,2 0 -azo-bis-(2-amidinopropane) hydrochloride (final concentration 1 mmol l À 1 ). Oxidation was expressed as the concentration of lipid hydroperoxides with conjugated diene structure in LDL.  hyperalphalipoproteinemia of apoC-III-D25V carriers confer an atheroprotective profile. This was corroborated by clinical data (detailed imaging of coronary angiography, pulse arterial Doppler ultrasound of the lower limbs and supra-aortic trunks, computed tomography (CT) angiography of the aorta and iliac vessels) establishing the absence of atheroma or vascular wall calcification, whereas carotid intima-media thickness was normal in the proband and his sister (III.5). Similarly, there was no record of ischaemic CVD in D25V-patients I.2 or II.1, who died aged 74 and 82 years, respectively.
Analysis of structure, function and aggregation propensity. Predictive analysis of the secondary structure of the apoC-III polypeptide in its lipid-free state revealed that the D25V mutation enhances both its beta-sheet content and aggregation propensity (Fig. 5a,b). Although wild-type and D25V apoC-III initially showed similar circular dichroic spectra, the mutation appears to cause a transition towards prominent beta-sheet content (Fig. 5c,d) associated with rapid formation of aggregates (Fig. 6a), which were shown to increase the thioflavin T fluorescence 16 , the typical signature of amyloid fibrils (Fig. 6b). Transmission electron microscopy (Fig. 6c) and atomic force microscopy ( Fig. 6d) confirmed the presence of fibrillar structures in the D25V apoC-III aggregated material.
The effect of the mutation on the lipid binding was also carried out using dimyristoylphosphatidylcholine (DMPC) multilamellar vesicles 17 . Analysis of the binding curves based on the decrease in turbidity at 325 nm showed that the time required to reduce by 50% the initial turbidity of the vesicles was 168.7±8.1 s for the variant and 114.7 ± 4.2 s for the wild-type, respectively (Fig. 7a), suggesting that the mutation may cause a less efficient lipid binding similar to other apoC-III variants 17 . At the equilibrium,     Fig. 7a). The native electrophoresis revealed the strong propensity of the variant to self-aggregate in its free form when it aggregated and precipitated at the site of deposition (inset, Fig. 7a). Solubility of D25V apoC-III was, however, rescued when the variant was integrated within the vesicles. We have also investigated the effect of the mutation on the well-known inhibition of LPL activity by apoC- III (refs 3,18). The D25V apoC-III variant exhibited the same inhibitory activity when compared with the wild-type (Fig. 7b). It is worth noting that another mutation in position 25 (D25N, designed in silico), with the change in charge but unmodified hydrophobicity, was as efficient as the wild-type in inhibiting LPL activity 19 .
We finally compared the 3D NMR structure of D25V apoC-III with that of the wild-type counterpart already characterized in the presence of SDS 9 . Two-dimensional 1 H, 15 N NMR spectra (Fig. 8a) showed small (o0.2 p.p.m., excluding residue 25) amide chemical shift changes in the variant (Fig. 8b,c), which were clustered within approximately one helical turn (3.6 residues) of the mutation site. No chemical shift perturbations were observed elsewhere in the sequence. The hydrodynamic radii of wild-type and D25V apoC-III micellar complexes, recorded by TRACT relaxation interference experiments, were virtually indistinguishable (Fig. 8d). Moreover, there was no difference in the helical conformation of residues 20-30 (Fig. 8e).

Discussion
Here we report a previously unappreciated, naturally occurring APOC3 mutation 6-8 , which causes low levels of plasma TG and apoC-III and a favourable lipoprotein profile in a French kindred affected with amyloidosis. Concordance between the APOC3 mutation and the presence of amyloid among members of this kindred, coupled with identification of D25V apoC-III within amyloid fibrils from numerous affected tissues and in vitro demonstration of its amyloidogenicity, establishes that the D25V variant also causes amyloidosis.
The phenotype of this new form of amyloidosis is mainly characterized by onset with sicca syndrome as early as age 20 years, secondary to diffuse salivary gland deposits and progressive renal insufficiency. We showed that these clinical symptoms are caused by amyloid deposition in the respective tissues with progressive and widespread amyloid accumulation within different organs. In this novel form of renal amyloidosis, amyloid deposition is characterized by prominent vascular involvement in all compartments of the kidney and ischaemic glomerular lesions that probably account for low-grade proteinuria and severe hypertension. This clinical renal presentation contrasts those described for Leu75Pro apoA1 amyloidosis 20 and A-alpha chain amyloidosis 21 , which typically manifests as a chronic tubulointerstitial nephritis with amyloid deposits restricted to the inner medulla, or as proteinuria/nephrotic syndrome because of massive glomerular amyloid deposition, respectively, ApoC-III  Our in vitro studies clearly show that, despite the remarkable differences in folding dynamics and kinetics of aggregation between wild-type and D25V apoC-III in their lipid-free state, their conformation and colloidal stability are similar in the lipidbound state. Wild-type apoC-III was recently reported to selfaggregate into polymeric ordered structures with a peculiar triangular geometry and a Möbius strip conformation after 3 days of shaking in physiological buffer. Although these structures do not resemble the genuine amyloid fibrils that we obtained under physiological conditions with the D25V variant, it is worth noting that aggregation of lipid-free wild-type apoC-III was preceded by a gradual structural transition that resembles, in circular dichroism (CD) spectroscopy, the conformation rapidly adopted by the D25V variant. There is no evidence suggesting in vivo amyloidogenicity of wild-type apoC-III and demonstration of aggregation in vitro does not necessarily imply amyloidogenicity in vivo. Importantly, the absence of wild-type apoC-III within the ex vivo amyloid fibrils from three patients in this kindred suggests that the wild-type protein does not contribute to amyloid deposition even in the presence of abundant D25V fibrillar seeds, analogous to what we have recently observed in patients with familial b 2 -microglobulin amyloidosis 14 .
We hypothesize that the preferential incorporation of D25V apoC-III into amyloid fibrils is likely the main cause of the imbalance in the ratio of wild-type/D25V apoC-III observed within HDL particles. It will be worth investigating further in appropriate cellular and/or animal models whether the secretion of the variant may be partially impaired by the intracellular quality control, similar to what we previously described for some of the amyloidogenic variants of apoA-I (ref. 25), in which we found an imbalance in the wild type/variant ratio.
D25V-carriers displayed a reduction in plasma and HDL apoCIII concentration, hypotriglyceridemia associated with a dramatic decrease in the number of VLDL particles, and a concomitant massive increase in the larger HDL2 fraction. We propose that these are indirect effects of the D25V variant because of its very low concentration in plasma (and HDL fractions) because of its preferential association with amyloid fibrils. The lower plasma apoC-III concentration in D25Vcarriers probably results in lack of inhibition of LPL and LDL-R. As a consequence, VLDL-TG hydrolysis proceeds at high rate and is followed by rapid uptake of VLDL-remnants (or intermediate density lipoprotein (IDL) in Fig. 4b), provoking the drastic decreases in VLDL and IDL of D25V-carriers (Fig. 4a,b). It has been well established that surface lipids of VLDL, which are liberated after LPL-mediated VLDL-TG hydrolysis, are transferred to HDL by phospholipid transfer protein, resulting in an increase in HDL size and particle number, as confirmed in mice deficient either in LPL 26 or in phospholipid transfer protein 27 and which lack HDL. Importantly, subjects heterozygous for other loss-of-function APOC3 mutations (K58E 28 and R19X 5,6 ) also present with hypotriglyceridemia and increased plasma HDL.
The favourable lipid/lipoprotein profile of D25V-carriers is remarkable in the context of CKD and/or ESRD, pathological conditions predisposing to premature atherosclerosis and major ischaemic CVD events and commonly associated with an unfavourable lipid profile, an elevated plasma apoC-III concentration and dysfunctional HDL carrying increased amounts of apoCIII, serum amyloid A1 and lipoprotein-associated phospholipase A2 (refs 29-31). Hypertriglyceridemia, often due to accumulation of VLDL remnants/IDL in the circulation, is an independent CVD risk factor 32 , and low plasma HDL is a second independent CVD risk factor 33 . D25V-carriers displaying hypotriglyceridemia and very high HDL2 are obviously lacking in CVD risk factors. Moreover, their HDL is functional, as evidenced by their normal antioxidant activity and cholesterol efflux capacity possibly related to their enrichment in sphingomyelin (SM). Furthermore, the lower apoC-III content of HDL may contribute to atheroprotection as HDL particles without apoC-III reduce monocyte adhesion to vascular endothelial cells, whereas apoC-III-rich HDL do not, suggesting that apoC-III may have additional proatherogenic effects on vascular endothelial cells 34 . Thus, the pathologic amyloidogenic properties of D25V apoC-III unexpectedly confer a favourable cardioprotective lipoprotein profile indirectly, most probably through lowering of plasma apoC-III concentration, extending the evidence for an important antiatherogenic role of APOC3 deficiency in ESRD patients.
As reducing the amyloid fibril precursor protein concentration is known to slow amyloid formation and improve prognosis among patients with a variety of systemic amyloidoses, and as PPARa agonists (fibrates) significantly reduce hepatic APOC3 transcription 35,36 , we postulate that fibrate therapy may have therapeutic potential for treatment of this new form of amyloidosis. Because the safety and side-effect profiles of fibrates are well known, it might be reasonable to prescribe fibrates to the affected individuals from this kindred in an attempt to slow down amyloid formation. If successful, this therapy could represent an alternative to the gene silencing approach currently under evaluation in other systemic amyloidoses 37 .   Genetic analysis. Genomic DNAs were extracted from peripheral blood samples, using the QIAamp DNA mini kit (Qiagen). Coding and consensus splice site  regions of APOA1, APOA2, TTR, B2M, FGA, LYZ, GSN, SAA, APOE, APOC3,  APOC2, APOC1, APOC4, APOA4 and APOA5 genes were amplified by PCR (AmpliTaq Gold; ABI, or Platinium Taq; Invitrogen), according to the manufacturer's instructions, in a DNA thermal cycler 9700 (Perkin-Elmer). The primer sequences for APOC3 amplification are listed in Supplementary Table 2.
The resulting PCR fragments were purified by QIAquick PCR Purification kit (Qiagen), and sequenced using the Big Dye terminator cycle sequencing kit (DNA sequencing kit; Applied Biosystems) on an automatic genetic analyser (ABI PRISM 3100 genetic analyzer; Applied Biosystems) 14 . APOC3 polymorphims at positions À 455 (rs 2854116), and À 482 (rs2854117) in the promoter region were genotyped by direct sequencing. Mutation nomenclature was based on the APOC3 transcript reference (NCBI RefSeqcDNA accession number NM_000040.1), and nucleotides were numbered according to the cDNA with þ 1 corresponding to the A of ATG translation initiation codon according to the Human Genome Variation Society guidelines, http://www.hgvs.org/mutnom//. Therefore, the apoC-III variant is described as p.Asp45Val (with the signal peptide), whereas the mature protein is designed as D25V.
Histology and immunohistochemistry. Amyloid was detected by Congo red staining of formalin-fixed wax-embedded biopsy sections 14 (kidney, salivary glands, heart, bowel, fat and bronchia from III.3; kidney, salivary glands, skin, heart, liver from III.5; salivary glands, kidney, skin, kidney, bowel from II.3; salivary glands from IV.1, IV.3, and IV.4). Amyloid typing was performed by indirect immunofluorescence and by immunoelectron microscopy 14 . For indirect immunofluorescence, 3 mm cryostat sections of frozen tissue were air dried, incubated 1 h with primary goat anti-human ApoC3 polyclonal antibody (Abnova, 50 mg ml À 1 in PBS), washed with PBS and stained with fluorescent isothiocyanateconjugated donkey anti-goat antibody IgG, F(ab,)2 (Santa Cruz, 20 mg ml À 1 in PBS). For the immunogold labelling procedure, thin sections (80 nm) were cut from the routine (glutaraldehyde-fixed, osmicated, Araldite embedded) electron microscopic blocks, collected on nickel grids. Grids were floated on one drop of 5% sodium metaperiodate (Sigma Chemical Co) in distilled water for 30 min. Sections were washed with distilled water three times and then saturated for 1 h with 3% BSA in PBS in order to minimize nonspecific labelling. The grids were then incubated overnight at 4°C with primary goat anti-human ApoC3 polyclonal antibody (Abnova, 100 mg ml À 1 in PBS). Sections were washed six times with PBS, incubated for 1 h with a rabbit polyclonal biotin-conjugated anti-goat IgG secondary antibody (Clinisciences, 100 mg ml À 1 in PBS) and 1 h with gold-labelled strepatavidin (Zymed Laboratories, 6 mg ml À 1 in PBS). Immunolabelled sections were contrasted using uranyl acetate and observed under a JEOL 1010 transmission electron microscope (Jeol Ltd). In all cases, parallel control negative sections consisting of the absence of the primary apoC-III antibody were performed, demonstrating the specificity of the apoC-III antibody.
LMD and tandem MS-based proteomics. Congo red-stained sections from salivary glands (IV.3), heart and digestive tract (III.3) were examined for the presence of amyloid under fluorescence (Leica, B/G/R filter cube). Positive areas were microdissected into 0.5 ml microcentrifuge tube caps containing 10 mM Tris, 1 mM EDTA, 0.002% Zwittergent 3-16 (Calbiochem) using a Leica DM6000B Microdissection System (Leica). For each case, -two to four different microdissections (60,000 mm 2 each) were collected. Collected tissues were heated at 98°C for 90 min with occasional vortexing. Following 60 min of sonication in a waterbath, samples were digested overnight at 37°C with 1.5 ml of 1 mg ml À 1 trypsin (Promega) 15 . The trypsin generated digests were reduced with dithiothreitol and separated by nanoflow liquid chromatography-electrospray tandem mass spectrometry using a ThermoFinnigan LTQ Orbitrap Hybrid Mass Spectrometer (Thermo Electron) coupled to an Eksigent nanoLC-2D HPLC system (Eksigent). A 0.25-m1 trap (Optimize Technologies) packed with Michrom Magic C-8 was plumbed into a 10-port valve. A 75 mm Â B15 cm C-18 column was utilized for the separation utilizing an organic gradient from 6 to 86% in 55 min at 400 nl min À 1 . The Thermo-Fisher MS/MS raw data files were searched using three different algorithms (Mascot, Sequest and X!Tandem) and the results assigned peptide and protein probability scores. The results were then combined and displayed using Scaffold (Proteome Software). All searches were conducted with variable modifications and restricted to full trypsin-generated peptides allowing for two missed cleavages. Peptide mass search tolerances were set to 10 p.p.m. and fragment mass tolerance to 1.00 Da. The human SwissProt database (24 July 2007) was complemented with frequently observed contaminants (porcine trypsin and human keratins). Protein identifications below the 90% confidence level and those with single peptide identification were not considered Lipid and lipoprotein analysis. Plasma levels of TGs and HDL-cholesterol were measured in the non-fasting state using standard hospital biochemical assays. Plasma apoC-III concentrations were measured by electro-immunodiffusion. Lipoprotein classes were prepared by sequential ultracentrifugation of sera from the proband III-3 and a control subject run in parallel. Sequential ultracentrifugations were performed at 100,000g and 10°C to isolate VLDL, IDL and LDL (for 18 h, at densities 1.006, 1.020 and 1.063 g ml À 1 , respectively) and HDL (for 40 h, at density 1.21 g ml À 1 ) in an OPTIMA XL 100 K Beckman ultracentrifuge 38 . After extensive dialysis of all lipoproteins against PBS containing/0.3 mM EDTA, their protein contents were measured with the Bio-Rad DC Protein assay kit II (Bio-Rad), and their TG, total cholesterol and phospholipid with commercial kits (DiaSys). HDL apolipoproteins were analysed by 15% SDS-PAGE under non-reducing conditions to distinguish the apoA-II homodimer (17 kDa) from apo Cs 38 , and stained with Coomassie Brilliant Blue G 250. Purified human apoA-I and apoA-II, as major HDL apolipoproteins, were run as standards (4 mg per lane) in the gel. For apoC-III detection, proteins were transferred to a nitrocellulose membrane and immunoblotted with rabbit anti-human apoC-III antibody (1 mg ml À 1 in PBS, Abnova) and secondary ECL anti-rabbit IgG-HRP-Linked specific whole antibody (5,000-fold diluted in PBS, GE Healthcare). Purified human apoC-III (4 mg per lane) was added as standard in the western blotting. Protein bands were visualized with ECL Western Blotting Detection Reagent (GE Healthcare). Standard human apoA-I, apoA-II and apoC-III were purified from delipidated HDL by preparative isoelectric focusing with a pH gradient 4-6.
Fractionation of HDL subpopulations. HDL subpopulations were isolated from plasma (3 ml) by single step, isopycnic non-denaturing density gradient ultracentrifugation in a Beckman SW41 Ti rotor at 40,000 r.p.m. for 44 h in a Beckman XL70 ultracentrifuge at 15°C by a slight modification of the method of Chapman et al. 39,40 . After centrifugation, each gradient was fractionated in predefined volumes from the meniscus downwards with an Eppendorf precision pipette into 11 fractions corresponding to very-low-density lipoprotein and intermediate-density lipoprotein (VLDL þ IDL; 800 ml, do1.019 g ml À 1 ), LDL (5 subfractions, 800 ml each, LDL1, d ¼ 1.019-1.023 g ml À 1 ; LDL2, d ¼ 1.023-1.029 g ml À 1 ; LDL3, d ¼ 1.029-1.039 g ml À 1 ; LDL4, d ¼ 1.039-1.050 g ml À 1 and LDL5, d ¼ 1.050-1.063 g ml À 1 ) and HDL. Five major HDL subclasses were isolated, that is, large, light HDL2b (1,200 ml, d ¼ 1.063-1.087 g ml À 1 ) and HDL2a (800 ml, d ¼ 1.088-1.110 g ml À 1 ), and small, dense HDL3a (d ¼ 1.110-1.129 g ml À 1 ), HDL3b (800 ml, d ¼ 1.129-1.154 g ml À 1 ) and HDL3c (800 ml, d ¼ 1.154-1.170 g ml À 1 ). The validity and reproducibility of this density gradient procedure, which facilitates preparative fractionation of HDL particle subspecies in a non-denaturated, native state, have been extensively documented [39][40][41] . All HDL subclasses isolated by this procedure are essentially albumin free (o1% of total protein, that is, o0.05 mg dl À 1 ) 42 43 . A blank and a control sample were extracted in parallel with each batch to ensure for quality control. K 4 EDTA (200 mM) solution was added (1:5 (v/v)) and the mixture was vortexed and centrifuged. The organic phase was dried under nitrogen and lipids were reconstituted into isopropanol/hexane/water (10:5:2 (v/v)), transferred into LC/MS vials, dried under nitrogen and resuspended in isopropanol/hexane/water (10:5:2 (v/v)). Seven principal phospholipid (PL) subclasses (phosphatidylcholine, lysophosphatidylcholine, phosphatidylethanolamine, PI, phosphatidylglycerol, phosphatidylserine and phosphatidic acid) and two principal sphingolipid (SL) subclasses, SM and ceramide, which together comprise 4160 individual molecular lipid species and account for 495% of total plasma PL and SM 44 were assayed by LC/MS/MS. Lipids were quantified using a QTrap 4000 mass spectrometer (AB Sciex), an LC20AD HPLC system, and the Analyst 1.5 data acquisition system (AB Sciex). Quantification of PLs and SLs was performed in positive-ion mode, except for PI species that were detected in negative-ion mode. Sample (4 ml) was injected onto a Symmetry Shield RP8 3.5 mm, 2.1 Â 50 mm 2 reverse phase column (Waters Corporation) using a gradient from 85:15 to 91:9 (v/v) methanol/water containing 5 mM ammonium formate and 0.1% formic acid at a flow rate of 0.1 ml min À 1 for 30 min. Lipid species were detected using multiple reaction monitoring reflecting the head-group fragmentation of each lipid class and quantified using calibration curves specific for the nine individual lipid classes with up to 12 component fatty acid moieties; 23 calibration curves were generated in non-diluted and tenfold diluted matrices to correct for matrix-induced ion suppression effects. More abundant lipid species that displayed a nonlinear response in non-diluted extracts were quantified from a 10-or 100-fold diluted sample. An in-house developed script was used to compile data from the three successive injections.
Cellular cholesterol efflux capacity of HDL. The cholesterol efflux capacity of HDL was characterized in a human THP-1 monocytic cell system (ATCC) 45 . In this assay, HDL particles were compared on the basis of their PL concentrations NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10353 ARTICLE because PL was shown to represent the key component determining cholesterol efflux capacity of HDL 46 . Briefly, THP-1 monocytes were cultured on 24-well tissue culture plates, grown in RPMI-1640 media with 10% FBS and differentiated into macrophage-like cells with 50 ng ml À 1 phorbol 12-myristate 13-acetate for 48 h and 37°C. The cells were washed and loaded for 24 h with [3H]cholesterol-labelled acetylated LDL (1 mCi ml À 1 ) in serum-free RPMI-1640 culture medium supplemented with 50 mM glucose, 2 mM glutamine, 0.2% BSA, 100 mg ml À 1 penicillin and 100 mg ml À 1 streptomycin (further abbreviated as RGGB) to allow equilibration of cellular cholesterol pools. The labelling medium was removed and human macrophages were then equilibrated in RGGB for an additional 16-24 h period. Cellular cholesterol efflux to 15 mg ml À 1 HDL-PL was assayed in 300 ml serum-free medium for a 4-h chase period. Finally, culture media were harvested and cleared of cellular debris by brief centrifugation. Cell radioactivity was determined by extraction in hexane-isopropanol (3:2), evaporation of the solvent under nitrogen and liquid scintillation counting (Wallac Trilux 1450 Microbeta, Perkin Elmer). The percentage of cholesterol efflux was calculated as (medium cpm)/ (medium cpm þ cell cpm) Â 100%. Specific cholesterol efflux was determined by subtracting nonspecific cholesterol efflux occurring in the absence of cholesterol acceptors.
Antioxidative activity of HDL. Antioxidative activity of HDL (final concentration of each, 10 mg total mass per dl) was assessed at a physiological HDL to LDL ratio towards reference LDL (d ¼ 1.019-1.063 g ml À 1 ; final concentration, 10 mg dl À 1 TC) isolated from one healthy normolipidemic control subject 42,47,48 . In this assay, HDL particles were compared on the basis of their total mass concentrations because both protein and lipid components were shown to contribute to the capacity of HDL to inhibit LDL oxidation 43,49,50 . HDL subfractions were added to LDL directly before oxidation. Lipoprotein oxidation was induced by an azo-initiator 2,2 0 -azo-bis-(2-amidinopropane) hydrochloride (final concentration 1 mmol l À 1 ) as model of mild oxidation induced by free radicals in the arterial intima 42,47,48 . Serum was used as a source of HDL for this assay to ensure intact paraoxonase activity, which is inhibited by EDTA. Thereby, this assay employs mild oxidative conditions and integrates the antioxidative activities of several HDL components, that is, apoA-I, antioxidative enzymes and lipophilic low-molecularweight antioxidants 43,49,50 . Accumulation of conjugated dienes was measured as the increment in absorbance at 234 nm. Absorbance kinetics values were corrected for the absorbance of 2,2 0 -azo-bis-(2-amidinopropane) hydrochloride itself run in parallel as a blank. Three consecutive phases were identified, the lag, the propagation and the decomposition phases for each curve and duration of the propagation rate, oxidation rate in the propagation phase and amount of dienes formed at the end of the propagation phase (maximal concentration of dienes) were calculated as markers of antioxidative activity of HDL 42,47,48 .
MS analysis of apoC-III isoforms from HDL. Twenty micrograms of HDL fraction was dissolved and acidified in 20 ml of 0.1% trifluoroacetic acid for 15 min with agitation. The protein was desalted using ZipTips C18 (Millipore) following the manufacturer's instructions and eluted with 20 ml of saturated matrix solution of sinapinic acid in 50% acetonitrile and 0.5% (v/v) trifluoroacetic acid. The mass spectrometric measurements were performed using a MALDI TOF/TOF 4800 Proteomics Analyzer mass spectrometer 51 . The MS spectra from m/z 8,000 to 10,000 Da were acquired both in reflectron and in linear positive ion modes using 1,000 laser shots. For mass spectrometric identification of ApoC-III in HDL2/3 using LC-MS/MS, 25 mg of HDL fraction was dissolved into 500 mM triethylammonium bicarbonate, 1% SDS. Proteins were reduced with 10 mM dithiothreitol, alkylated with 20 mM iodoacetamide and trypsin digested overnight. The resulting peptides were cleaned up using OMIX C18 100 ml pipette tips (Agilent), and lyophilized before being reconstituted for the LC-MS/MS analysis. The peptides were separated using an Eksigent Ultra nano-LC HPLC system coupled with an AB Sciex Triple TOF 5600 mass spectrometer. The LC separations were performed using a Discovery Bio Wide Pore HPLC column (C18, 3 mm, 100 Â 5 mm 2 ). The mobile phases used were 0.1% formic acid in water (A) and 100% acetonitrile with 0.1% formic acid (B). The gradient elution steps were performed with a flow rate of 5 ml min À 1 as follows: 0-40% B for 106 min, 40-80% for 5 min and then 80% B for an additional 5 min. All data were acquired using Analyst software (AB Sciex) in the data-dependent mode. Peptide profiling was performed using a mass range of 350-1,600 Da, followed by a MS/MS product ion scan from 100 to 1,500 Da. For each survey MS1 scan (accumulation time of 250 ms), MS/MS spectra (accumulation time of 75 ms per MS/MS) were obtained for the 30 most abundant precursor ions. The protein identification was performed with the ProteinPilot Software 4.5.0 (AB Sciex). The MS/MS spectra obtained were searched against the UniProt database (release 2014_03). The search parameters for tryptic cleavage and accuracy are built-in functions of the software. Other data analysis parameters were as follows: sample type: identification; Cys-alkylation: iodoacetamide; digestion: trypsin; instrument: TripleTOF 5600; special factor: biological modifications; species: Homo sapiens; search effort: thorough ID. Proteins comprising one or more peptides with a high confidence score (95%) and a low false discovery rate (estimated local false discovery rate of 1%) were considered positively identified.
Radiolabelled SAP scintigraphy. Patients III.3 and IV.3 underwent 123 I-labelled SAP scintigraphy 52 (http://www.amyloidosis.org.uk/diagnosis/sap-scintigraphy-3/). Purified human SAP was labelled with 123 I using the N-bromosuccinimide method. Each patient received B100 mg SAP bearing 200 MBq of 123 I, corresponding to an effective dose equivalent of B3 mSv. Thyroid uptake was blocked by oral ingestion of potassium iodide for 2 days before and following isotope injection. Anterior and posterior whole-body scintigraphy was performed 24 h after isotope administration using a GE Healthcare-Discovery gamma-camera.
Expression and purification of recombinant ApoC-III. ApoC-III was expressed from a pET23b vector containing the full-length cDNA for human ApoC-III, including the sequence encoding a C-terminal His6-tag preceded by Leu and Glu additional residues 17 . pET23b/D25V was generated by using a QuikChange site-directed mutagenesis kit (Stratagene). In a 50 ml reaction, 50 ng of plasmid DNA pET23b/ApoC-III was used as template and two synthetic oligonucleotide primers, the forward primer (5 0 -CCAAGACCGCCAAGGtTGCACTGAGCA GCG-3 0 ) and the reverse primer (5 0 -CGCTGCTCAGTGCAaCCTTGGCGGTCT TGG-3 0 ) with the A-to-T base change complementary to opposite strands of the vector, were extended by Pfu Turbo DNA polymerase (Stratagene). For PCR, there was an initial cycle of 95°C for 30 s, followed by 12 cycles of 95°C for 30 s, 55°C for 1 min and 68°C for 2 min. After DpnI digestion of the parental deoxyadenosine methylase-methylated template, the synthesized mutated DNA was transformed into E. coli XL1Blue supercompetent cells. Candidate clones were screened by sequencing the ApoC3 insert on an ABI 377 prism DNA sequencer.
Prediction of beta aggregation propensity. The Garnier-Osguthorpe-Robson (GOR) method version IV (ref. 53) was used to predict the beta propensity at the mutation site. The aggregation propensities of wild-type and D25V ApoC-III were predicted using the Zyggregator method 54 .
Fibrillogenesis. Fibrillogenesis experiments were performed in standard quartz cells stirred at 1,500 r.p.m. (IKA magnetic stirrer) at 37°C using 100 mM ApoC-III isoforms in PBS, pH 7.4. Aggregation was carried out without seeds of preformed fibrils and the increase in turbidity at 350 nm was monitored. Thioflavin T fluorescence emission 16 was measured at the end-point aggregation. Transmission electron microscopy analysis was performed using protein samples stained with 2% (w/v) uranyl acetate using a CM120 microscope at 80 keV. Atomic force microscopy analysis was carried out on 10 ml of fibrillar sample incubated on a freshly cleaved mica substrate for 5 min, then rinsed and dried. Conformational modifications during aggregation of apoC-III were monitored by far-UV CD.
Lipid-binding properties of apoC-III. Association of recombinant wild-type and D25V apoC-III to DMPC multilamellar vesicles 55 , at a lipid-to-protein ratio of 2:1 (w/w), was followed by monitoring the decrease in turbidity of DMPC liposomes at 325 nm as a function of time at 30°C (ref. 17) in 0.01 M Tris-HCl buffer pH 8.0 containing 150 mM NaCl, 8.5% KBr, 0.01% NaN 3 and 0.01% EDTA using a 1 cm path length quartz cell in a Jasco V-650 spectrophotometer equipped with a Peltier temperature controller. The lipid/protein association was expressed as percentage of the turbidity of DMPC vesicles in the absence of apoC-III; the time required to reduce the initial absorbance at 325 nm by 50% was determined from the curves to compare the lipid-binding properties of the two apoC-III isoforms. Electrophoretic mobility of wild-type and D25V apoC-III with and without DMPC was analysed using 1% medium electroendosmosis agarose gel in 70 mM barbiturate-sodium barbiturate buffer, pH 8.6 containing 2 mM calcium lactate. Samples (3 ml) were applied to the gel including fivefold diluted human plasma in 2% bromophenol blue solution used as marker for protein electrophoretic mobility. Electrophoresis was carried out in a 10°C water-thermostated chamber at 20 V cm À 1 . Gel was then fixed for 15 min in a 15% acetic acid solution containing 1% picric acid, dried and stained with Coomassie Blue.
LPL inhibition. Bovine LPL catalytic activity, in the presence of apoC-III was measured using an emulsion with the same composition as Intralipid 10% (Fresenius-Kabi, Sweden) containing [3H]triolein substrate 56 . The incubation medium contained 2% (v/v) of the emulsion, 0.1 M NaCl, 60 mg ml À 1 BSA, 0.15 M Tris-HCl, pH 8.5, and 16.7 U ml À 1 heparin. The medium was pre-incubated with or without each apoC-III sample for at least 10 min before addition of bovine LPL (0.25 mg ml À 1 ) in a final 200 ml mixture. After 30 min incubation under agitation at 25°C, the reaction was stopped by adding 2 ml of isopropanol, heptane and 1 M H 2 SO 4 (40:48:3:1) and 0.5 ml of water. Free fatty acids were extracted using sequential centrifugations in glass tubes for 3 min at 2,580g to separate each mixture in two phases. After the first centrifugation, 800 ml of the upper phase was transferred to other vials containing alkaline ethanol and then added with 3 ml of heptane before mixing and centrifuging again. After removal of the upper phase, 3 ml of heptane were mixed to the lower phase, which was then centrifuged as above. Supernatant was removed and 800 ml of the lower phase transferred to a vial containing 2 ml of scintillation liquid to count radioactivity and quantify the [3H] free fatty acids (nmol ml min À 1 ). The LPL activity was then expressed as percentage of the enzyme activity in the absence of apoC-III.
NMR spectroscopy. NMR spectra were acquired for 0.5 mM samples of wild-type and D25V apoC-III (10 mM sodium acetate (pH 5.0), 180 mM SDS, 10% D 2 O) at 315 K, using a 700-MHz BrukerAvance III spectrometer equipped with a TXI cryoprobe, and processed with nmrPipe and CCPN Analysis.