Serum amyloid A forms stable oligomers that disruptvesicles at lysosomal pH and contribute to thepathogenesis of reactive amyloidosisShobini Jayaramana,1, Donald L. Gantza, Christian Hauptb, and Olga Gurskya

aDepartment of Physiology & Biophysics, Boston University School of Medicine, Boston, MA 02118; and bInstitute of Protein Biochemistry, University of Ulm,89081 Ulm, Germany

Edited by Susan Marqusee, University of California, Berkeley, CA, and approved June 29, 2017 (received for review April 28, 2017)

Serum amyloid A (SAA) is an acute-phase plasma protein that functionsin innate immunity and lipid homeostasis. SAA is a protein precursor ofreactive AA amyloidosis, the major complication of chronic inflamma-tion and one of the most common human systemic amyloid diseasesworldwide. Most circulating SAA is protected from proteolysis andmisfolding by binding to plasma high-density lipoproteins. However,unbound soluble SAA is intrinsically disordered and is either rapidlydegraded or forms amyloid in a lysosome-initiated process. Althoughacidic pH promotes amyloid fibril formation by this and many otherproteins, the molecular underpinnings are unclear. We used an array ofspectroscopic, biochemical, and structural methods to uncover that atpH 3.5–4.5, murine SAA1 forms stable soluble oligomers that are max-imally folded at pH 4.3 with ∼35% α-helix and are unusually resistantto proteolysis. In solution, these oligomers neither readily convert intomature fibrils nor bind lipid surfaces via their amphipathic α-helices in amanner typical of apolipoproteins. Rather, these oligomers undergo anα-helix to β-sheet conversion catalyzed by lipid vesicles and disruptthese vesicles, suggesting a membranolytic potential. Our results pro-vide an explanation for the lysosomal origin of AA amyloidosis. Theysuggest that high structural stability and resistance to proteolysis ofSAA oligomers at pH 3.5–4.5 help them escape lysosomal degradation,promote SAA accumulation in lysosomes, and ultimately damage cel-lular membranes and liberate intracellular amyloid. We posit that thesesoluble prefibrillar oligomers provide a missing link in our understand-ing of the development of AA amyloidosis.

prefibrillar amyloid oligomers | intrinsically disordered proteins |pH-induced conformational changes | hydrophobic cavities |acute-phase reactant

Serum amyloid A (SAA, 12 kDa) is an intrinsically disorderedplasma protein that functions in innate immunity and lipid

homeostasis and is up-regulated in inflammation (1–3). Mostcirculating SAA binds to its major plasma carrier, high-densitylipoprotein (HDL), and thereby reroutes HDL transport andpromotes lipid clearance from the sites of injury and lipid recyclingfor tissue repair (2, 3). Binding to HDL stabilizes the α-helicalstructure in SAA and protects it from proteolysis and misfolding.Such a misfolding underlies amyloid A (AA) amyloidosis, a majorcomplication of chronic inflammation and a common form ofhuman systemic amyloidosis (4, 5). In this life-threatening disease,N-terminal fragments of SAA, termed AA, deposit as fibrils inkidneys, spleen, and other organs, causing organ damage (4–6).Although binding to the lipid surface mediates normal functionsof SAA in lipid transport, SAA interactions with cell membranesmay also modulate pathologic effects in AA amyloidosis (7, 8).Molecular underpinnings for these effects are beginning to beelucidated (9) and are explored in the current study.Aberrant interactions of amyloidogenic proteins with cell mem-

branes have been implicated in the pathogenesis of other amyloiddiseases such as Alzheimer’s, Parkinson’s, type 2 diabetes, etc. (seerefs. 10–12 and references therein). Furthermore, extensive evi-dence suggests that the primary cytotoxic species in various forms ofsystemic and neurodegenerative amyloidoses are compact soluble

prefibrillar oligomers with diverse structural and pathogenic fea-tures (see refs. 13–15 and references therein). Such polymorphicoligomers can exert toxicity through multiple mechanisms in-cluding perforation of cell membranes (7, 13); in addition, maturefibrils can mediate a range of pathological processes (16, 17).Certain “on-path” oligomers can also trigger fibril formation viathe crystallization-like nucleation-growth mechanism (16, 18) in aprocess that can be affected by cell membranes.Unlike neurodegenerative disorders such as Alzheimer’s, in

AA amyloidosis, fibril removal restores organ function, whichsuggests that AA fibrils per se are pathogenic and sets the basisfor the current treatment of AA amyloidosis (4, 5, 19, 20).Nevertheless, small soluble SAA-based oligomers can also con-tribute to the AA pathogenesis (21–24). Clinically, once AA depositshave been cleared by administering anti-inflammatory therapies todecrease the plasma levels of SAA, new fibrils form again muchfaster on repeated exposure to inflammatory stimuli that up-regulateSAA (18–20). Such a recurring amyloidosis is probably triggeredby particularly stable SAA-derived species that cannot be com-pletely eliminated by therapeutic interventions. These speciesare perhaps akin to the amyloid-enhancing factor (AEF) derivedfrom amyloid deposits in vivo, which can nucleate fibril forma-tion in vitro (21, 22). Oral administration of AEF can induce AAamyloidosis in mice, indicating prion-like transmission (18, 25). Ahallmark of such a transmission is the high stability of prions, whichconfers resistance to denaturation and proteolytic degradation inthe digestive tract. Therefore, high stability of SAA-based AEF isprobably important for the onset of AA amyloidosis in humans and

Significance

Although acidic conditions favor themisfolding of various proteinsin amyloid diseases, the molecular underpinnings of this processare unclear. We used an array of spectroscopic, biochemical, andelectron microscopic methods to unravel the effects of pH on se-rum amyloid A, a protein precursor of reactive amyloidosis, themajor complication of chronic inflammation and one of the majorhuman systemic amyloidoses worldwide. We found that at lyso-somal pH this protein forms unusually stable proteolysis-resistantsoluble oligomers that have solvent-accessible apolar surfaces,disrupt lipid vesicles, and undergo an α-helix to β-sheet transitionin the presence of lipids. Such oligomers are likely to escape ly-sosomal degradation, accumulate in the lysosomes, and disruptcellular membranes, thereby contributing to the development ofamyloid A amyloidosis.

Author contributions: S.J. designed research; S.J. and D.L.G. performed research; C.H.contributed new reagents/analytic tools; S.J. and O.G. analyzed data; and S.J. and O.G.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: shobini@bu.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1707120114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1707120114 PNAS | Published online July 25, 2017 | E6507–E6515

BIOCH

EMISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0

its transmission in mice. The molecular entity responsible for theAEF activity in AA amyloidosis is unclear and was proposed tocomprise degradation-resistant species, including prefibrillar dif-fusible oligomers or small fibril fragments (23, 25). Here we de-scribe soluble SAA oligomers formed at lysosomal pH in vitro,which have unusually high stability, disrupt lipid vesicles, andhence may contribute to the SAA accumulation in the cells, cellmembrane damage, and AA pathogenesis.Acidic pH, particularly in the lysosomal or endosomal com-

partment (pH 4.5–5.5), is well-known to favor fibril formation invivo and in vitro by many protein precursors of major humanamyloidoses, including SAA, transthyretin, Ig light chain, Aβ pep-tide, etc. (see ref. 26 and references therein). In vitro fibril forma-tion by SAA and AA requires acidic conditions (27) that areoptimal circa pH 3 (9). Moreover, ex vivo and cell-based immu-nohistochemical and electron microscopic (EM) studies have clearlyestablished lysosomes as the initial sites of AA fibril formation inmacrophages and other cells (26, 28, 29). Besides acidic pH, proteinconcentration in lysosomes and the activity of acidic proteasesprobably contribute to the lysosomal origin of fibrillogenesis in thisand other amyloid diseases. To elucidate molecular underpinningsfor the lysosomal origin of AA fibrillogenesis, here we address therole of acidic pH in SAA folding, aggregation, fibril formation,proteolytic degradation, and interactions with lipid bilayers in vitro.

ResultsEffects of pH on the Secondary Structure and Thermal Stability of SAAin Solution.Recombinant murine SAA isoform 1.1 (mSAA1, 103 aa,11.6 kDa) used in this study is a major amyloidogenic isoform thatbinds HDL and has 80% sequence similarity to human SAA1. Fig.1A shows far-UV circular dichroism (CD) spectra of freshly preparedsolutions of mSAA1 (termed SAA for brevity) at 25 °C, pH 3.0–7.5,and the α-helical content determined from these spectra. The resultswere independent of the protein or salt concentration in the rangeexplored (0.02–0.5 mg/mL SAA, 0–150 mM NaCl in 50 mM phos-phate buffer). At pH 7.5 in the absence of bound ligands, SAA waspredominantly in a random coil conformation characterized by a CDminimum circa 200 nm. At pH 3.0, SAA was also largely unfolded.Surprisingly, at pH 4.3, the CD minimum shifted to ∼210 nm and astrong maximum appeared circa 190 nm, indicating a substantiallyfolded structure. This structure was detected at pH 3.5–4.5, whichoverlaps the lysosomal pH range, with the α-helical content reachinga maximum of 35 ± 5% at pH 4.3. At higher and lower pH, thehelical content declined to ∼10% (Fig. 1A, Inset). To our knowledge,

such a pH-dependent structural transition in SAA has not beenpreviously reported.The temperature dependence of the secondary structure at var-

ious pH was monitored by CD at 210 nm during protein heatingand cooling at a constant rate (Fig. 1B). At pH 7.5, SAA waspartially α-helical at 5 °C and showed reversible thermal unfoldingwith a midpoint Tm = 18 °C. At pH 3.0, SAA was also substantiallyα-helical at 5 °C and unfolded reversibly with a Tm = 18 °C. Un-expectedly, at pH 4.3 the unfolding shifted to much higher tem-peratures (Tm = 50 °C) and became thermodynamically irreversible,with a large hysteresis in the melting data (Fig. 1B). Such a hys-teresis is typical of lipid-bound but not lipid-free apolipoproteins,including SAA (30, 31). Irreversible unfolding of SAA at pH 4.3was also evident from far-UV CD spectra recorded before andafter heating to 80 °C, and the difference spectrum showed anirreversible loss of ordered secondary structure (Fig. 1C, Inset).Further, the heating data of lipid-free SAA at any pH exploredshowed no scan rate dependence, suggesting the absence of highactivation energy (30).In summary, far-UV CD results in Fig. 1 revealed that lipid-free

mSAA1 acquired a partially folded conformation at pH 3.5–4.5.The maximally folded state was observed at pH 4.3 and had 35 ±5% α-helical content, 50 ± 5% random coil, with the rest forming aβ-structure. This structure unfolded on heating circa 50 °C, which isunusually high for lipid-free mSAA. The unfolding was ther-modynamically irreversible but did not involve high activationenergy. Further studies focused on pH 4.3 wherein the secondarystructure was maximal; however, experiments at pH 3.5–4.5showed similar trends.

Effects of pH on the Hydrophobic Residue Packing and SAA Aggregationin Solution. To monitor pH-dependent aromatic packing, we usedintrinsic Trp fluorescence by taking advantage of W18, W29, andW53 in mSAA1. These tryptophans are located in or near theconcave apolar face of the SAA monomer that was proposed tobind lipids (32) (see SI Results for details). At pH 3.0 and 7.5, theemission maximum was centered at λmax = 350 nm, indicatingexposed Trp (Fig. 2A), which was consistent with the disorderedsecondary structure observed at these pHs by CD (Fig. 1A). AtpH 4.3, a blue shift to λmax = 342 nm indicated a structural re-organization with partial shielding of Trp (Fig. 2A), consistent withmore ordered secondary structure observed at pH 4.3 (Fig. 1A).To probe for solvent-accessible hydrophobic cavities, we used

a fluorescent probe 8-anilinonaphthalene-1-sulfonic acid (ANS).

Fig. 1. Effects of pH on the secondary structure and thermal stability of SAA in solution. (A) Far-UV CD spectra of SAA at 25 °C. Sample conditions in all panelswere 0.1 mg/mL SAA in 50 mM sodium phosphate buffer (which is the standard buffer used throughout this study) at indicated pH. (Inset) α-helical content atvarious pHs was determined by spectral deconvolution using the K2D program in Dichroweb server as described in Materials and Methods. Error bars indicateSD from the mean of five independent measurements. (B) Heat-induced secondary structural changes in SAA at selected pHs. Far-UV CD melting data, Θ210(T),were recorded at 210 nm during heating and cooling at a rate of 70 °C/h; similar results were obtained at a rate of 10 °C/h. Arrows indicate directions oftemperature changes. (C) Far-UV CD spectra at pH 4.3, 25 °C, recorded from the same SAA sample that was either intact (premelt, solid line) or had beenheated to 80 °C to record the melting data in B (postmelt, dotted line). SAA heating caused an irreversible loss of the secondary structure characterized by thedifference CD spectrum before and after heating (Inset), which apparently involved α-helices and β-sheets.

E6508 | www.pnas.org/cgi/doi/10.1073/pnas.1707120114 Jayaraman et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0

ANS alone showed very weak emission with λmax = 525 nm, whichdid not change on addition of SAA at pH 7.5 and changed verylittle at pH 3, 25 °C (Fig. 2B), indicating lack of ANS bindingto SAA at these pHs. This result was consistent with the lackof ordered secondary structure in SAA observed by CD atpH 7.5 and pH 3, 25 °C (Fig. 1A). In contrast, at pH 4.3, theemission greatly increased and showed a 45-nm blue shift toλmax = 480 nm, indicating ANS binding (Fig. 2B). Conse-quently, the ordered secondary structure in SAA detected byCD at pH 4.3 contained hydrophobic cavities large enough tobind ANS. ANS binding to SAA at pH 4.3, but not at pH 3.0 or7.5, was confirmed in the energy transfer experiments. Trp wasexcited at λex = 295 nm, and the ANS emission was recorded at300–600 nm. No energy transfer from Trp to ANS was detectedat either pH 3.0 or pH 7.5. In stark contrast, a nearly completeenergy transfer was observed at pH 4.3, indicating ANS bindingin close proximity to tryptophans (Fig. 2C).Together, the results in Figs. 1 and 2 revealed that at pH 4.3,

25 °C, SAA acquired a partially folded structure with high α-helicalcontent, partially buried Trp, and large solvent-accessible hydro-phobic cavities. Given the small size of SAA and its high aggregatingpropensity, such cavities likely formed within protein oligomers.To test for oligomerization, we performed nondenaturing gel

electrophoresis (NDGE) of SAA at pH 3.0–7.5. In this pH range,freshly prepared SAA (1 mg/mL protein) migrated as a singlebroad band with a hydrodynamic size circa 7.5–8 nm (Fig. 3A),indicating self-association. A potential caveat is that the NDGEloading and running buffers, which have pH 8.5, may alter proteinconformation and self-association. Next, SAA (0.5 mg/mL) wascross-linked with 0.01–0.06% glutaraldehyde. At pH 4.3, SDS/PAGE showed sharp bands corresponding to SAAmonomer, dimer,and trimer but no higher order oligomers (Fig. 3B). At pH 3.0 and7.5, additional smears corresponding to hexamers (∼50 kDa) andhigher order oligomers were detected at 0.03–0.06% glutaraldehyde,indicating aggregation. A caveat of glutaraldehyde cross-linking isthat it may miss some oligomeric species (e.g., if lysines are protectedfrom cross-linking). Nevertheless, the results in Figs. 1–3 clearlyshowed that formation of the partially folded stable structure in SAAat pH 4.3 involved compact low-order oligomers and was accom-panied by a decrease in nonspecific protein aggregation.

Effects of pH on SAA Interactions with Phospholipids and Their MutualRemodeling. To test whether the pH-induced structural changesin SAA affected its interactions with lipid bilayers, we analyzedthe ability of SAA to remodel multilamellar vesicles (MLVs)of a model zwitterionic phospholipid 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC; C14:0, C14:0). SAA was previouslyshown to rapidly clear DMPC MLVs (>100 nm) at pH 7.5 and

form SAA–DMPC complexes ∼10 nm in size (31, 33). Here, theclearance time course at various pH was monitored by turbidity at24 °C for 1 h. Rapid clearance was observed at pH 7.5 and 3.0 butnot at pH 4.3 (Fig. 4A). Next, SAA was incubated overnight at24 °C with DMPC MLVs at pH 3.0, 4.3, or 7.5; excess lipid wasremoved by centrifugation; and the samples were subjectedto NDGE. At pH 3.0 and 7.5, SAA formed lipoproteins circa11–18 nm in size (Fig. 4B), consistent with previous studies atpH 7.5 wherein such lipoproteins were explored in detail (31).In contrast, at pH 4.3, such lipoproteins were not detected andprotein migration was consistent with lipid-free SAA (Fig. 4B).Therefore, the ability of SAA to remodel DMPC into lipoproteinswas retained at pH 7.5 and pH 3.0 but was abolished at pH 4.3.Next, we explored SAA interactions with a longer chain mono-

unsaturated lipid, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine(POPC; C16:0, C18:1), which mimics zwitterionic phospholipidsin the plasma membrane. To facilitate the bilayer remodeling ofthis relatively long-chain lipid, POPC was used in the form ofsmall unilamellar vesicles (SUVs, ∼25 nm). SAA and POPCSUVs at 1:100 mol:mol protein:lipid were coincubated overnightat 25 °C. NDGE showed that at pH 7.5 and pH 3.0, all SAAscomigrated with SUVs, indicating binding. At pH 4.3, no bindingwas observed and SAA migrated as a free protein (Fig. 5A).Consequently, regardless of the exact nature of the lipid as-sembly (DMPC MLVs or POPC SUVs), SAA could readily bind

Fig. 2. Effects of pH on the intrinsic Trp fluorescence and ANS binding to SAA in solution. (A) Normalized Trp emission spectra showed a pH-dependent shiftin λmax (vertical lines). Freshly prepared samples contained 0.1 mg/mL SAA in standard buffer. Trp residues in mSAA1.1 (W18, W29, and W53) were excited atλex = 295 nm. The emission was recorded at 25 °C in all panels. (B) ANS binding to SAA in solution monitored by fluorescence. ANS was excited at λex = 350 nm.Sample conditions were 40 μM ANS, 10 μM SAA at indicated pH. Emission spectrum of 40 μM ANS in buffer was shown for comparison (dashed line).(C) Förster resonance energy transfer from Trp to ANS. Sample conditions were as in B. Trp was excited at λex = 295 nm.

Fig. 3. Effects of pH on the self-association of lipid-free SAA. (A) NDGE (4–20%gradient) of SAA (1 mg/mL protein) at pH 3.0–7.5. (B) SDS/PAGE of SAA that hasbeen cross-linked with glutaraldehyde as described in Materials and Methods.The cross-linker concentrations and pH are indicated. All gels were stained withDenville Blue protein stain. The figure represents two composite gels whereinlanes were combined by alignment of the MW markers. White lines indicateboundaries between the spliced parts of the gels.

Jayaraman et al. PNAS | Published online July 25, 2017 | E6509

BIOCH

EMISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0

to a phospholipid surface at pH 7.5 and pH 3.0 but not at pH 4.3.To our knowledge, such pH-dependent apolipoprotein–PC in-teractions have not been previously reported.Cross-linking studies showed that, similar to lipid-free SAAs,

SAAs in the presence of POPC SUVs formed low-order oligomers(dimers and trimers) and large nonspecific aggregates at pH 3.0 andpH 7.5 (Figs. 3B and 5B). However, at pH 4.3 in the presence ofPOPC, both specific and nonspecific self-association of SAA wasgreatly diminished and most SAAs either remained monomeric orformed large aggregates (>200 nm) (Fig. 5B). These pH-dependentchanges in protein oligomerization were accompanied by structuralchanges in the protein and lipid described below.Previously, we showed that binding to either DMPC or POPC

induced α-helical folding in mSAA1 at pH 7.5 (31). Here, weexplored the effects of pH on the secondary structure in SAAand its interactions with POPC. SAA was incubated with POPCSUV at 25 °C overnight at either pH 7.5 or 4.3, and far-UV CDspectra were recorded before and after the incubation. Asexpected, at pH 7.5, binding to POPC induced α-helix folding inSAA, as evidenced by a strong CD maximum observed at 193 nmand minima at 208 nm and 222 nm, which is a signature of anα-helix (Fig. 6A). Surprisingly, at pH 4.3, a very different lipid-induced conformational change was observed: The ∼35% α-helicalstructure characteristic of free SAA at pH 4.3 was partially lost,and the CD spectrum showed a weaker maximum circa 190 nmand a minimum circa 218 nm, which is a signature of a β-sheet(Fig. 6B). Together, CD spectra in Fig. 6 A and B revealed that,in contrast to lipid-induced random coil to α-helix transitionobserved at pH 7.5, SAA underwent a lipid-induced α-helix toβ-sheet transition at pH 4.3. To our knowledge, such a secondarystructural transition in apolipoproteins at the lipid surface hasnot been previously reported.The kinetics of this unusual transition was explored next. SAA

was incubated at 25 °C at either pH 4.3 or pH 7.5, POPC SUVsprepared in the same buffer were added to protein, and the timecourse of the secondary structural changes was monitored for upto 2 h by CD at 210 nm, Θ210(t). The dead time in these ex-periments was ∼1 min. At pH 7.5, the value of Θ210 droppedswiftly during the dead time, and no additional changes weredetected (Fig. 6C, black line), indicating rapid α-helical foldingon the lipid (Fig. 6A). In contrast, at pH 4.3, a gradual increase in

Θ210(t) was observed, indicating a slow α-helix to β-sheet con-version in SAA on the lipid surface, which took many minutes tocomplete (Fig. 6C, gray). Similar pH-dependent trends wereobserved in SAA on addition of DMPC MLV (Fig. S1).To test for mutual remodeling of the protein and lipid, we used

ANS that avidly bound SAA oligomers at pH 3.5–4.5 (Fig. 2 B andC). First, POPC SUVs were equilibrated with ANS for 30 min at24 °C, and the ANS emission was recorded using λex = 350 nm. Atany pH explored, in the presence of POPC SUVs, the emissionshowed a blue shift and a large increase in intensity, indicatingANS binding to SUVs (Fig. 6 D and E and Fig. S2, solid lines).Next, SAA was added at 1:100 mol:mol SAA:POPC at pH 4.3; as aresult, ANS emission gradually decreased, reaching a baseline in∼1 h (Fig. 6E, open symbols). In the presence of either SAA aloneor POPC SUV alone, ANS emission remained constant; Fig. S2shows representative data for POPC SUV alone at pH 4.3, in-dicating that the vesicle–ANS interactions remained invariant atthis pH. Hence, loss of ANS binding by both protein and lipid ontheir coincubation at pH 4.3 indicated their mutual structuralremodeling. The protein remodeling by the lipid was also seen byCD showing the α-helix to β-sheet conversion (Fig. 6B). In starkcontrast, at pH 7.5, when no β-sheet was induced by the lipid (Fig.6A), addition of SAA to SUV did not change the ANS emission(Fig. 6D, open symbols). Therefore, the mutual remodeling of theprotein and lipid was very different at pH 4.3 and pH 7.5.In fact, EM revealed striking pH-dependent structural changes

in lipid vesicles, which were induced by the protein: After 1 h ofincubation with SAA at pH 4.3 but not at pH 7.5, most POPCSUVs disappeared and were replaced with elongated curvy fea-tures (Fig. 6F). Similar remodeling of DMPC MLVs by SAA wasobserved at pH 4.3 but not at pH 7.5 (Fig. 6F). In the absence ofthe protein, no significant vesicle remodeling was observed atthese pHs. Consequently, at pH 4.3, SAA drastically disruptedPOPC SUVs and DMPC MLVs. Together, the results in Fig. 6revealed mutual structural remodeling of SAA oligomers andlipid vesicles. At pH 4.3 but not at pH 7.5, this remodeling in-volved α-helix to β-sheet transition in the protein, a major dis-ruption of lipid vesicles, and a loss of ANS affinity for bothprotein and lipid. All together, these effects suggest strongly

Fig. 4. Effects of pH on SAA interactions with DMPC MLV and their mutualremodeling. (A) Time course of SAA-induced clearance of DMPC MLV (>100 nm)into smaller particles monitored at 24 °C by turbidity at 325 nm. At t = 0, SAAsolution in standard buffer at various pH was added to the freshly preparedDMPC MLV in the same buffer to the final protein:lipid molar ratio of 1:80. Thedata for DMPC alone, which were similar at various pHs, are shown for com-parison. (B) NDGE (4–20% gradient, Denville Blue protein stain) of SAA-DMPCcomplexes formed at various pHs. Mixtures of SAA and DMPC, which were pre-pared as described in A, were incubated overnight at 24 °C before analysis. SAAalone (shown for comparison) migrated similarly at pH 3.0–7.5.

Fig. 5. Effects of pH on SAA interactions with POPC SUVs and their mutualremodeling. (A) NDGE of SAA mixtures with POPC SUVs (>20 nm in size). Ateach pH, SAA and POPC SUV stock solutions were prepared in the same buffer.SAA (43 μM) was incubatedwith POPC SUVs (1:100 mol/mol SAA:POPC) at 25 °Cfor 12 h before analysis. SAA alone was shown for comparison. (B) SDS/PAGE(Denville Blue protein stain) of SAA-POPC mixtures followed cross-linking withglutaraldehyde as described in Materials and Methods. The cross-linker con-centrations and pH are indicated; lipid-free SAA at pH 4.3 is shown for com-parison. White lines indicate boundaries between the spliced part of the gel.

E6510 | www.pnas.org/cgi/doi/10.1073/pnas.1707120114 Jayaraman et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0

SAA’s potential to disrupt cell membranes at near-lysosomal pHbut not in plasma.

Effects of pH and Lipids on SAA Proteolysis. To explore the effects ofpH on the protein susceptibility to limited proteolysis, SAA wasincubated with trypsin at 1:200 enzyme:substrate weight ratio at24 °C, and the aliquots taken at various intervals were analyzed bySDS/PAGE. Lipid-free SAA was explored first. At pH 4.3, freeSAA was unusually resistant to proteolysis: No fragments weredetected even after overnight incubation with trypsin (Fig. 7A). Instark contrast, at either pH 3 or pH 7.5 (Fig. 7B), rapid frag-mentation was observed, and most SAA was converted into8–10 kDa fragments after 5 min. These results were consistent witha well-ordered thermostable secondary structure in lipid-free SAAseen at pH 4.3 vis-à-vis the largely unfolded secondary structureseen at either pH 3 or pH 7.5 at 25 °C (Fig. 1 A and B).Interestingly, the pH dependence of SAA proteolysis was re-

versed on addition of lipid vesicles. SDS/PAGE clearly showed

that in the presence of either DMPC MLVs or POPC SUVs, SAAwas cleaved faster at pH 4.3 than at either pH 3 or pH 7.5 (Fig. 7C).This observation suggested that at pH 7.5 and pH 3, SAA binding tolipid surface observed by NDGE (Figs. 4B and 5A), which inducedα-helical conformation observed by CD (Fig. 6A), protected theprotein against proteolysis. In contrast, at pH 4.3, interactions of SAAwith the lipid surface, which induced an α-helix to β-sheet conversion(Fig. 6B) and altered protein aggregation (Fig. 5B), greatly decreasedSAA protection against proteolysis. This finding is an exception fromthe general trend: Lipid binding by apolipoproteins protects themfrom proteolysis (3, 33, 34).SAA and its tryptic fragments, which were generated either at

pH 7.5 in the absence of lipids (Fig. 7B) or at pH 4.3 in the presenceof either DMPC or POPC (Fig. 7C), were subjected to MALDITOF. The exact mass and, hence, the nature of the protein frag-ments was similar, showing three major peaks with a molecular massof 10.31, 9.65, and 7.81 kDa (Fig. 7D). Therefore, similar flexible

Fig. 6. Mutual structural remodeling of the protein and lipid vesicles at various pHs. (A and B) Lipid-induced pH-dependent secondary structural changes inSAA. Far-UV CD spectra of SAA (0.1 mg/mL) alone or on incubation with POPC SUVs (1:100 molar ratio of SAA:POPC) for 1 h at 25 °C. SAA and POPC sampleswere at either pH 7.5 (A) or pH 4.3 (B). Arrows indicate the directions of lipid-induced changes in the CD signal at 210 nm. (C) Time course of secondarystructural changes in SAA on addition of POPC SUVs. SAA (0.1 mg/mL) was pre-equilibrated for 10 min at 25 °C; at t = 0 min, POPC SUV was added to SAA at a1:100 SAA:POPC molar ratio, and the CD signal was recorded at 210 nm, [Θ210](t). Short horizontal bars show [Θ210] before the addition of lipid; arrowsindicate directions of changes on addition of lipid. (D and E) Fluorescence emission spectra monitor ANS binding to POPC SUV alone (solid lines) or in thepresence of SAA (open triangles) at pH 7.5 (D) and pH 4.3 (E). Dashed line in D shows the emission spectrum of ANS in buffer (λex = 350 nm). Arrow in E showsthe direction of time-dependent spectral changes at pH 4.3 during SAA incubation with POPC SUV and ANS. Sample conditions are 40 μM ANS, 10 μM SAA,1:100 mol:mol SAA:POPC. (F) Electron micrographs of negatively stained POPC SUVs (Left) and DMPC MLVs (Right) after 1 h of incubation with SAA at pH 7.5(Bottom) or pH 4.3 (Top) at 25 °C. Protein:lipid molar ratio was 1:100 for SAA:POPC and 1:80 for SAA:DMPC.

Jayaraman et al. PNAS | Published online July 25, 2017 | E6511

BIOCH

EMISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0

sites in SAA were susceptible to proteolysis at various pHs in theabsence and in the presence of lipids.

Effects of pH on Amyloid Formation by SAA.To determine the effects ofpH on fibril formation, lipid-free SAA (1 mg/mL) was incubated at37 °C for up to 3 d and was analyzed for microscopic and macro-scopic structure using spectroscopic, EM, and biochemical methods.Due to low solubility of SAA at pH ∼7, these studies were per-formed only at an acidic pH; Figs. 8 and 9 show representative re-sults at pH 4.3 and pH 3.0.First, formation of amyloid-like structure in SAA was monitored

for 72 h by using thioflavin T (ThT), a fluorescent dye whoseemission increases on binding to amyloid. At pH 3.0, an increase inThT emission on binding to SAA followed a sigmoidal kinetics(Fig. 8A), which is a signature of the nucleation-growth mechanismcharacteristic of amyloid formation by many proteins (16, 35). Incontrast, at pH 4.3, a larger initial ThT emission was detected thatchanged little over time. These results suggest that during in-cubation SAA underwent larger structural changes at pH 3 thanat pH 4.3 and that these changes were on path to amyloid fibrils atpH 3 but apparently not at pH 4.3.

These conclusions were supported by far-UV CD spectra ofSAA at acidic pH recorded before and after 72 h of incubationat 37 °C. At pH 3.0, SAA showed large spectral changes on in-cubation, indicating a conversion from the predominantly ran-dom coil to the β-sheet–rich conformation (Fig. 8B), which ischaracteristic of amyloid. In contrast, at pH 4.3, little spectralchange was detected (Fig. 8C), and the CD spectra remainedinvariant even after 1 mo of incubation at pH 4.3.SAA aggregates were visualized by negative stain EM. At

pH 3.0, the morphology significantly changed over time, fromcurvy protofibrils observed in a freshly prepared sample to longnonbranching fibrils ∼10 nm in width characteristic of amyloidobserved after 72 h at 37 °C (Fig. 8D, Top). In contrast, at pH 4.3,worm-like heterogeneous aggregates and occasional globularstructures were observed that showed no major morphologicalchanges after 72 h or longer incubation (Fig. 8D, Bottom). To-gether, the results in Fig. 8 revealed high temporal structuralstability of SAA oligomers at pH 4.3, which contrasts with thegradual structural changes culminating in fibril formation at pH 3.The microscopic structure in SAA aggregates at pH 3 and

pH 4.3 was explored using ANS binding. At pH 3, ANS did not

Fig. 7. Effects of pH and lipids on the limited proteolysis of SAA by trypsin. (A) SAA at pH 4.3 was resistance to trypsin digestion. Freshly prepared samples of SAA(0.5 mg/mL) were incubated with trypsin (1:200 enzyme:SAA weight ratio) at 25 °C. At indicated time points, 10 μL of mixture was removed, the reaction wasterminated, and the products were analyzed by 18% SDS/PAGE (Denville Blue protein stain). (B) In contrast to pH 4.3, SAA at pH 7.5 and pH 3 was labile to trypsindigestion. SAA solutions at either pH 4.3 or pH 7.5 were incubated with trypsin for 5 min and analyzed as in A. White line indicates boundaries between thespliced part of the gel. (C) Effects of lipids on the SAA susceptibility to limited proteolysis. Freshly prepared samples containing 0.5 mg/mL SAA were incubated for12 h with either DMPC MLVs or POPC SUVs at a protein:lipid molar ratio of 1:80. Trypsin was added to the incubation mixture, the reaction was terminated after10 min, and the products were analyzed as in A. The results at pH 7.5 and pH 3 were similar. (D) MALDI TOF analysis of the tryptic fragments of SAA generated atpH 4.3 in the presence of POPC as described in C. Intact SAA is shown for comparison. Molecular mass in Da is indicated for each peak. Similar fragments weredetected on digestion of lipid-free SAA at pH 7.5.

Fig. 8. Effects of acidic pH on SAA aggregation at 37 °C probed by ThT fluorescence, far-UV CD, and EM. (A) Time course of amyloid formation monitored by ThTbinding to SAA. Solutions containing 1 mg/mL SAA at either pH 3.0 or pH 4.3 were incubated at 37 °C for up to 72 h. At indicated time points, 20 μM ThT wasadded to sample aliquots containing 10 μM SAA. ThT emission spectra were recorded using λex = 444 nm at 25 °C; emission spectra of ThT alone were subtractedfrom the data. The ThT emission peak centered at 480 nm was integrated from 450 nm to 500 nm, and the integrated intensity was plotted. Each data pointrepresents SD of mean for four independent measurements. (B and C) Secondary structural changes in SAA on incubation at 37 °C. SAA solutions were incubatedat 37 °C for 72 h as described in A, the samples were diluted to 0.1 mg/mL, and far-UV CD spectra were recorded at 25 °C (72 h, dotted lines). These spectraremained invariant on protein incubation for up to 1 mo. The spectra of freshly prepared SAA before incubation were shown for comparison (0 h, solid lines).(D) Electron micrographs of negatively stained SAA (1 mg/mL) before and after incubation at 37 °C. Representative images at 0 h and 72 h were shown. (Scale bar,25 nm.) The aggregate morphology did not change on prolonged incubation for up to 1 mo.

E6512 | www.pnas.org/cgi/doi/10.1073/pnas.1707120114 Jayaraman et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0

bind to freshly prepared SAA (Fig. 9A). After 72 h of incubationat pH 3, a blue shift from λmax = 525 nm to 480 nm and a smallincrease in the emission intensity was observed, indicating in-creased ANS binding. In contrast, at pH 4.3, the emission intensitygreatly increased and was centered at λmax = 480 nm in both thefreshly prepared and incubated samples, indicating much greaterburial of ANS on rapid binding to SAA oligomers at pH 4.3 (Fig.9A, dark gray). Again, no large spectral changes were observed onprolonged incubation at pH 4.3, supporting high temporal struc-tural stability of SAA oligomers at this pH.Protein integrity during incubation at 37 °C was analyzed by

SDS/PAGE. At pH 3, most soluble protein was hydrolyzed intofragments after 72 h of incubation during which the random coil toβ-sheet transition and fibril formation was observed (Fig. 8B). Incontrast, at pH 4.3, most SAA remained intact (Fig. 9B), consis-tent with the formation of a substantially folded stable confor-mation at this pH (Fig. 8C).Finally, to differentiate between soluble oligomers and amyloid

fibrils, we used conformation-specific antibodies A11 and OC.A11 binds generic epitopes in prefibrillar protein oligomers,whereas OC recognizes amyloid fibrils formed by various proteinsbut does not bind to monomers or nonfibrillar aggregates (36).Dot blot analysis showed that at pH 3, freshly prepared SAA didnot bind either A11 or OC (Fig. 9C). However, after SAA hasbeen incubated at pH 3 for 72 h, strong binding of both antibodieswas observed, suggesting formation of prefibrillar oilgomers andmature fibrils. This result was consistent with the EM datashowing mature amyloid fibrils formed on SAA incubation atpH 3.0 (Fig. 8D, Upper Right). In stark contrast, SAA incubated atpH 4.3 for up to 72 h showed A11 binding to both freshly preparedand incubated protein, whereas OC did not bind strongly (Fig.9C). Therefore, at pH 4.3, SAA rapidly formed stable prefibrillaroligomers that did not convert into mature fibrils over time. Thisresult was in excellent agreement with high temporal stability of

SAA aggregates observed at pH 4.3 by CD and fluorescencespectroscopy as well as with the worm-like curvilinear rather thanfibrillar morphology observed by EM before and after SAA in-cubation at pH 4.3 (Figs. 8 and 9A).To test whether the high structural stability of SAA oligomers

formed at pH 4.3 was preserved on changes in pH, SAA was dis-solved at pH 4.3 and then diluted into buffers at pH 3.0–7.5. Fibrilformation, DMPC MLV clearance, and trypsin digestion wereperformed as described in Fig. S3. The results clearly showed thatthe protein’s ability to form amyloid fibrils, remodel lipid vesicles,and resist proteolysis was determined by the final pH and was notsignificantly influenced by the prior exposure to pH 4.3.In summary, our fibrillation studies at acidic pH, including sec-

ondary structural analysis, ThT and ANS binding, negative-stainEM, and antibody binding, revealed distinctly different structuraland aggregation properties of lipid-free SAA at pH 3.0 and pH 4.3(Figs. 8 and 9). At pH 3.0, SAA gradually formed amyloid in aprocess involving a sigmoidal increase in ThT fluorescence; thisprocess culminated in formation of classic amyloid fibrils 10 nm inwidth that bound fibril-specific OC antibody. This process involveda gradual random coil to β-sheet conversion (Fig. 8B) renderingSAA labile to proteolysis (Fig. 9B). In contrast, at pH 4.3, SAArapidly formed soluble prefibrillar oligomers that bound A11 butnot OC, showed unusual temporal structural stability in lipid-freestate in solution, had a substantial α-helical content, were protectedfrom proteolysis, and had solvent-accessible hydrophobic cavitieslarge enough to bind ANS (Figs. 8 and 9). Such compact solubleprefibrillar oligomers are particularly relevant to amyloid diseasebecause of their potential to perturb cell membranes, which issuggested by the vesicle disruption (Fig. 6F), as well as their re-sistance to proteolytic degradation at near-lysosomal pH (Fig. 7A),which may facilitate lysosomal accumulation of SAA.

DiscussionSAA Forms a Potentially Cytotoxic Species at pH 3.5–4.5. This studyreports a compact soluble oligomer formed by lipid-free SAA atpH 3.5–4.5. This oligomer is maximally folded at pH 4.3, with 35%α-helix at 25 °C, and unfolds irreversibly circa 50 °C, thus showingan unusually high thermostability for free SAA (Fig. 1). In theabsence of lipids, this oligomer is remarkably resistant to pro-teolysis (Fig. 7A), and its secondary structure remains invariant fordays on incubation at pH 4.3, 37 °C (Fig. 8C). Further, in contrastto pH 3.0 or pH 7.5, lipid-free SAA at pH 4.3 does not form largenonspecific aggregates (Fig. 3B). Rather, it forms low-order pre-amyloid oligomers that bind the diagnostic dye ThT and theA11 antibody but do not readily convert into mature fibrils (Figs. 8and 9). These oligomers contain large solvent-accessible hydro-phobic cavities that bind ANS (Fig. 2). The hydrodynamic size ofthese oligomers is circa 7.5–8 nm, consistent with the partially foldedSAA hexamers, although larger species such as octamers and dec-amers cannot be excluded; however, glutaraldehyde cross-linkingsuggests that dimers and trimers are the key building blocks (Fig.3B). Unlike similar-size mSAA1 oligomers described in previousdetailed studies (24), the pH 4.3 oligomers are distinct in their highstructural stability (Fig. 1B) and affinity for A11 seen in the freshlydissolved SAA (Fig. 9C). Such stable soluble preamyloid oligo-mers of SAA, which may represent a potential lysosomal storageform of SAA, have not been previously reported.An important property of these oligomers is their aberrant

interactions with lipid bilayers. Despite its 35% α-helical content,SAA at pH 4.3 does not bind to phospholipid vesicles via itsamphipathic α-helices and does not form lipoprotein particles ina manner typical of apolipoproteins (Figs. 4 and 5A). Instead,SAA undergoes a gradual lipid-induced α-helix to β-sheet tran-sition (Fig. 6 B and C), which increases protein’s susceptibility toproteolysis (Fig. 7C). The slow rate of this transition suggeststhat energy barriers must be overcome to reorganize the helicalstructure in oligomeric SAA and convert it into a β-sheet (Fig. 6

Fig. 9. Effects of pH on the structural properties of SAA aggregates formed at37 °C. (A) ANS binding to SAA monitored by fluorescence. Solutions containing1 mg/mL SAA at indicated pH were incubated at 37 °C for 72 h, aliquots con-taining 10 μM SAA were mixed with 40 μM ANS, and the ANS emission spectrawere recorded using λex = 350 nm (72 h, dotted lines). These spectra remainedinvariant on prolonged incubation (up to 1 wk). Spectra of freshly preparedSAA before incubation were shown for comparison (0 h, solid lines). (B) SDS/PAGE (18% gradient, Denville Blue protein stain) of freshly prepared (0 h) andincubated (72 h) SAA samples at indicated pH. (C) Binding of conformation-specific antibodies to SAA oligomers that were formed as described in A. Dotblot using A11, which is specific to prefibrillar oligomers, and OC, which pref-erentially binds to mature fibrils, was performed as described in Materials andMethods. White lines indicate boundaries between the spliced part of the blots.

Jayaraman et al. PNAS | Published online July 25, 2017 | E6513

BIOCH

EMISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0

B and C). Importantly, this transition involves mutual structuralremodeling of the protein and lipid, including disruption ofphospholipid vesicles (Fig. 6F). We hypothesize that such anα-helix to β-sheet conversion in soluble oligomeric SAA and theconcomitant disruption of lipid bilayers may be an early step inthe cytotoxic pathway of protein misfolding in amyloid. Becauseno cells are viable in the pH range of 3.5–4.5 where the compactSAA oligomers were stable, this hypothesis could not be testeddirectly in cytotoxicity studies; similarly, the effects of pH ontoxicity of other amyloidogenic proteins such as amylin cannot bedetermined directly in cell cultures (37). However, strong credibilityto our idea is lent by the observation that SAA at pH 4.3 disruptsPOPC SUVs and DMPC MLVs (Fig. 6F) and, hence, has a highpotential to perturb cell membranes.To our knowledge, such unusual pH-dependent interactions

with lipids have not been previously reported for SAA or any otherHDL proteins. In fact, the accepted concept is that lipid surfacebinding by apolipoproteins stabilizes their amphipathic α-helicesand protects them from proteolytic degradation and misfolding(see refs. 3, 33, 34 and references therein). SAA interactions withPCs at pH 4.3 provide an exception from this general rule.Other structural details of the SAA oligomers are presented in

SI Results. They provide evidence for partial folding of SAA helicesh1 and h3 (residues 1–27 and 51–69) at pH 3.5–4.5 and proposethe titration of acidic residues as a possible driving force for thepH-dependent structural changes in lipid-free SAA.

Potential Relevance to Amyloid Disease. Studies of other intrinsicallydisordered amyloidogenic proteins and peptides, such as Alzheimer’sAβ peptide and synuclein, have postulated that an α-helix to β-sheetconversion in partially folded soluble protein oligomers can be anearly step in amyloid formation (see refs. 38, 39 and referencestherein). Furthermore, compact soluble polymorphic low-orderoligomers of various amyloidogenic proteins, including SAA (21–24), are thought to comprise the pathogenic species that can disruptcell membranes and/or mediate prion-like transmission (7, 8, 13–15,40). Notably, studies of soluble low-order oligomeric forms of Aβhave revealed a direct correlation between the ANS binding andcytotoxicity (9), whereas studies of a model amyloidogenic proteinHypF reported that “structural flexibility and hydrophobic exposureare primary determinants of the ability of oligomeric assemblies tocause cellular dysfunction” (13). We hypothesize that small solubleoligomers of SAA, such as those formed circa pH 4.3, which bindANS, disrupt phospholipid bilayers, and convert from an α-helix toβ-sheet on interacting with these bilayers, may be involved in thepathogenesis of AA amyloidosis. We speculate that such SAAoligomers can disrupt cellular membranes and liberate the amyloidinto the extracellular space.

Although in vivo formation of such SAA oligomers remains to beestablished, their observation in vitro only at pH 3.5–4.5 suggestslimited sites for their possible formation in the body. Such oligo-mers may be stable in the gastric tract, which might contribute tooral transmission of AA amyloidosis in animals (25). Besides thegastric tract, pH 3.5–4.5 can be found in vivo only in urine or inlysosomes. A fraction of SAA that is filtered in the kidney may bereabsorbed via the receptor-mediated endocytosis, followed byacidification of endocytic vesicles and their fusion with lysosomes(41). Because kidney is the primary site of AA deposition, thispathway of SAA exposure to lysosomal pH in the kidney maycontribute to renal AA amyloidosis. Moreover, SAA internalizationin lysosomes of various cells, including macrophages that are as-sociated with tissue amyloid (21, 42), and the lysosomal origin ofAA fibrillogenesis is well established (26, 28, 29). Our finding ofunusual SAA oligomers at pH 3.5–4.5 helps provide molecular basisfor the lysosomal origin of AA fibrillogenesis. We posit that highstructural stability of SAA oligomers at lysosomal pH and theirresistance to proteolysis help them escape lysosomal degradationand accumulate in lysosomes, reaching high SAA concentrationsthat favor amyloid fibril formation. Further, the membranolyticpotential of SAA oligomers helps this amyloid to escape from thecells, ultimately leading to massive deposition of extracellular am-yloid in AA disease. Moreover, SAA oligomers that disrupt lipidbilayers may potentially contribute to the cytotoxic effects in AAamyloidosis. We posit that such soluble prefibrillar oligomers, whichare stable at pH 3.5–4.5 but not in plasma, provide a missing link inour understanding of AA amyloid formation.

Materials and MethodsRecombinant murine SAA1 (mSAA1, 103 aa, 11.6 kDa) was expressed inEscherichia coli and purified to 95% purity as described previously (9). All othermaterials, which were commercially obtained as described in SI Materials andMethods, were of highest available purity. Methods of sample preparationand analysis by using an array of biophysical and biochemical techniques aredescribed in SI Materials and Methods.

Note Added in Proof. While this paper was in press, a cell-based study cameout demonstrating that mSAA1 accumulation in lysosomes leads to membranedisruption, lysosomal leakage, and cell death, while the preformed amyloid isdeposited outside the cells and seeds the fibrillation of extracellular SAA (43).The current paper provides the biophysical basis for these cell-based studiesthat are in excellent agreement with our proposed mechanism.

ACKNOWLEDGMENTS. We are grateful to Dr. Andrea Havasi for very helpfulcomments and all members of the O.G. laboratory for useful discussions andhelp. This work was supported by National Institutes of Health Grant GM067260 (to S.J., D.L.G., and O.G.), the Stewart Amyloidosis Research Endow-ment Fund (O.G.), and Deutsche Forschungsgemeinschaft Foundation GrantHA 7138/2-1 (to C.H.).

1. De Buck M, et al. (2016) Structure and expression of different serum amyloid A (SAA)variants and their concentration-dependent functions during host insults. Curr MedChem 23:1725–1755.

2. van der Westhuyzen DR, de Beer FC, Webb NR (2007) HDL cholesterol transport duringinflammation. Curr Opin Lipidol 18:147–151.

3. Kisilevsky R, Manley PN (2012) Acute-phase serum amyloid A: Perspectives on its physi-ological and pathological roles. Amyloid 19:5–14.

4. Obici L, Merlini G (2012) AA amyloidosis: Basic knowledge, unmet needs and futuretreatments. Swiss Med Wkly 142:w13580.

5. Westermark GT, Fändrich M, Westermark P (2015) AA amyloidosis: Pathogenesis andtargeted therapy. Annu Rev Pathol 10:321–344.

6. Röcken C, Shakespeare A (2002) Pathology, diagnosis and pathogenesis of AA amy-loidosis. Virchows Arch 440:111–122.

7. Hirakura Y, Carreras I, Sipe JD, Kagan BL (2002) Channel formation by serum amyloid A:A potential mechanism for amyloid pathogenesis and host defense. Amyloid 9:13–23.

8. Gellermann GP, et al. (2005) Raft lipids as common components of human extracel-lular amyloid fibrils. Proc Natl Acad Sci USA 102:6297–6302.

9. Kollmer M, et al. (2016) Electron tomography reveals the fibril structure and lipidinteractions in amyloid deposits. Proc Natl Acad Sci USA 113:5604–5609.

10. Pineda A, Burré J (2017) Modulating membrane binding of α-synuclein as a thera-peutic strategy. Proc Natl Acad Sci USA 114:1223–1225.

11. Morgado I, Garvey M (2015) Lipids in amyloid-β processing, aggregation, and toxicity.Adv Exp Med Biol 855:67–94.

12. Singh S, Trikha S, Bhowmick DC, Sarkar AA, Jeremic AM (2015) Role of cholesterol andphospholipids in amylin misfolding, aggregation and etiology of islet amyloidosis.Adv Exp Med Biol 855:95–116.

13. Campioni S, et al. (2010) A causative link between the structure of aberrant proteinoligomers and their toxicity. Nat Chem Biol 6:140–147.

14. Stefani M (2012) Structural features and cytotoxicity of amyloid oligomers: Implications inAlzheimer’s disease and other diseases with amyloid deposits. Prog Neurobiol 99:226–245.

15. Guerrero-Muñoz MJ, Castillo-Carranza DL, Kayed R (2014) Therapeutic approachesagainst common structural features of toxic oligomers shared by multiple amyloi-dogenic proteins. Biochem Pharmacol 88:468–478.

16. Chiti F, Dobson CM (2006) Protein misfolding, functional amyloid, and human disease.Annu Rev Biochem 75:333–366.

17. Tipping KW, van Oosten-Hawle P, Hewitt EW, Radford SE (2015) Amyloid fibres: Inertend-stage aggregates or key players in disease? Trends Biochem Sci 40:719–727.

18. Westermark GT, Westermark P (2010) Prion-like aggregates: Infectious agents inhuman disease. Trends Mol Med 16:501–507.

19. Gillmore JD, Lovat LB, Persey MR, Pepys MB, Hawkins PN (2001) Amyloid load andclinical outcome in AA amyloidosis in relation to circulating concentration of serumamyloid A protein. Lancet 358:24–29.

20. Lachmann HJ, et al. (2007) Natural history and outcome in systemic AA amyloidosis. NEngl J Med 356:2361–2371.

21. Shirahama T, et al. (1990) Amyloid enhancing factor-loaded macrophages in amyloidfibril formation. Lab Invest 62:61–68.

E6514 | www.pnas.org/cgi/doi/10.1073/pnas.1707120114 Jayaraman et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0

22. Cathcart ES, Elliott-Bryant R (1999) Diet, amyloid enhancing factor (AEF) and amy-loidogenesis: An hypothesis. Amyloid 6:107–113.

23. Senthilkumar S, Chang E, Jayakumar R (2008) Diffusible amyloid oligomers triggersystemic amyloidosis in mice. Biochem J 415:207–215.

24. Srinivasan S, et al. (2013) Pathogenic serum amyloid A 1.1 shows a long oligomer-richfibrillation lag phase contrary to the highly amyloidogenic non-pathogenic SAA2.2.J Biol Chem 288:2744–2755.

25. Lundmark K, et al. (2002) Transmissibility of systemic amyloidosis by a prion-likemechanism. Proc Natl Acad Sci USA 99:6979–6984.

26. Teng J, Turbat-Herrera EA, Herrera GA (2014) Extrusion of amyloid fibrils to the ex-tracellular space in experimental mesangial AL-amyloidosis: Transmission and scan-ning electron microscopy studies and correlation with renal biopsy observations.Ultrastruct Pathol 38:104–115.

27. Yamada T, Kluve-Beckerman B, Liepnieks JJ, Benson MD (1994) Fibril formation fromrecombinant human serum amyloid A. Biochim Biophys Acta 1226:323–329.

28. Shirahama T, Cohen AS (1975) Intralysosomal formation of amyloid fibrils. Am JPathol 81:101–116.

29. Chronopoulos S, Laird DW, Ali-Khan Z (1994) Immunolocalization of serum amyloid Aand AA amyloid in lysosomes in murine monocytoid cells: Confocal and immunogoldelectron microscopic studies. J Pathol 173:361–369.

30. Gursky O, Ranjana, Gantz DL (2002) Complex of human apolipoprotein C-1 withphospholipid: Thermodynamic or kinetic stability? Biochemistry 41:7373–7384.

31. Jayaraman S, Haupt C, Gursky O (2015) Thermal transitions in serum amyloid A insolution and on the lipid: Implications for structure and stability of acute-phase HDL.J Lipid Res 56:1531–1542.

32. Frame NM, Gursky O (2016) Structure of serum amyloid A suggests a mechanism forselective lipoprotein binding and functions: SAA as a hub in macromolecular in-teraction networks. FEBS Lett 590:866–879.

33. Takase H, et al. (2014) Characterization of reconstituted high-density lipoproteinparticles formed by lipid interactions with human serum amyloid A. Biochim BiophysActa 1842:1467–1474.

34. Das M, Gursky O (2015) Amyloid-forming properties of human apolipoproteins: Se-quence analyses and structural insights. Adv Exp Med Biol 855:175–211.

35. Colón W, Aguilera JJ, Srinivasan S (2015) Intrinsic stability, oligomerization, and amyloi-dogenicity of HDL-free serum amyloid A. Adv Exp Med Biol 855:117–134.

36. Kayed R, et al. (2007) Fibril specific, conformation dependent antibodies recognize ageneric epitope common to amyloid fibrils and fibrillar oligomers that is absent inprefibrillar oligomers. Mol Neurodegener 2:18.

37. Tomasello MF, Sinopoli A, Pappalardo G (2015) On the environmental factors affectingthe structural and cytotoxic properties of IAPP peptides. J Diabetes Res 2015:918573.

38. Fezoui Y, Teplow DB (2002) Kinetic studies of amyloid beta-protein fibril assembly.Differential effects of alpha-helix stabilization. J Biol Chem 277:36948–36954.

39. Uversky VN (2015) Protein misfolding in lipid-mimetic environments. Adv Exp MedBiol 855:33–66.

40. Kayed R, et al. (2004) Permeabilization of lipid bilayers is a common conformation-dependent activity of soluble amyloid oligomers in protein misfolding diseases. J BiolChem 279:46363–46366.

41. Miura K, Takahashi Y, Shirasawa H (1985) Immunohistochemical detection of serumamyloid A protein in the liver and the kidney after casein injection. Lab Invest 53:453–463.

42. Claus S, et al. (2017) Cell-to-cell transfer of SAA1 protein in a cell culture model ofsystemic AA amyloidosis. Sci Rep 7:45683.

43. Claus S, et al. (June 21, 2017) Cellular mechanism of fibril formation from serumamyloid A1 protein. EMBO Rep, 10.15252/embr.201643411.

44. Böhm G, Muhr R, Jaenicke R (1992) Quantitative analysis of protein far UV circulardichroism spectra by neural networks. Protein Eng 5:191–195.

45. Andrade MA, Chacón P, Merelo JJ, Morán F (1993) Evaluation of secondary structureof proteins from UV circular dichroism spectra using an unsupervised learning neuralnetwork. Protein Eng 6:383–390.

46. Micsonai A, et al. (2015) Accurate secondary structure prediction and fold recognitionfor circular dichroism spectroscopy. Proc Natl Acad Sci USA 112:E3095–E3103.

47. McLean LR, Phillips MC (1984) Cholesterol transfer from small and large unilamellarvesicles. Biochim Biophys Acta 776:21–26.

48. Swaney JB, O’Brien K (1978) Cross-linking studies of the self-association properties ofapo-A-I and apo-A-II from human high density lipoprotein. J Biol Chem 253:7069–7077.

49. Lu J, Yu Y, Zhu I, Cheng Y, Sun PD (2014) Structural mechanism of serum amyloidA-mediated inflammatory amyloidosis. Proc Natl Acad Sci USA 111:5189–5194.

50. Derebe MG, et al. (2014) Serum amyloid A is a retinol binding protein that transportsretinol during bacterial infection. eLife 3:e03206.

Jayaraman et al. PNAS | Published online July 25, 2017 | E6515

BIOCH

EMISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

0