Screening Random Peptide Libraries with Subacute Sclerosing Panencephalitis Brain-Derived Recombinant Antibodies Identifies Multiple Epitopes in the C-Terminal Region of the Measles Virus Nucleocapsid Protein

Abstract

Infectious and inflammatory diseases of the CNS are often characterized by a robust B-cell response that manifests as increased intrathecal immunoglobulin G (IgG) synthesis and the presence of oligoclonal bands. We previously used laser capture microdissection and single-cell PCR to analyze the IgG variable regions of plasma cells from the brain of a patient with subacute sclerosing panencephalitis (SSPE). Five of eight human IgG1 recombinant antibodies (rAbs) derived from SSPE brain plasma cell clones recognized the measles virus (MV) nucleocapsid protein, confirming that the antibody response in SSPE targets primarily the agent causing disease. In this study, as part of our work on antigen identification, we used four rAbs to probe a random phage-displayed peptide library to determine if epitopes within the MV nucleocapsid protein could be identified with SSPE brain rAbs. All four of the SSPE rAbs enriched phage-displayed peptide sequences that reacted specifically to their panning rAb by enzyme-linked immunosorbent assay. BLASTP searches of the NCBI protein database revealed clear homologies in three peptides and different amino acid stretches within the 65 C-terminal amino acids of the MV nucleocapsid protein. The specificities of SSPE rAbs to these regions of the MV nucleocapsid protein were confirmed by binding to synthetic peptides or to short cDNA expression products. These results indicate the feasibility of using peptide screening for antigen discovery in central nervous system inflammatory diseases of unknown etiology, such as multiple sclerosis, neurosarcoidosis, or Behcet's syndrome.

Panning of phage-displayed random peptide libraries allows an unbiased selection of antibody epitopes/mimotopes without preconceptions about the nature of the target antigens (reviewed in reference 23). In central nervous system (CNS) inflammatory diseases, where access to active diseased tissue is limited or where the levels of tissue antigen may be extremely low, phage peptide panning provides an alternative and sensitive avenue for antigen identification. Panning of phage-displayed random peptide libraries has successfully identified rheumatoid factor-specific mimotopes (22) and allergen mimotopes (19) and has mapped both linear and discontinuous viral epitopes recognized by antibodies specific for various infectious agents (5, 10, 11, 16, 20).

Infectious and inflammatory diseases of the CNS are often characterized by increased intrathecal immunoglobulin G (IgG) synthesis that is seen as discrete bands of oligoclonal IgG when cerebrospinal fluid (CSF) or brain IgG is separated by isoelectric focusing. In CNS infectious diseases, such as subacute sclerosing panencephalitis (SSPE), neurosyphilis, mumps meningitis, progressive rubella panencephalitis, cryptococcal meningitis, and varicella zoster virus vasculitis, the oligoclonal IgG is directed largely against the infectious agent that causes the disease (reviewed in reference 12). Increased CNS IgG synthesis is accompanied by an elevated number of CD19⁺ B cells and the appearance of post-germinal center reaction plasmablast/plasma cells (4, 6, 18). We previously used laser capture microdissection and single-cell PCR to identify expanded clones among the CD38⁺ B cells found within the parenchyma of SSPE brain (3). Five of the eight human IgG1 recombinant antibodies (rAbs) prepared from these expanded plasma cell clones recognized the measles virus (MV) nucleocapsid (N) protein, confirming that infiltrating B cells in SSPE brain target primarily the causative agent of the disease. In this study, four MV-specific rAbs derived from SSPE brain B cells were used to probe a random phage-displayed peptide library. To further define the specificity of the intrathecal response in SSPE, our goal was to map the MV-specific protein epitopes recognized by these rAbs. A second aim was to determine whether the peptide sequences selected by panning could identify target antigens in database searches and thus be applicable to antigen identification in diseases, such as multiple sclerosis (MS), where the target of the intrathecal antibody response is unknown.

MATERIALS AND METHODS

Construction and generation of recombinant antibodies from plasma cell clones.

Full-length human IgG1 rAbs were produced in mammalian tissue culture cells (Invitrogen, Carlsbad, Calif.) from V region sequences of expanded SSPE brain plasma cell clones as described previously (3). Briefly, V_H regions were cloned into pIgG1Flag, a modified version of the episomal expression vector pCEP4 (Invitrogen) that contains a full-length human IgG1 C region domain and a C-terminal Flag epitope (21). Amplified full-length light-chain (L-chain) sequences were cloned independently into the unmodified pCEP4 vector.

Endonuclease-freelarge-scale plasmid Maxiprep or Megaprep kits (QIAGEN, Valencia, Calif.) were used to purify expression vector DNA for each heavy-chain (H-chain) and L-chain construct. Recombinant IgG was produced either in suspension cultures using Freestyle 293 (293 F) cells or in adherent HEK 293-EBNA cells. 293 F cells were grown and maintained in FreeStyle 293 expression medium at 37°C with 8% CO₂, 85% humidity, and shaking at 125 rpm. 293 F cells were seeded in 42 ml of fresh medium at a density of 1 × 10⁶ cells/ml and cotransfected with 20 to 25 μg each of H- and L-chain plasmid DNA and 60 μl of 293fectin (Invitrogen) according to the manufacturer's recommendations. The cells were subsequently grown for an additional 72 to 96 h, and the supernatant was collected and recombinant-IgG affinity purified using protein A-Sepharose beads (Sigma, St. Louis, Mo.). The purified IgG was concentrated to ∼1 ml using Centricon YM 30 centrifugal filter devices (Millipore) and dialyzed against phosphate-buffered saline (PBS) overnight at 4°C. The dialyzed antibody was quantified using the bicinchoninic acid protein assay kit (Pierce Chemical Co., Rockford, Ill.) and stored at 4°C in PBS supplemented with 0.1% protease-free and IgG-free bovine serum albumin (BSA), 0.002% NaN₃.

293-EBNA cells were propagated in Dulbecco modified Eagle high-glucose medium supplemented with 10% fetal calf serum, penicillin-streptomycin, and 250 μg/ml of Geneticin. To transfect 293-EBNA cells, 18 ml of fresh medium without antibiotics was added to 70%-confluent cells in T-150 flasks, followed by 6 μg each of H- and L-chain plasmid DNA and 12 μl of Lipofectamine 2000 (Invitrogen) according to the manufacturer's recommendations. Supernatants were harvested 3 to 5 days after transfection, and recombinant IgG was affinity purified and stored as noted above.

The production of intact, disulfide-bonded rAbs was confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) under nonreducing conditions and immunoblot detection with anti-human IgG or anti-Flag antibodies. Purified rAb resolved as a dominant 160-kDa protein band.

Biopanning.

The PhD.-12 phage display (random) peptide library (New England BioLabs Inc., Beverly, Mass.) was panned against SSPE rAbs using an alternate epitope-mapping procedure. In this method, the random peptide library is incubated with rAb in solution, followed by affinity capture of rAb-phage complexes onto protein A-Sepharose or protein G-agarose beads. Three rounds of panning were carried out for each rAb. To prepare the affinity resin, protein A-Sepharose (50 μl of a 50% aqueous suspension in Tris-buffered saline [TBS]) was combined with 1 ml of Tris-buffered saline-0.1% Tween 20 (TBST) in a microcentrifuge tube and collected by centrifugation for 30 s at 6,000 rpm. The supernatant was removed, and the protein A pellet was blocked by suspension in 1 ml of TBS-1% BSA (blocking buffer) for 60 min at 4°C with occasional mixing. The protein A-Sepharose was collected by centrifugation and washed four times with TBST (1 ml), and the resin was pelleted after each wash. To initiate each panning experiment, 1.5 × 10¹¹ PFU of random-peptide-library M13 phage was combined with rAb (300 ng) in a 200-μl final volume of TBST and incubated at room temperature (RT) for 20 min. The phage-antibody mixture was incubated with the BSA-blocked and -washed protein A resin at room temperature for 15 min with occasional mixing. The suspension was centrifuged for 30 s, and the supernatant was removed from the pellet. After the resin was washed 10 times with TBST (1 ml), bound phage was eluted by suspending the protein A beads in 1 ml of 0.2 M glycine-HCl (pH 2.2)-0.1% BSA (1 ml) (elution buffer) at room temperature for 8 min. The elution mixture was centrifuged for 1 min and the eluate transferred to a new tube and immediately neutralized with 150 μl of 1 M Tris-HCl, pH 9.1. The titer of the eluted phage particles was determined, and the phage was amplified as described below. A second round of panning was performed using amplified phage (1.5 × 10¹¹ PFU) recovered from the first-round eluate. To capture antibody-phage complexes during the second-round panning, protein G-agarose was substituted for protein A-Sepharose. To reduce nonspecific interactions, the concentration of Tween 20 in TBST was increased to 0.5% in the second and third pannings. The third round of panning was carried out using amplified second-round eluate (1.5 × 10¹¹ PFU) and protein A-Sepharose beads to capture phage-IgG complexes. The titer of the third-round eluate was determined, but it was not amplified. Single plaques from the third-round titer plates were amplified, and the DNA was sequenced.

Phage titration and amplification and DNA sequencing.

Mid-log-phase Escherichia coli ER 2738 cells (optical density at 600 nm, ∼0.5) were infected with serial dilutions of M13 phage, added to 3 ml of top agarose, and plated on 100-mm plates of LB agar supplemented with IPTG (isopropyl-â-d-thiogalactopyranoside) and X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside). Blue-colored plaques were counted, and the M13 titer was determined. To amplify phage for subsequent rounds of panning, eluted phage (100 μl) was incubated for 4.5 h at 37°C with 20 ml of a 1:100 dilution of E. coli cells grown overnight in LB medium. To amplify single, randomly picked clones, individual phage plaques were excised and amplified in 1 ml of diluted E. coli cells for 4.5 h at 37°C. For large-scale amplification, 5 μl of phage suspensions from the previous 1-ml amplified-phage preparations were grown in 20 ml of diluted E. coli cells.

Amplified phage was purified from supernatants of phage-infected cell cultures collected after centrifugation at 10,000 rpm for 15 min at 4°C in a Sorvall SS-34 rotor. Amplified phage was precipitated overnight at 4°C with one-sixth volume of 20% polyethylene glycol 8000 (PEG 8000), 2.5 M NaCl (PEG 8000-NaCI), and the phage-PEG 8000 pellet was collected by centrifugation at 10,000 rpm for 15 min at 4°C. Purified phage was suspended in 200 μl TBS and the titer determined. For phage DNA purification, 500 μl of the 1-ml single-plaque suspension was precipitated with PEG 8000, and the phage-PEG 8000 pellet was suspended in 100 μl iodide buffer (10 mM Tris-HCl [pH 8.0], 1 mM EDTA, 4 M NaI). Single-stranded phage DNA was precipitated by incubation with 250 μl of 100% ethyl alcohol at room temperature for 10 min. DNA was collected by centrifugation, washed with 70% ethyl alcohol, air dried, and dissolved in 30 μl TE buffer (10 mM Tris [pH 8.0], 1 mM EDTA). Phage DNA (5 μl) was sequenced with the M13 sequencing primer −96 gIII (New England BioLabs).

Peptide synthesis.

Commercially synthesized peptides were obtained from Sigma Genosys (The Woodlands, Tex.) or Global Peptide (Ft. Collins, Colo.). Purity was estimated to be greater than 75 to 90% by reverse-phase high-performance liquid chromatography. Each peptide contained 10 to 12 amino acids (aa) of the specific peptide sequence of interest followed by the C-terminal sequence GGGC. The cysteine residue was used to couple the peptide tomaleimide-activated enzyme-linked immunosorbent assay (ELISA) plates (Pierce, Rockford, Ill.). The C-terminal carboxyl group on the cysteine residue was amidated to block the negative charge.

M13 phage ELISAs.

Each well of a radioimmunoassay (RIA) plate was coated overnight at 4°C with a 100-μl volume of rAb (5 μg/ml in TBS) or control human IgG (Alpha Diagnostics International, San Antonio, Tex.). Wells were blocked for 1 to 2 h with 3% BSA, and the amounts of selected phage indicated in the legend to Fig. 1 (usually 1 × 10¹⁰ M13 phage particles) were added to each well. After 1 h at RT, the wells were washed with 0.5% Tween 20-TBS, incubated with a 1:500 dilution of horseradish peroxidase (HRP)-conjugated mouse anti-M13 antibody for 1 h, washed again with TBST, and incubated with the peroxidase substrate ABTS [2,2-azinobis(3-ethylbenzthiazolinesulfonic acid)] (Zymed Laboratories Inc., San Francisco, Calif.) for 20 min at room temperature. Absorbance was determined at 415 nm using a Benchmark microplate reader (Bio-Rad). All samples were tested at least twice in duplicate or triplicate.

FIG. 1.

ELISA binding of phage-displayed peptides to rAb IgG and control IgG. Wells of an RIA plate were coated with brain-derived SSPE rAbs (1, 2B4, 3B, and 5), control preimmune human IgG (Alpha Diagnostic International, San Antonio, Tex.), or BSA alone and assayed for the level of binding to 1 × 10¹⁰ particles (5 × 10¹⁰ phage were used in the ELISAs with rAb 1 whose results are shown in panel A) of the indicated M13 phage clones. The M13 clones 1-C and 1-R, (A); 2B4-B, 2B4-C, 2B4-N, and 2B4-X (B); 3B-C and 3B-J (C); and 5-A, 5-D, and 5-L (D) used in the ELISAs (see displayed peptide sequences in Table 3) were assessed. The binding of a control phage clone (RAND) randomly selected from the unpanned M13 library to each rAb was also assayed. All samples were tested in duplicate or triplicate, and all tests were repeated at least once. At least one enriched phage population from each panning experiment showed specific binding to its panning rAb compared to its binding to control human IgG. None of the rAbs showed significant nonspecific binding to the randomly selected phage populations.

Peptide ELISAs.

Wells of preblocked Reacti-Bind maleimide-activated clear strip plates (Pierce, Rockford, Ill.) were coated overnight at 4°C with synthetic peptides (100 μl of a 50-μg/ml solution in PBS). After a brief washing with PBS, the wells were blocked with cysteine (25 μg/ml) for 1 h at RT. For dose-response curves, the wells were incubated for 60 to 90 min at room temperature with increasing concentrations of SSPE rAbs in a final 100-μl volume. After six washes with TBST, the wells were incubated for 60 min at room temperature with a 1:500 dilution of HRP-conjugated anti-human IgG Fc-specific antibody. The wells were again washed with TBST, and IgG binding was detected with ABTS substrate. In competitive ELISAs, SSPE rAb 2B4 (10 ng/ml) was preincubated for 30 min at room temperature with various concentrations of antibody-specific and control phages before being added to peptide-coated wells. Bound antibody was detected as described above.

Screening of an SSPE brain-derived phage display cDNA library.

To obtain cDNA clones expressing short linear stretches of the MV N protein, a T7 phage cDNA expression library prepared from SSPE brain mRNA was screened using SSPE rAb 2B4 or IgG purified from SSPE brain. Briefly, an aliquot (<5,000 PFU/plate) of library phage was used to infect BLT5615 cells and plated onto LB plates supplemented with carbenicillin (50 μg/ml) as described previously (2). Plaque proteins were transferred from each plate onto nitrocellulose filters. After being blocked in 3% BSA-TBS, the filters were incubated with antibody (1 to 5 μg/ml) for at least 2 h in TBS blocking buffer, washed five times for 5 min each in TBST, and incubated for 1 h with a 1:500 dilution of alkaline phosphatase-conjugated goat anti-human IgG. After additional washes in TBST, positive plaques were detected using a nitroblue tetrazolium substrate. Positive plaques were randomly selected and placed into sterile LB medium. The cDNA insert from each clone was amplified using T7Select primers flanking the cDNA cloning site and sequenced as described previously (2). For ELISAs, rAbs were coated onto microtiter plates and assayed for binding to 2 μg (1 × 10⁹ PFU) of purified T7 phage particles as described above for the M13 phage ELISAs. Phage binding was detected using a 1:1,000 dilution of HRP-conjugated T7 Tag antibody (Novagen, Madison, Wis.) and ABTS as the substrate.

Sequence analyses.

Single-stranded phage DNA was sequenced by the University of Colorado Cancer Center DNA Sequence Core. The deduced amino acid sequence of each selected peptide sequence was obtained using the DNASIS MAX software package (MiraiBio, Alameda, Calif.). BLASTP searches for short, nearly identical sequence matches were conducted using the NCBI (http://www.ncbi.nlm.nih.gov) website.

RESULTS

SSPE brain-derived rAbs.

Four brain-derived SSPE rAbs directed against the MV N protein were used to pan a phage-displayed random peptide X₁₂ library. Three of the MV N-specific rAbs (rAB 1, rAb 3B, and rAb 5) were generated from laser capture microdissection analysis of SSPE brain (3), and the fourth MV N-specific rAb (rAb 2B4) was obtained by screening an SSPE brain-derived Fab antibody library (1). The rAbs were chosen for epitope mapping because they represented three of the major plasma cell clones identified in the SSPE 83 brain repertoire and bound strongly to the MV N protein in Western blots. Table 1 lists the V region germ line segments, their relative abundances in the SSPE brain repertoire, and the extent of somatic mutation for the expressed H- and L-chain V region sequences of each rAb used in this study.

TABLE 1.

SSPE recombinant antibodies used for affinity selection of phage-displayed peptides^a

SSPE rAb	V region	Germ line	CDR3 amino acid sequence	Homology (%)	Frequencyb
1	VH3	3-23	ALKKGEGGLRFLEYYFD	92.1	14/65
	Vλ4	4b	QTWGSGMGV	93.7
3B	VH3	3-09	IRAGAFD	95.2	6/65
	Vκ2	A23	MQATQSWTF	98.2
5	VH4	4-34	GGLAARARLVLARMD	93.8	4/65
	Vκ1	O12	QQSYNTPITF	95.1
2B4	VH3	3-30	IAGRAY	91.7	NAc
	Vκ4	B3	QQYYSTPLT	96.1

^aFor each rAb, the heavy-chain sequence is listed above its corresponding light-chain partner. Each row contains the IgG variable heavy-or light-chain CDR3 amino acid sequence, the most homologous V region germ line segment, and the extent of germ line homology.

^bFrequency indicates the number of plasma cells out of the total number analyzed in that repertoire expressing the same VDJ rearrangement as indicated by the CDR3 amino acid sequence (3).

^cNA, not applicable. rAb 2B4 was obtained by panning a phage-displayed Fab antibody library prepared from SSPE brain (1).

Enrichment of specific phage peptides.

To screen the phage-displayed library for high-affinity peptide ligands, solution phase binding to individual rAbs was performed as described in Materials and Methods. After each round of panning, the titer of the eluted phage was determined. Comparison of the ratios of eluted phage to input phage from round 3 to round 1 (Table 2) revealed a marked enrichment of phage in all panning experiments. The most striking was the 10⁴- to 10⁵-fold enrichment observed for rAbs 2B4 and 3B.

TABLE 2.

Phage enrichment after panning of random peptide libraries with SSPE brain-derived recombinant antibodies^a

rAb	Panning round			Enrichment (fold)b
	1	2	3
1	5.4 × 104	9.6 × 104	4.7 × 106	87
2B4	9 × 103	9 × 107	5 × 108	5.6 × 104
3B	6.1 × 104	5 × 108	2.5 × 1010	4 × 105
5	3.3 × 104	7.3 × 104	2.2 × 108	6.7 × 103

^aAfter each round of panning, the titer of the eluted phage selected by each rAb was determined.

^bEnrichment represents the ratios of the output to the input phage titers of the third round of panning compared to the ratios from the first round of panning.

After the third round of panning, ∼20 randomly picked plaques from each antibody screen were amplified and their phage DNA sequenced. Table 3 lists the deduced amino acid sequence and relative abundance of a representative subset of the peptide sequences recovered from each panning experiment. One or more specific peptide sequences were enriched in each panning experiment. Most interesting were peptides selected by rAbs 2B4 and 3B; each experiment yielded several different peptide sequences with shared common structural features (Table 3). Alignment of phage-displayed peptide sequences enriched by SSPE rAb 2B4 revealed two distinct amino acid motifs. The first was the consensus sequence S-(T/S)-W-Y-(D/E)-W-(Q/N)-P detected in peptides expressed by 16/18 phage isolates. The amino acids indicated in boldface were highly conserved and located at the same positions in 50% or more of the unique peptides sequenced. A second, less-abundant peptide motif, N-(Q/A)-L-L-R-(I/Q)-Q-A, was detected in 1/18 sequences (2B4-N) from the solution phase panning experiment and in 1/20 sequences analyzed from a separate solid phase panning experiment. Phage-displayed peptide sequences similar to the group 1 motif were also detected by solid phase screening. For example, a peptide sequence identical to that expressed by 2B4-F phage was a dominant sequence recovered from the rAb 2B4 solid phase screening (4/20 phage sequences).

TABLE 3.

Phage-displayed peptide sequences selected by SSPE recombinant antibodies^a

SSPE rAb	Phage	Selected peptide sequence	Abundance
1	1-R	ARLLIGTSPDFF	6/17
	1-C	TQTTVSSTQKPY	2/17
2B4	2B4-B	EDMSYWHEWNP	3/18
	2B4-C	DSSWYTWQPPTW	2/18
	2B4-A	HLQSMSSRWIEW	1/18
	2B4-E	EKVTSTWTDWQY	1/18
	2B4-Fb	TSKNTSWFDWQA	1/18
	2B4-G	TSFWADWQPQGR	1/18
	2B4-I	HSLSMYFDWVPH	1/18
	2B4-J	NSTWHDWNSWTH	1/18
	2B4-K	ATWYEWNPPWVS	1/18
	2B4-0	HLNSLYYDWTEP	1/18
	2B4-P	ESTWYMWQSRSE	1/18
	2B4-Q	HPMYPTWYSWQP	1/18
	2B4-R	HSIYSEWQPMWP	1/18
	2B4-1-5b	AHSSPRSVWYEW
	Consensus1	STWYD/EWQP
	2B4-N	SHNQLLRIQALD	1/18
	2B4-Xb	SHMNALLRQQAF
	Consensus2	NXLLRXQA
3B	3B-C	TTTSPYNDRQLL	5/20
	3B-F	NLGEKYNTLSLL	1/20
	3B-J	NPWYNDLSLLTA	1/20
	3B-M	SINLNYNSPSLL	1/20
	3B-S	GPLYQMNDALLL	1/20
	Consensus	YNDXSLL
5	5-A	RTLDSHWIRVGG	16/18
	5-D	NYSPLHNSDITV	1/18
	5-L	ALPLHGSPWKWY	1/18

^aAfter the third round of panning, phage DNA from ∼20 randomly selected plaques of each panning experiment was purified and sequenced. Selected peptide sequences are listed in order of their relative frequencies from the total number of clones successfully sequenced. Deduced consensus peptide motifs selected by rAbs 2B4 and 3B are also listed.

^bSequences are identical to peptides identified in a separate solid phase panning of the random peptide library with rAb 2B4. Phage expressing an identical peptide sequence to phage 2B4-F was found in 4/20 randomly sampled phage clones from the solid phase round 3 panning. Phage 2B4-1-5 and 2B4-X were not found in the solution phase panning but represented 15/20 and 1/20 phage clones, respectively, from the solid phase panning.

Peptides sharing core amino acid sequence homology were also selected by rAb 3B. Amino acids at conserved positions of its consensus peptide sequence, Y-N-D/S-X-X-L-L, were found in 19/20 sequences analyzed. The asparagine and two leucine residues were essentially invariant among the different phage clones. Table 3 gives a partial list of the 19 related peptides selected by rAb 3B. In contrast, the different peptide sequences selected by rAb 1 and rAb 5 did not identify a conserved amino acid motif. The most-abundant peptide sequences selected by rAb 1 and rAb 5 are also listed in Table 3 but were less intriguing because they did not reveal shared structural motifs.

Enriched phage-displayed peptides bind to their panning SSPE rAbs in ELISAs.

To demonstrate the reactivities and specificities of SSPE rAbs used in each panning experiment, selected phage-displayed peptides listed in Table 3 were assayed for binding to their panning SSPE rAbs, control preimmune human IgG (irrelevant negative-control antibody), and BSA blocking buffer alone. Unless indicated in the legend to Fig. 1, ELISA wells were coated with 0.5 μg of recombinant IgG or control IgG and incubated with 10¹⁰ phage particles per well. A second negative control to demonstrate the specificity of each rAb for peptides selected in the panning experiment was also included, and the binding of a randomly selected phage clone from the original M13 phage library to each rAb was measured. As shown in Fig. 1, one or more phage-displayed peptides from each panning experiment, but not randomly selected control peptide, bound to their respective panning rAbs. Phage 1-R, the most abundant phage population enriched by panning on SSPE rAb 1, showed increased binding to rAb 1 compared to that observed with control human IgG or BSA alone. Because phage 1-R displayed high background binding to both controls (Fig. 1A), specific binding to rAb 1 was evident only when phage input in the ELISA was increased to 5 × 10¹⁰ phage particles. Phage 1-C, however, showed no specific binding to rAb 1 at these same phage concentrations. Each of the four phage-displayed peptides (2B4-B, 2B4-C, 2B4-N, and 2B4-X) showed enhanced binding to rAb 2B-4 compared to that to control human IgG (Fig. 1B). Thus, phage-displayed peptides representing both structural motifs identified by sequence analysis (Table 3) bound strongly to rAb 2B-4. Correspondingly, phage-displayed peptides 3B-C and 3B-J and 5-A and 5-D demonstrated specific binding to their respective cognate rAbs (Fig. 1C and D). Some phage-displayed peptides showed additional binding to control IgG above the binding observed to BSA alone. This control polyclonal IgG was included in the ELISA to identify phage clones that might bind to IgG indiscriminately but also showed weak reactivity to MV. Thus, we cannot rule out the possibility that some of the phage binding observed with the control IgG was due to a low titer of anti-MV IgG antibody.

Peptide sequences selected by SSPE rAbs share homology with the MV N protein.

To identify regions of sequence homology to the MV N protein and other sequences in the NCBI database, a BLASTP search was conducted using the deduced consensus peptide recognition sequences identified for rAbs 2B4 and 3B. Searches of the entire nonredundant protein database and databases restricted to viral or mammalian protein sequences were conducted to identify short, nearly identical sequence matches. Because of the low sequence complexity of the target peptides (6 to 8 amino acids), high expect values were observed and many proteins sharing some homology to the consensus peptides were identified in most searches. To limit the number of potential protein targets aligning to the consensus peptide sequences, we omitted database hits that contained gaps or insertions in their alignment. We also eliminated alignments that had amino acid sequence differences at highly conserved residues (indicated in bold in Table 3) in the consensus peptide. Table 4 lists the proteins with the highest homologies to the consensus peptide sequences selected by rAbs 2B4 and 3B. The first and most dominant peptide sequence (ST/SWYD/EWQP) enriched by rAb 2B4 aligned to proteins from an array of different species. No homology to MV proteins was detected even when the database was restricted to viral sequences only. However, the second consensus motif recognized by rAb 2B4, NALLRIQA, produced several matches to the MV N protein. Furthermore, when the search was restricted to viral protein sequences in the nonredundant database, homology to the MV N protein was the only match detected. Alignment to a linear portion of the MV N protein was also found when searching with the conserved rAb 3B consensus binding peptide YNDRQLL. Although homology to two bacterial proteins was also noted, most database hits were to the N protein from either MV or canine distemper virus, a related paramyxovirus. Again, when the search was restricted to viral sequences, the only matches were to variants or different strains of the MV and canine distemper virus N proteins.

TABLE 4.

Alignment of peptide sequences enriched by SSPE brain-derived rAbs to sequences in the NCBI protein database^a

rAb	Protein	Organism	Homology	DBb
2B4 consensus 1			STWY(D/E)WQP
	Agarase	Microbulbifer sp. strain JAMB-A94	STWHDWQP	NR
	Cel48A precursor	Piromyces equi	WYDWQP	NR
	IMP dehydrogenase/GMP reductase	Desulfuromonas acetoxidans	WYDWQP	NR
	TRRAP-like protein	Caenorhabditis elegans	STWYDW	NR
	Hypothetical protein XP_579779	Rattus norvegicus	STWYDW	Mam
2B4 consensus 2			NALLRQQA
	Nucleocapsid protein	Measles virus	ALLRQQA	NR
		Measles virus	ALLRQQA	Virus
3B consensus			YNDRQLL
	Putative helicase	Bacteroides fragilis YCH46	YNDRQLL	NR
	Cytochrome c	Burkholderia pseudomallei	YNNRQLL	NR
	Nucleocapsid protein	Measles virus	YNDRNLL	NR
	Nucleocapsid protein	Canine distemper virus	YNDRELL	NR
	Nucleocapsid protein	Measles virus	YNDRNLL	Virus
	Nucleocapsid protein	Measles virus	YNDRHLL	Virus

^aConsensus peptide sequences (underlined in the Homology column) selected by SSPE brain-derived rAbs 2B4 and 3B were used to search the nonredundant NCBI protein database for short, nearly identical sequence matches. Proteins showing the highest homology and their alignments to the peptide sequences are listed.

^bDB, NCBI database used for the BLASTP search; NR, nonredundant; Mam, a database restricted to mammalian protein sequences; Virus, a database restricted to viral protein sequences.

Because of the limited number of ELISA-positive peptide sequences enriched by rAbs 1 and 5, a putative consensus recognition motif was not apparent. BLASTP searches of the protein databases using the entire 12-aa peptide sequences expressed by phages 1-R and 5-A did not identify the MV N protein among the 100 best alignments. BLAST searches using the entire 12-aa peptide sequences are significantly compromised by being constrained to peptide regions that may not be required for antibody recognition. Interestingly, a region of weak homology between the first 10 aa of peptide 1-R and a region of the MV N protein was found when the BLASTP search was restricted to only MV proteins (Fig. 2).

FIG. 2.

Alignment of peptide sequences enriched by SSPE rAbs to regions of the MV N protein. Regions of homology between the MV N protein and ELISA-positive peptides enriched by panning on rAbs 1, 2B4, and 3B lie in the C-terminal region of the MV N protein. Amino acids indicated in boldface represent identities between enriched peptide sequences and the N protein sequence. The N sequence shown is from the Edmonston strain of MV. Alignments of the rAb 2B4 and 3B peptides were obtained from BLASTP searches of the nonredundant NCBI protein database; the region of weaker homology (aa 465 to 474) to peptide 1-R was identified by limiting searches with this peptide to MV proteins only. The bracketed regions of the N protein identified as A, B, and C are products of N gene cDNA clones expressed as in-frame T7 capsid fusion proteins by the phage expression vector T7Select 3-b (Novagen, Madison, Wis.). These clones were obtained by screening an SSPE brain T7 cDNA library with SSPE brain IgG or rAb 2B4 prepared from SSPE brain (17).

Mapping of epitopes to the C terminus of the MV N protein.

Peptides enriched by rAbs 1, 3B, and 2B4 aligned to different regions of the MV N protein within the 60 C-terminal amino acids (Fig. 2). Two strategies were used to confirm the binding of each SSPE rAb to its putative epitope on the MV N protein. Peptides representing the suspected recognition site for rAb 2B4 were synthesized and assayed for binding by ELISA. rAb 2B4 exhibited concentration-dependent binding to peptide 2B4-1 (SAEALLRLQAGGGC[underlining indicates amino acid sequences of interest]),a peptide matching amino acids 491 to 500 of the MV N protein, whereas no binding was observed to an irrelevant control peptide (AAKLTIPAPQHTGGGC) (Fig. 3A). rAb 2B4 also bound a synthetic peptide of phage 2B4-F (TSKNTSWFDWQAGGGC) (Fig. 3B) which represents the most prominent structural motif enriched by rAb 2B4 (Table 3). Phage 2B4-F was selected by both solution phase and solid phase panning and does not share sequence homology with the MV N protein.

FIG. 3.

Binding of SSPE rAb 2B4 to synthetic peptides representing each structural motif selected by panning. Wells of Reacti-Bind maleimide-activated clear strip plates were coated with synthetic peptides (100 μl of a 5-μg/ml solution) and assayed for binding to rAb 2B4 as described in Materials and Methods. The synthetic peptides used were 2B4-1 (SAEALLRLQAGGGC),corresponding to aa 491 to 500 of the MV N protein and characteristic of the second binding motif enriched by panning on rAb-2B4; 2B4-F (TSKNTSWFDWQAGGGC),representing dominant motif 1, enriched by rAb-2B4 binding but with no sequence homology to the MV N protein; and the nonrelated control peptide 2-1-2 (AAKLTIPAPQHTGGGC).The underlined residues represent the amino acid sequences of interest. Shown is the concentration-dependent binding of rAb 2B4 to peptide 2B4-1 (diamonds) but not to the control peptide (circles) (A) and of rAb 2B4 to peptide 2B4-F (B). For the competitive ELISAs, rAb-2B4 was assayed for binding to peptides 2B4-1 and 2B4-F (C and D, respectively) at a concentration of 10 ng/ml, which is within the linear portion of the dose-response curves shown in panels A and B. rAb 2B4 was preincubated for 30 min at room temperature with the indicated numbers of antibody-specific phage particles (clones 2B4-B and 2B4-N) or irrelevant control phage and added to peptide-coated ELISA plate wells.

Binding to the synthetic peptides could be inhibited in a dose-dependent manner by preincubation of rAb 2B4 with ELISA-positive phage particles representing either structural motif (Table 3) selected in the panning experiments (Fig. 3C and D). The binding of rAb 2B4 to peptide 2B4-1 was competitively inhibited using both 2B4-B phage (motif 1) and 2B4-N phage (motif 2) (Fig. 3C). Similarly, the binding of rAb 2B4 to synthetic peptide 2B4-F was inhibited by phage 2B4-B and 2B4-N (Fig. 3D). In each assay, inhibition was more efficient using phage 2B4-B than 2B4-N, most likely reflecting the higher affinity of dominant motif 1 for rAb 2B4. In both experiments, near-complete inhibition of binding was obtained using ∼1 × 10¹⁰ phage particles. Assuming that each phage particle displays five peptides, the estimated peptide concentration of 1 × 10¹⁰ phage particles was ∼1 nM.

We also screened a T7 phage-displayed cDNA library generated from SSPE brain with rAb 2B4 and IgG eluted from SSPE brain to obtain a panel of T7Select phage clones (Fig. 2) expressing portions of the MV N gene in the translational reading frame of T7 capsid fusion proteins (2). An ELISA of purified phage from each SSPE T7 cDNA clone and a negative-control phage randomly selected from the T7 library confirmed the recognition sites of rAbs 2B4 and 3B to be in regions of the MV N protein predicted from the random-peptide-library panning. As shown in Fig. 4, only clone A and clone B phage bound strongly to rAb 2B4. The differential binding of rAb 3B to T7 clone A phage but not to T7 clone B confirmed the binding site of rAb 3B to be in the predicted C-terminal region of the MV N protein. Because clone B terminates within the putative binding site for rAb 3B, the absence of the final 3 C-terminal amino acids (LLD) of the MV N protein was sufficient to completely abrogate antibody binding. None of the SSPE T7 phage bound to rAb 1 in an ELISA (Fig. 4). Therefore, the putative binding region for rAb 1 highlighted in Fig. 2 could not be confirmed from these experiments. We also did not detect any binding of SSPE T7 clones to rAb 5. Identical binding profiles were obtained when purified T7 phage proteins were separated by SDS-PAGE and assayed for binding to individual SSPE rAbs by protein immunoblotting (data not shown).

FIG. 4.

Binding of the products of MV N gene cDNA clones to SSPE brain-derived rAbs. T7 clones A, B, and C express regions of the MV N protein (Fig. 2) as T7 capsid fusion proteins. In this ELISA, wells of an RIA plate were coated with 100 μl of rAbs 1, 2B4, and 3B (5 μg/ml in TBS) or BSA only and purified T7 phage particles (2 μg protein or ∼10⁹ phage particles) of MV N clones A, B, and C and the negative-control T7 cDNA clone product (Rand) were added to each well. Binding with HRP-conjugated mouse anti-T7 capsid protein IgG was assessed.

DISCUSSION

The combination of laser microdissection and sensitive PCR technologies now enables analysis of gene expression in cells from diseased human tissue at the single-cell level. This approach is particularly useful in recapitulating in vitro the intrathecal antibody response in human CNS inflammatory diseases. Our previous sequence analysis of H- and L-chain V regions of plasma cells microdissected from SSPE brain revealed features of a targeted antigen-driven response, including clonal expansion, somatic mutation, intraclonal variation, and receptor editing (3). Furthermore, we demonstrated that the antibody response in SSPE brain is directed largely against MV, particularly the N protein, as evidenced by the specificity of five of eight rAbs generated from SSPE brain plasma cell clones. Indeed, reactivity against the MV N protein is the dominant response after natural MV infection (13) and in the CSF of SSPE patients (9, 15). In this study, we began to map the epitopes recognized by these dominant plasma cell clones. There are several distinct advantages to using rAbs derived from B cells that have migrated and differentiated within the CNS compartment, including the feasibility of analyzing immune responses occurring at the actual site of disease without dilution or confounding non-disease-related antibody specificities found in serum and to a lesser extent in CSF.

Our present study demonstrates the utility of random peptide library screening for epitope and antigen identification. Based on a sequence comparison of the selected peptide sequences (Table 3), phage-peptide ELISAs, and the binding of each panning rAb to phage-displayed cDNA sequences representing different portions of the MV N protein (Fig. 4), the rAbs tested in this study recognize different epitopes on the MV N protein. Each of the SSPE brain-derived rAbs examined enriched a distinct set of specific immunoreactive peptide sequences. Most striking was the diverse array of peptides selected by rAbs 2B4 and 3B. Panning on each of these rAbs yielded a large number of different but structurally similar peptide sequences (Table 3), allowing us to deduce for each rAb short putative consensus binding motifs that share sequence homology to antibody epitopes on the MV N protein. Two very different motifs at the amino acid sequence level were selected by rAb 2B4. Peptides similar to the first consensus motif (STWYD/EWQP) mimicked the 2B4 epitope and accounted for 16/18 M13 phage clones analyzed. Sequences identified by immunopanning of peptides that cross-react with the panning antibody but do not share sequence homology (peptide mimics) with the true target antigen are commonly found in random peptide pannings (23). Although there was no sequence homology between this peptide mimic and the MV N protein, phage expressing this motif did inhibit the binding of rAb 2B4 to a synthetic peptide of the 2B4-N protein epitope (Fig. 3C). The predominance of phage expressing this motif probably reflects a higher affinity for rAb 2B4 than phage peptide sequences homologous to the true 2B4 epitope as suggested by inhibition assays (Fig. 3). Interestingly, the second 2B4 consensus peptide, NXLLRXQA, which was used to identify the 2B4 epitope, accounted for only 1/18 peptides found after the third round of panning. The reactivity of phage particles expressing this peptide was assayed only because a similar peptide sequence was detected in a previous panning experiment with rAb 2B4 (Table 3). This observation underscores the importance of assaying even minor phage populations for immunoreactivity and of performing multiple panning experiments. While we confirmed the epitopes recognized by rAbs 2B4 and 3B by demonstrating the binding of rAbs either to synthetic peptides (Fig. 3) or to short MV N fusion proteins (Fig. 4), the putative epitope for rAb 1 could not be confirmed. A likely explanation lies in the sequence variation among different MV strains found in the 125 C-terminal aa of the MV N protein (14). The T7 cDNA clones used in the ELISAs whose results are shown in Fig. 4 are from a different SSPE brain than the brain-derived SSPE rAbs and differ from the Edmonston strain at 11 amino acid positions within this region (data not shown). Furthermore, the putative rAb 1 N protein epitope (AHLSTDTPLD) expressed by T7 clones A and B shares only weak homology (indicated in bold) to peptide 1-R (ARLLIGTSPD), while the sequence of this region in N proteins (AHLPIGTSLD) from other SSPE brains is more homologous to peptide 1-R (17). The homology of the N proteins from these strains identified this region of the N protein as the probable rAb 1 epitope. The N protein sequence and MV strain from the SSPE brain that was the source of the rAbs is not known.

Several factors contributed to the successful identification of MV epitopes in this study and could be useful in predicting which panning experiments might lead to antigen identification. We knew that the SSPE brain-derived rAbs were most likely directed against linear sequences of the N protein, since they all reacted in immunoblots following separation by SDS-PAGE, and that rAbs 2B4 and 3B in particular appeared to have high affinity as judged by their immunoreactivity in several different binding assays. Both the 2B4 and 3B pannings showed large increases in the numbers of recovered phage after three rounds of panning (Table 2), indicating that phage enrichment is a reliable marker for the selection of reactive high-affinity peptides. It was somewhat surprising that none of the peptides enriched by rAb 5 aligned to the MV N protein, despite the significant increases in round 3 phage titers. This could be due to the enrichment of mimotopes by rAb 5 or to inherent difficulties associated with BLASTP peptide searches. The alignment of peptides selected by rAb 2B4 and rAB 3B to the MV N protein was facilitated by identifying short linear binding motifs that simplified the database search and eliminated alignments to regions of enriched peptide sequences that do not contribute to the antibody binding site. Thus, mutational or positional peptide scanning analysis to identify those amino acids contributing to antigen recognition should be performed when a binding motif is not readily discernible.

Based on our results, panning of random peptide libraries with CNS-derived rAbs could potentially be applied to antigen identification in other inflammatory CNS diseases in which the target of the intrathecal Ig response is unknown, such as multiple sclerosis, sarcoidosis, and Behcet's syndrome. Because successful antigen identification is dependent on many factors, including the binding affinity, the nature of the epitope, and the presence of the target antigen sequence or a closely related peptide sequence in the protein database, this approach should be used in conjunction with other screening methods. Antigen identification and database searches can also be confounded by the selection of peptide mimotopes. Identifying the antigenic specificities of MS oligoclonal IgG has been problematic, and the use of random peptide screening as a tool for antigen identification in MS is not novel. Panning of libraries with MS CSF in these earlier studies enriched several peptide sequences (7, 8). However, none of the selected peptides reacted universally and specifically with MS CSF and reactivity was also found in serum from a subset of MS patients and healthy controls. An advantage of using CSF-derived rAbs rather than whole CSF for panning experiments is that individual antibodies known to be intrathecally synthesized can now be examined for specificity.

In our current study, the peptides selected by two SSPE brain-derived rAbs (2B4 and 3B) when viewed together would have identified MV as a likely target of the intrathecal Ig response. Because both rAbs 2B4 and 3B recognize different linear protein epitopes within the MV N protein, the utility of rAbs recognizing discontinuous epitopes for identification of unknown antigens is still uncertain. The screening of libraries containing longer peptide sequences or that are constrained by cysteine residues may aid in the identification of more-complex, discontinuous epitopes (4, 9). It is possible that many of the rAbs derived from MS and other inflammatory disease CSF samples will recognize discontinuous epitopes and perhaps even carbohydrates. Nevertheless, a large panel of CNS-derived rAbs will likely contain antibodies of various affinities and specificities, some of which will recognize linear sequences and be amenable to random peptide biopanning as a means of identifying novel candidate antigens for further evaluation.