The presence of increased IgG in the brains of humans with infectious and inflammatory CNS diseases of unknown etiology such as multiple sclerosis may be a clue to the cause of disease. For example, the intrathecally synthesized oligoclonal bands (OGBs) in diseases such as subacute sclerosing panencephalitis (SSPE) or cryptococcal meningitis have been shown to represent Ab directed against the causative agents, measles virus (MV) or Cryptococcus neoformans, respectively. Using SSPE as a model system, we have developed a PCR-based strategy to analyze the repertoire of IgG V region sequences expressed in SSPE brain. We observed abnormal expression of germline V segments, overrepresentation of particular sequences that correspond to the oligoclonal bands, and substantial somatic mutation of most clones from the germline, which, taken together, constitute features of Ag-driven selection in the IgG response. Using the most abundant or most highly mutated γ H chain and κ or lambda L chain sequences in various combinations, we constructed functional Abs in IgG mammalian expression vectors. Three Abs specifically stained MV-infected cells. One Ab also stained cells transfected with the MV nucleoprotein, and a second Ab stained cells transfected with the MV-fusion protein. This technique demonstrates that functional Abs produced from putative disease-relevant IgG sequences can be used to recognize their corresponding Ags.
The increased IgG and oligoclonal bands (OGBs)3 found in the brains of humans with infectious CNS diseases have been shown to be Ab directed against the causative agent. For example, OGBs in the cerebrospinal fluid (CSF) of patients with disorders such as cryptococcal meningitis, mumps meningitis, progressive rubella panencephalitis, herpes simplex virus encephalitis, HTLV-1 myelopathy, subacute sclerosing panencephalitis (SSPE), and Lyme disease are directed against the respective fungus, virus, or bacterium (1, 2, 3, 4, 5, 6, 7). In SSPE, a chronic progressive measles virus (MV) infection of brain (8), ∼95% of the CNS IgG is synthesized locally (9), and 25–75% of the Ab in the CNS is directed against MV (6, 10). OGBs are also present in the brains of several inflammatory CNS diseases of unknown cause, such as multiple sclerosis and CNS sarcoidosis. Although OGBs are found in 88–100% of CSF from multiple sclerosis (MS) patients, their corresponding Ags are unknown (11, 12).
Herein, we describe a strategy to examine the IgG repertoire expressed in the CNS of a patient with chronic progressive encephalitis, and to construct recombinant Abs to study Ab reactivity based on that repertoire. Our previous analysis of the IgG response in MS brain demonstrated restricted expression of particular IgG sequences, clonal expansion, and somatic hypermutation from germline V segments, features consistent with an Ag-driven response (13). With available IgG expression vectors, it is possible to construct recombinant Abs from putative disease-relevant sequences and to characterize their corresponding Ags. Thus, we performed a sequence analysis of the IgG repertoire in an SSPE brain as a model system to determine whether recombinant Abs constructed from candidate sequences could identify MV-specific Ags.
Materials and Methods
Pathologically verified SSPE brain was frozen at −70°C within 12 h after death. A XhoI/oligo(dT)-primed cDNA library was constructed from multiple areas of gray and white matter and directionally cloned into the EcoRI/XhoI sites of the Lambda-ZAP vector (Stratagene, La Jolla, CA) as described (13, 14). The complexity of the library was 7.0 × 106, with an average insert size of 1.5 kb.
PCR amplification of IgG V regions and repertoire cloning
Representative IgG sequences were obtained using a nested PCR strategy to amplify the IgG H chain (γ) and L chain (κ or λ) V regions from Lambda Zap DNA purified from the SSPE cDNA library as described (14). Briefly, the first H chain PCR amplified the region between the 5′ polylinker site in the vector adjacent to the cDNA and a 3′ primer (CH1) complementary to a site in the first C region of all four human IgG isotypes. In the nested PCR amplification, the 5′ primer was still located in the vector polylinker upstream of the cDNA insert, and the 3′ primer (CHJ) was complementary to the 3′ region of all six J segments and the 5′ portion of the first C region genes. The PCR amplifications of V region sequences from IgG L chains were identical in reaction conditions to the H chain amplifications, but with 3′ primers matching equivalent sequences in κ or λ L chains. PCR products were electrophoresed on 1% agarose gels, and the nested products (450–500 bp) were excised and cloned into the TA cloning vector (Invitrogen, Carlsbad, CA) as described (14). After plating, well-separated white colonies were randomly picked and grown in liquid culture for plasmid DNA preparation.
Sequencing and analysis of clones
Plasmids containing inserts were sequenced by the dideoxy method (15) with Sequenase 2.0 (U.S. Biochemical Corp., Cleveland, OH) and primers adjacent to the insert (T7 or T3). Sequences were analyzed with the PCGene software package (Intelligenetics, Campbell, CA) and aligned to the closest VH, VK, and VL germline segments in the VBASE database as described (13). Group analyses at significant levels were calculated at 95% confidence intervals.
Recombinant Ab construction and purification
H chain and L chain V regions were chosen from the sequence analysis for expression as recombinant Abs in the mammalian expression vectors pNG1.16 and pGK.11 (16). These cassette vectors were developed to express complete Ig H and L chains with the subcloned V regions of choice. Selected SSPE V region sequences in the TA vector were individually PCR amplified as described (16), with slight modifications. The sense and antisense primers were modified to match exactly each template V region clone and were of a length to provide the entire V region for subcloning after a single PCR amplification. The 5′ primers contained restriction sites for subcloning, splice recognition signals for intron excision, a polypyrimidine tract, the 3′ exon of an Ig leader sequence, and regions encoding the framework 1 regions of the Ig template clones. The 3′ primers contained regions complementary to the 3′ ends of the respective V region templates, splice donor sequences, and restriction sites for subcloning (16). PCR amplification conditions included 10–20 ng of template, 15 pmol of each primer, 4 mM MgCl2, and 2.0–2.5 U of Taq polymerase (Perkin-Elmer Cetus, Emeryville, CA). Reaction parameters were 34 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 1 min (10 min in last cycle). PCR products were electrophoresed on 1% agarose gels, excised, purified, subcloned into the TA vector, and resequenced to insure fidelity and the correct reading frames. The verified H chain V region DNA was digested with Asp718I/XhoI, excised from agarose gels and subcloned into the pNG1.16 H chain vector. The κ or λ V region DNA was similarly digested with NruI/XhoI and subcloned into those sites of the pGK.11 vector.
The γ H chain constructs were cotransfected with each of the κ or λ L chain constructs by electroporation into Ag8.653 myeloma cells as described (17) and diluted in 96-well microtiter plates at densities between 103 and 2 × 104 cells per well. After 14–17 days of coselection with G418 (0.6 mg/ml) and mycophenolic acid, (Life Technologies, Gaithersburg, MD) (17), culture supernatants from wells containing 1–2 colonies were tested for the presence of IgG H and L chains by slot blot analysis (18). Wells producing maximal amounts of both IgG chains were transferred to flasks and grown to high density with coselection in 100 ml IMDM, 10% FCS (Life Technologies). IgG was purified from the culture supernatants by protein A chromatography, with yields of 100–300 μg per Ab.
Abs were tested for reactivity against MV by immunostaining uninfected Vero cells or cells previously infected with the Edmonston strain MV, grown on glass coverslips to near confluence, and fixed in acetone for 1 min. Immunocytochemistry was performed as described (19).
Each Ab was also used to stain Vero cells transiently transfected with expression constructs encoding each of six MV proteins (phosphoprotein, hemagglutinin, nucleoprotein, fusion, matrix, and polymerase). Vero cells were grown to 50% confluence on glass coverslips in 60-mm dishes with DMEM/10% FCS and then transfected with 1 μg MV expression construct DNA using Lipofectin, according to the manufacturer’s recommendations (Life Technologies). After 48 h, coverslips were fixed in acetone for 1 min and used for immunocytochemistry as described (19).
Immunoprecipitation and immunoblotting
MV-infected Vero cell lysates were immunoblotted and immunoprecipitated with recombinant Abs as described (19).
The repertoire of IgG expressed in the SSPE brain was analyzed for sequence variability by randomly selecting and sequencing IgG variable domains after PCR amplification from an SSPE cDNA library. The cDNA library was created from multiple regions of white and gray matter from a single SSPE brain (13). The sequences were PCR amplified from a cDNA library to ensure that the IgG repertoire analyzed was produced by resident B cells in the brain and not by transient IgG protein trafficking through the brain. Nested PCR amplification from the cDNA library also enabled the use of universal vector-specific 5′ primers and avoided the need for a panel of Ig family-specific primers at the 5′ end (20) that may preferentially amplify specific sequences or fail to amplify sequences that have mutated at the 5′ primer binding sites.
Particular IgG H and L chain sequences encountered multiple times in our sequence analysis (Fig. 1) probably indicate overrepresented IgGs in the brain repertoire. This method of repertoire analysis yielded similar results in a study of the B cell response in an MS brain but did not reveal groups of sequences when adult human blood, an abundant source of heterogeneous sequences, was analyzed (13). In the SSPE H chain sequence analysis, 23 distinct V regions are identified by their CDR 3 sequences in Fig. 1 and are grouped by their relative abundance in the 52 clones analyzed. Sequence variability was not generally observed within members of each group, except occasionally for shorter 5′ ends that demonstrated the different endpoints of reverse transcriptase during the cDNA synthesis of those clones. Nucleotide differences were noted between members of one group at specific sites, but these differences occurred in less than 1% of the entire sequences. To determine the germline origin of these V region segments, all of the complete sequences were aligned with the V BASE database of human IgG germline VH segments (21). The SSPE IgG sequences were only 83–97% identical to their closest germline V segments, a greater difference than the usual 1–2% observed for IgG allelic polymorphisms in humans (21). In contrast to the prevalence of VH3 family V segments in normal humans (21, 22), the VH1 germline family was found more often than other families in the SSPE H chain analysis (48%). The two germline VH segments DP79 and DP10 are represented by 10 of the 23 sequences analyzed in this SSPE brain. To predict which IgG sequences were overabundant in the brain, at least two of the H chain sequences analyzed (No. 1 and 2) were overrepresented at significant levels (19 ± 11%, 17 ± 10%) and may represent several of the OGBs observed in this SSPE brain. Most of the H chain sequences appeared once in the analysis. As each sequence was aligned to its most similar germline VH segment, the nucleotide differences in different regions of the clones (the framework or CDR regions 1 and 2) were defined as replacement or silent mutations, based on their effect on the translated IgG sequence. These accumulated differences were included in Fig. 1 as ratios of replacement to silent mutations for each sequence, and for several sequences demonstrated an extraordinarily high proportion of replacement mutations in the CDR regions 1 and 2. For example, the group 2 sequence had 10 changes in the CDR 1 and 2 regions differing from the germline V segment, all of which contribute to replacement mutations in the translated IgG. The group 14 sequence had 11 changes in the CDR 1 and 2 regions, all representing amino acid changes in the translated IgG sequence. This accumulation of replacement mutations in the CDR regions is characteristic of affinity maturation in the humoral immune response (23, 24).
The Ig L chains expressed in the SSPE brain were also analyzed. The κ and λ L chains were PCR amplified and sequenced in a fashion similar to the H chain analysis. Although L chains derived from IgG- or IgM-producing B cells would be indistinguishable from each other by this PCR amplification protocol, the IgG greatly exceeds the amount of IgM in the SSPE inflammatory CNS response (10, 25) and should account for the majority of L chains in this analysis. The κ chains revealed 2–3 overrepresented sequences and an overrepresentation of the VK3 germline family. Although 11 distinct sequences were encountered, κ groups 1 and 2 were the most abundant and constituted 50% of all the clones sequenced. The κ sequences were 91–97% identical to their closest germline V segments, also suggesting somatic mutation from the germline segments, rather than polymorphisms (26). But the κ V region sequences were less mutated from their respective κ germline segments than the γ H chain sequences above. Like the γ V regions, many κ sequences showed higher proportions of replacement mutations in the CDR regions 1 and 2 compared with the framework regions. The λ L chain sequence analysis was more cursory due to an incomplete published inventory of the germline VL segments but also demonstrated a clear overabundance of at least one λ V region sequence at significant levels (group 1, 35 ± 8%).
From the IgG sequence analyses of SSPE, several H and L chain sequences that were either overrepresented or contained unusually high numbers of CDR replacement mutations were chosen to produce intact recombinant Abs for functional studies (Fig. 2). IgG expression vectors were used to express each inserted V region in the context of a functional Ig promoter and enhancer and joined to the appropriate C regions to assemble a functional IgG after cotransfection into myeloma cells. Each of the V regions of γ groups 1, 2, and 14 (Fig. 1, and denoted G1, G2, and G3 in Fig. 2) was PCR amplified and cloned into the pNG1.16 expression vector. The κ V region groups 1 and 2, or the λ group 1 V region (Fig. 1 and denoted K1, K2, and L1 in Fig. 2) were PCR amplified and cloned into the pGk.11 expression vector. The H and L chain expression constructs were cotransfected in every combination by electroporation into the mouse myeloma cell line P3 × 63Ag8.653 (Fig. 2). For example, the γ 1 V region sequence was coexpressed in separate experiments with the κ 1, κ 2, or λ 1 L chain V region, to be tested as IgG Ab C1, C2, or C5.
We tested each of the purified recombinant Abs for activity by immunostaining MV-infected Vero cells. As a positive control, Ab eluted from SSPE brain stained MV-infected cells (Fig. 3,J), but not uninfected cells (Fig. 3,K). Constructs C1, C4, and C7 specifically stained MV-infected cells (Fig. 3, A, D, and G), but not uninfected cells (Fig. 3, B, E, and H), or cells infected with the unrelated varicella zoster virus (VZV) (Fig. 3, C, F, and I). Although the group 1 γ H chain expressed with κ 1 (Ab C1) stained MV-infected cells, the same γ 1 did not stain cells when paired with either the κ 2 or λ 1 L chains (C2 or C5, data not shown). Similarly, the γ 2 H chain specifically stained MV-infected cells when paired with the κ 2 L chain (C4; Fig. 3,D), but not when paired with the κ 1 or λ 1 L chains (C3 or C6; Fig. 3,L, and data not shown). The γ 14 V region sequence, which did not appear to be overrepresented in this SSPE analysis, but which contained a large number of replacement mutations in the CDR regions, also stained MV-infected cells when paired with the K1 L chain (C7; Fig. 3,G), but not with the K2 or L1 L chain (C8 or C9; data not shown). None of the Abs stained uninfected Vero cells or cells infected with VZV (Fig. 3, C, F, and I), and none of the other pairs of H/L chains in IgG constructs stained infected or uninfected cells (data not shown). These results suggested that each γ H chain required some L chain specificity to demonstrate activity.
To establish the Ag specificity of the recombinant Abs more precisely, each Ab was used as a probe to recognize specific MV components. The C1, C4, and C7 Abs were used to immunoprecipitate MV-infected Vero cell lysates but did not bind specific Ags. Abs eluted from the SSPE brain did precipitate specific MV proteins. In addition, the recombinant Abs did not recognize specific proteins in immunoblots of MV-infected cell lysates (data not shown).
Each Ab was also used to stain cells transfected with each of six specific MV proteins. Abs eluted from the SSPE brain stained the phosphoprotein, hemagglutinin, nucleoprotein, and fusion proteins transfected into cells (Fig. 4, A-D), as well as the polymerase protein (data not shown). SSPE Abs did not stain the MV matrix protein transfected into cells, which may reflect the lack of Abs to the M protein in SSPE brain (27). Recombinant Ab C4 stained only cells transiently transfected with the MV nucleoprotein (Fig. 4,G), but not cells expressing the phosphoprotein (Fig. 4,E), hemagglutinin (Fig. 4,F), or fusion protein (Fig. 4,H). Ab C7 stained only cells transfected with the MV fusion protein (Fig. 4,L), but not cells transfected with phosphoprotein, hemagglutinin, or nucleoprotein (Fig. 4, I-K). Ab C1 did not stain specific MV proteins transfected into cells. In addition, each recombinant Ab was negative on cells transfected with the MV polymerase and matrix proteins (data not shown).
To develop optimal techniques to study specific Ab reactivity in infectious and inflammatory CNS diseases, we investigated the IgG response in SSPE brain. Our analysis of the IgG sequences expressed in SSPE brain revealed a heterogeneity of IgG clones. Specific clones were mutated from their closest germline V segment and accumulated replacement mutations in their CDR regions, features characteristic of Ag-driven selection and affinity maturation in the B cell response. Based on sequence analysis, we constructed recombinant Abs from several of the abundant or highly mutated H and L chain sequences and demonstrated their reactivity to disease-relevant Ags. Each H chain that we tested reacted with MV (the cause of SSPE), but only when combined with particular L chains.
We used nested PCR to obtain a representative sampling of the γ H chain and κ or λ L chain V regions from an SSPE brain cDNA library. This strategy has been successfully used to examine the IgG repertoire expressed in an acute MS brain and in normal human PBLs (13). Specific VH, VK, and VL sequences were overrepresented in our SSPE brain. Many sequences were encountered only once. The number of clones that we analyzed allows a statistically significant assignment as a group to only the largest groups (groups γ 1 and 2, κ 1 and 2, and λ 1). The occurrence of multiple IgG sequence groups from the SSPE brain is consistent with the presence of OGBs in SSPE and was also found in a similar analysis of IgG synthesized in MS brain (13). Analysis of larger numbers of IgG clones may reveal additional sequence groups corresponding to OGBs.
Each of the VH region sequences differed in homology from its closest germline V segment, ranging from 83–97% identity. These differences are greater than the usual 1–2% polymorphism observed for IgG V segment alleles in humans (21). They are not likely to be Taq polymerase errors during PCR amplification, since, for several selected clones, additional PCR amplification in preparation for recombinant Ab construction showed no additional nucleotide changes introduced into the clones. Rather, the differences seen in VH sequences more likely constitute somatic mutation of various IgGs expected in the prolonged B cell response to MV in SSPE brain.
The OGBs in SSPE brains are either κ or λ L chain predominant (28). Thus, it was important to amplify and characterize the contribution of sequences from both L chain isotypes, in reconstituting the functional recombinant Abs. Specific κ and λ L chains were overrepresented in the SSPE brain, as has been found in SSPE CSF (28). Like the H chain analyses, the VK sequences also differed from their closest germline V segment genes at levels (91–97% identity) indicating somatic mutation and affinity maturation in the L chain V regions. In striking contrast to VH sequences, the VK sequences mutated to a lesser degree from their germline V segments. If, as we expect, these L chain sequences were associated in vivo with the SSPE H chain sequences, then the apparently different mutation rates must be reconciled. It is possible that the less mutated L chain sequences are not associated with IgG H chains, but instead are expressed in IgM-producing B cells, or are expressed as free L chains. Indeed, other forms of L chains could not be distinguished from IgG-derived L chains by our PCR amplification and sequence analyses, and free L chains have been reported in SSPE brain (28). On the other hand, the predominant Ab in CSF of infectious CNS diseases is IgG and should contribute most L chain sequences (10, 25). The lower mutation rate of IgG κ V regions from their germline V segments agrees with previous findings (29). Several functional secondary hybridoma Abs against influenza virus hemagglutinin have been sequenced in which the VK L chain segments were mutated at rates 10- to 15-fold lower than the adjoining VH segments (30, 31). Thus, the most abundant VH, VK, and VL sequences analyzed here may be coexpressed in SSPE OGBs. Verification of these H/L chain pairings in vivo in the SSPE brain would require purification of sufficient amounts of specific OGB proteins to sequence the separate IgG chains.
The purpose of analyzing the IgG repertoire in SSPE brain and its utility in examining other inflammatory CNS diseases of unknown origin are the subsequent use of the Ab sequences to identify their corresponding Ags. From this SSPE analysis, the two most abundant H chain sequences and the group 14 sequence, containing a high proportion of replacement mutations, were expressed as Abs. All three of the γ sequences reacted against MV when coexpressed with particular L chains, suggesting that our selection criteria for H chain sequences were valid for determining disease-relevant Abs, or that the majority of sequences in SSPE brain are MV specific, or both. The most abundant IgG sequences and the most highly mutated sequence were IgGs directed against MV, the cause of SSPE. Furthermore, only one L chain conferred reactivity for each of the H chains; when two different L chains were combined with each H chain, specific immunostaining was lost. In addition, γ 1 and 2 sequences both specifically stained MV-infected cells, but only when combined with different L chains. Thus, for each functional Ab, the H chain has a relatively specific requirement for its coexpressed L chain, and the functional combinations in these recombinant Abs are likely to mimic the in vivo combinations originally encountered in brain. Note that the third VH sequence (γ 14) also stained MV-infected cells when combined with only one of the L chains tested, but worked with the same L chain (K1) as worked for the γ 1 sequence. Since both γ 1 and γ 14 were not likely to have been coexpressed with the same L chain in the brain, functional Ab reactivity must be permissive to some degree of L chain substitution. Although L chain usage has been shown to be critical in defining epitope specificity (32), the ability of an Ab’s H chain to be shuffled with several L chains and retain Ag specificity has also been noted for Ab reactivities to haptens, DNA, and cardiolipin (33, 34). The ability to shuffle L chains enhances the feasibility of recombinant Ab experiments if the pairing of H and L chains need not duplicate the exact pairing originally found in vivo. Since IgG expression is represented as OGBs in SSPE as well as in other infectious and inflammatory CNS diseases, these more abundant sequences should help guide the initial pairings of H and L chains in functional assays.
The SSPE recombinant Abs identified MV Ags. Ab C1 did not stain any of the transiently transfected MV proteins tested and may be directed against an untested MV protein or antigenic determinants that are altered or not expressed in these transfected cells. Ab C4 stained cells transfected with and expressing the MV nucleoprotein, but not cells transfected with the phosphoprotein, hemagglutinin, or fusion protein. A second recombinant Ab, C7, stained cells transfected with the MV fusion protein, but not cells transfected with the phosphoprotein, hemagglutinin, or nucleoprotein. C4 and C7 appear to reflect the predominant MV reactivity by IgG in SSPE brain (35, 36). Although both MV nucleocapsid and fusion protein can be detected in SSPE brains, the MV matrix protein as well as Ab to the matrix protein is greatly reduced or absent (37). The first Ab responses we detected with our most abundant or highly mutated sequences were also to the MV N and F proteins. This suggests that our strategy to characterize the IgGs most actively expressed in the SSPE brain reflects the characteristic immune response previously observed in SSPE and that similar Abs in other inflammatory CNS diseases may also be those involved in progression of disease.
C4 also mimics a N protein reactivity we obtained earlier by immunopanning a broad phage-displayed repertoire of SSPE Fabs on MV-infected cells, and which enriched disease-relevant Fabs that were subsequently used to identify their corresponding Ags (19). We previously isolated four Fabs from the same SSPE brain that reacted specifically with the MV P or N protein (19), but none of those Fab sequences appeared in our present survey of IgG sequences. Possibly those Fabs were derived from sequences in brain not observed in this limited repertoire analysis but were nevertheless high affinity Fabs that enriched by panning. The Fab immunopanning technique is more useful in quickly selecting high affinity Fabs from a large population of randomly combined H and L chain sequences (38, 39), and in purifying their corresponding Ags. The SSPE Fabs derived from panning immunoprecipitated their Ags, whereas the Abs generated here from the most abundant or highly mutated sequences were not of sufficiently high affinity to do so. Once the putative disease-relevant Abs are identified, the techniques described here will be useful to rapidly detect the Ags in other brains by immunostaining. The ability to construct bivalent recombinant IgG from Fab-derived sequences may also improve their binding to Ags, thus providing better reagents for the identification of those Ags.
The techniques presented here can be applied directly to other inflammatory CNS diseases of unknown etiology, such as MS. Overall, no fewer than 16 agents and viruses have been associated with MS, but none as yet has been tightly linked to the disease (40). The presence of other Ab reactivities in MS brain may obscure the search for a causative agent and would require studying more MS patients and correlating the disease course with brain-derived Ab reactivity.
SSPE brain was kindly provided by the National Neurological Research Specimen Bank, VAMC, Los Angeles, CA. We thank Jo Moore for excellent technical assistance, and Marina Hoffman for editorial review. The mammalian transfection vectors pNg1.16 and pGk.11 were generously provided by Dr. Linda Harris of Bristol-Myers Squibb, Seattle, WA. The expression constructs encoding the MV proteins were the generous gifts of Drs. S. Schneider-Schaulies and V. ter Meulen, Wurzburg, Germany.
This study was supported in part by Public Health Service Grant NS32623 and by Pilot Research Award PP0586 from the National Multiple Sclerosis Society to M.P.B.
Abbreviations used in this paper: OGB, oligoclonal band; CSF, cerebrospinal fluid; SSPE, subacute sclerosing panencephalitis; MV, measles virus; MS, multiple sclerosis; HTLV, human T cell leukemia virus; VZV, varicella zoster virus.