Isolated Human Germinal Center Centroblasts Have an Intact Mismatch Repair System¹

The humoral immune system of higher vertebrates has evolved to produce high-affinity Abs to an enormous variety of Ags. The Ab repertoire generated by the initial recombination of random germline gene segments is not efficient in protecting an organism because of the relatively low affinities of the resultant Abs (1). The affinity maturation of T cell-dependent Ab responses results from the accumulation of point mutations in the V region of rearranged Ig genes followed by Ag-driven selection of the B lymphocytes expressing high-affinity Abs (reviewed in 2 . Hypermutation in human tonsils occurs in centroblasts in the dark zone of the germinal center (GC)⁴ (3, 4). Several salient features of Ig somatic hypermutation have been reported: 1) a high rate of point mutations, 2) confinement of the mutations to the rearranged V genes and the immediate flanking sequences, 3) the presence of intrinsic hot spots, 4) DNA strand preference, 5) transcriptional dependence, and 6) a requirement for intronic and 3′ enhancer sequences (reviewed in Refs. 5–7).

Despite intensive study, the molecular mechanisms of Ig hypermutation remain unclear. One possibility is that hypermutation is a result of impaired DNA repair. Among the known DNA repair systems, the DNA mismatch repair (MMR) system deserves thorough examination for its involvement in Ig somatic hypermutation for the following reasons: First, point mutations are characteristics of somatic hypermutation in Ig V genes (reviewed in 8 . Second, unlike nucleotide excision repair (NER), which specifically removes bulky DNA adducts generated by certain types of DNA damage, the MMR system corrects point mutations introduced during DNA synthesis (reviewed in 9 . Experiments by two research groups excluded the involvement of transcription-coupled NER in Ig somatic hypermutation (10, 11). Third, inactivation of the genes encoding MMR activities results in a large increase in the spontaneous mutation rate and predisposition to tumor development (reviewed in 9 . Fourth, the connection found between MMR and transcription-coupled NER (12, 13) suggests a potential role of MMR in transcription-linked Ig somatic hypermutation. Finally, Cascalho et al. (14) showed recently that PMS2, a component of the MMR system, is involved in Ig somatic hypermutation. Interestingly, they showed that PMS2 may actually enhance hypermutation, suggesting that the MMR system may be altered in some way in hypermutating B cells such that it enhances rather than corrects somatic mutations at Ig loci.

Despite some evidence suggesting that the MMR system is involved in somatic hypermutation, it has not yet been shown definitively that this system is present and acts in the cells in which Ig hypermutation is taking place. Hence, we have assessed the function of the MMR system in human GC centroblasts. Our results show, for the first time, that GC centroblasts express all known components of the human DNA MMR system, and that their MMR system is functionally intact in correcting DNA mismatches in a strand-specific manner in vitro.

The human cell lines HeLa S3 (CCL2.2), HCT116 (CCL-247), and CA46 (CRL-1648) were purchased from the American Type Culture Collection (ATCC) (Manassas, VA). HCT116 and CA46 cells were grown in RPMI 1640 medium supplemented with l-glutamine (2 mM), 10% heat-inactivated FBS, penicillin (100 U/ml,), and streptomycin (0.1 mg/ml). HeLa S3 cells were grown in Ham’s F12 in the presence of the above supplements. All cell culture reagents were purchased from Life Technologies (Grand Island, NY).

Anti-CD44 mAbs (clones NKI-P1 and NKI-P2) were generous gifts of Dr. Carl. G. Figdor (The Netherlands Cancer Institute, Amsterdam, The Netherlands). Ascites fluids containing anti-IgD mAb (clone HJ9) or anti-CD38 mAb (OKT10) were produced by inoculating BALB/C mice with HJ9 or OKT10 hybridomas as described previously (15). The HJ9 hybridoma was kindly provided by Dr. Moon H. Nahm (University of Rochester, Rochester, NY). The OKT10 hybridoma was purchased from ATCC. The following Abs were purchased: phycoerythrin (PE)-conjugated anti-CD38 (Leu-17), Simultest anti-CD3-FITC and anti-CD19-PE (Leu4/12), Simultest control γ1/γ1 (FITC-conjugated and PE-conjugated isotype control mouse IgG1), and FITC-conjugated goat anti-mouse Ig (all from Becton Dickinson, San Jose, CA); FITC-conjugated anti-IgD (Sigma, St. Louis, MO); anti-hMSH2, anti-hMLH1, and anti-hPMS2 (all from Calbiochem, La Jolla, CA); and anti-hPMS1 and anti-GTBP (from Santa Cruz Biotechnology, Santa Cruz, CA).

Human tonsillar B cell subsets were prepared as reported by Lagresle et al. (16). In brief, tonsils that had been freshly obtained from routine tonsillectomy were teased; the resulting cell suspensions were subjected to gradient centrifugation on Ficoll-Hypaque (Pharmacia, Uppsala, Sweden) at 20°C. After washing in HBSS, tonsillar mononuclear cells were subjected to one round of T cell depletion by rosetting with 2-aminoethyl-isothiouronium bromide (Sigma, St. Louis, MO)-treated SRBCs. The resulting pool of cells contained >95% CD19⁺ B cells. These cells were then separated into GC centroblasts and follicular mantle (FM) B cells. GC centroblasts (CD44⁻ B cells) were isolated by incubation with anti-CD44 Abs (NKI-P1 or NKI-P2) for 30 min on ice, followed by depletion using magnetic beads coupled to rat anti-mouse IgG1 Dynal beads (Dynal, Oslo, Norway). FM B cells were isolated by incubating the B cell pool with mouse anti-human CD38, followed by depletion using rat anti-mouse IgG1 Dynal beads. The resulting CD38⁻ B cells were further incubated with anti-human IgD (clone HJ9) for 30 min and subsequently with microbeads conjugated with rat anti-mouse IgG1 (Miltenyi Biotec, Auburn, CA) for 15 min on ice. The IgD⁺ fraction was separated using a magnetic cell separation column. All separation procedures were performed at 4°C to delay the death of GC B cells by apoptosis. Isolated FM B cells (IgD⁺CD38⁻) and GC centroblasts (IgD⁻CD38⁺CD44⁻) were analyzed for purity on a FACS Vantage (Becton Dickinson, Sunnyvale, CA). To determine the repair activity of human primary cells, PBMCs were isolated by Ficoll-Hypaque gradient centrifugation as described above and stimulated with PMA (final 1 ng/ml; Sigma) and ionomycin (final 0.2 μM; Sigma) for 82 h.

NEs were prepared according to Holmes et al. (17) unless otherwise indicated. Cell lines in the log phase of growth, resting and stimulated human PBMCs, and human tonsillar B cell subsets (FM B cells and GC centroblasts) that had been isolated as described above were harvested to prepare NEs. After obtaining ≥0.5g of both GC centroblasts and FM B cell pellets, the cell pellets of each type were pooled. Each NE was frozen in small aliquots in liquid N₂ and stored at −80°C.

Equal amounts of each NE were lysed in sample buffer (125 mM Tris-HCl (pH 6.8), 2.5% glycerol, 0.4% SDS, 1% 2-ME, and 0.01% bromophenol blue) by boiling for 3 min and were separated on a 7.5% SDS polyacrylamide gel. After electrophoresis, protein bands were transferred to a nitrocellulose membrane (Hybond enhanced chemiluminescence; Amersham, Cleveland, OH). The membrane was rinsed briefly in Tris-buffered saline (TBS), blocked with blocking solution (5% nonfat dried milk in TBS with 0.1% (v/v) Tween-20 (TBST)) for 1 h at room temperature, washed briefly with TBST, and incubated with the primary Ab (1:100) for 2 h at room temperature. The filter was rinsed extensively with TBST and subsequently incubated with the secondary Ab, horseradish-peroxidase-conjugated anti-mouse or rabbit Ig (1:5000; Amersham) for 1 h at room temperature. The filter was washed with three changes of TBST. Signals were detected by enhanced chemiluminescence (Amersham) as recommended by the manufacturer. The same blot could be used to probe with the different Abs shown in the experiments by removing the probe in stripping buffer (100 mM 2-ME, 2% SDS, and 62.5 mM Tris-HCl (pH 6.7)) at 50°C for 30 min with occasional agitation followed by washing at room temperature with large volumes of TBST.

Immunoperoxidase staining was performed according to the protocols provided by the manufacturer (Dako LSAB kit; Dako, Carpinteria, CA). The 5-μm thick cryosections mounted on silanized slides (Dako, Kyoto, Japan) were air dried for 1 h and fixed in ice-cold acetone for 10 min at 4°C. After rehydration and preblocking with TBST containing 1% BSA (Sigma), the slides were incubated with the primary Abs (1:50) in TBST containing 1% BSA for 16 h at 4°C in a humidified chamber. Slides were then rinsed three times in TBS and incubated with biotinylated secondary Ab. Streptavidin-conjugated peroxidase was then added, and bound peroxidase was visualized by the addition of substrate-chromogen solution. Slides were counterstained with hematoxytin (Sigma) to stain the nuclei. Immunostained slides were observed by light microscopy. For the negative controls for the immunostaining, TBST was used in place of primary Ab.

The gel shift assay was performed as reported previously (18), with some modifications. In brief, oligonucleotide duplexes containing a single G-T mispair or a control A-T base pair were formed by annealing 5′-AATTCGCTAGCAAGCTTTCGATTCTAGAAATTCGGC-3′ with 5′-AATTCGCCGAATTTCTAGAATCGAGAGCTTGCTAGC-3′ or 5′-AATTCGCCGAATTTCTAGAATCGAAAGCTTGCTAGC-3′, respectively. Annealed duplexes were radiolabeled by filling in with Klenow DNA polymerase (Promega, Madison, WI) in the presence of [α-³²P]deoxyATP (3000 Ci/mmol; Amersham). ³²P-labeled oligonucleotide duplexes were incubated with 1 μg of each NE in 20-μL reactions containing 10 mM HEPES (pH 7.5), 1 mM EDTA, 1 mM DTT, 5 mM MgCl₂, BSA (50 μg/ml), and 4% glycerol. After a 20-min incubation on ice, 5 μL of 50% sucrose was added and the samples were subjected to electrophoresis at room temperature through 6% polyacrylamide in 6.7 mM Tris-acetate (pH 7.5) and 1 mM EDTA.

Phage stocks of f1 MR1 and f1 MR3 (19) were generously provided by Dr. Paul Modrich (Duke University Medical Center, Durham, NC). G-T and A-C heteroduplexes as well as A-T and G-C homoduplexes were prepared according to a protocol provided by Dr. James T. Drummond (Indiana University, Bloomington, IN). In brief, bacteriophage f1 replicative form was isolated from infected XL1-Blue cells grown in the presence of tetracycline (5 μg/ml; Sigma). Viral ssDNAs were isolated from purified virions. DNA heteroduplexes were prepared by mixing f1 duplexes (100 μg of replicative form, linearized with Sau96I) with a fivefold molar excess of viral strands followed by alkaline denaturation and renaturation. Circular duplexes containing a single nick were isolated by hydroxyapatite chromatography (Bio-Rad Laboratories, Hercules, CA) and dialyzed against 10 mM Tris-HCl (pH 8.0) with 1 mM EDTA. The resulting dsDNA containing a single base mismatch at position 5632 and a site-specific single-strand break at position 5757 in the cDNA strand (refer to figure 1 in Fang and Modrich (20)) was dissolved in distilled water.

MMR assays were performed as described in Holmes et al. (17). In brief, the MMR reaction (25 μL) contained 20 mM Tris-HCl (pH 7.6), 5 mM MgCl₂, 1 mM glutathione, 50 μg/ml BSA, 0.1 mM of deoxynucleoside triphosphate, 1.5 mM of ATP, 24 fmol (0.1 μg) of heteroduplex DNA, and 100 μg of NE. After 20 min of incubation at 37°C, three volumes of stop solution (25 mM EDTA, 0.67% SDS, and 100 μg/ml proteinase K; (Sigma)) were added; samples were extracted twice with phenol and twice with chloroform. The DNA was collected by ethanol precipitation, dissolved in water, and digested with diagnostic restriction enzymes (Bsp106 and XhoI for the A-C heteroduplex and Bsp106 and HindIII for the G-T heteroduplex). Digestion products were separated by electrophoresis on a 1% agarose gel stained with ethidium bromide.

As an initial step to explore whether GC centroblasts express components of DNA MMR and are capable of correcting mismatches, we isolated human tonsillar GC centroblasts. As a negative control for Ig hypermutation, naive FM B cells with nonmutated germline Ig genes were also isolated. Because GC B cells (IgD⁻CD38⁺) are a mixture of centroblasts and centrocytes, we further separated centroblasts and centrocytes. From the available surface markers that distinguish centroblasts from centrocytes, we chose anti-CD44 Abs over anti-CD77 Abs. Feuillard et al. (21) found that anti-CD44 Ab is more reliable than anti-CD77 Ab in discriminating between centroblasts and centrocytes. We assessed the effectiveness of this separation method by FACS analysis (Fig. 1). More than 97% of CD44⁻ B cells are IgD⁻CD38⁺ (Fig. 1,B) and show a high level of CD77 expression (data not shown). Thus, negative selection of B cells not expressing CD44 on their surface yields a preparation of GC B cells that is selectively depleted of centrocytes. We also routinely obtained FM B cells with >95% purity (Fig. 1 C).

Defects in MMR result in an elevated rate of spontaneous mutations, termed the “mutator phenotype” (22). Among the known human MMR proteins, hMSH2 is known to initiate MMR by recognizing mispairs after forming specific mispair-binding complexes with hMSH3 or hMSH6 (GTBP) (18, 23, 24). The hMLH1/hPMS2 heterodimer (hMutLα) is also required for functional MMR (25). NEs derived from immunomagnetically sorted GC centroblasts and FM B cells were used to examine the presence of human MMR proteins. MMR-proficient HeLa cells (17, 26) are included as a positive control for each human MMR protein. An immunoblot with increasing amounts of protein from each NE shows the expression of those five MMR proteins in GC centroblasts (Fig. 2). The intensity of each signal obtained from the NE derived from GC centroblasts was comparable with that of MMR-proficient HeLa NE. Alternatively, FM B cells expressed only low basal levels of hMSH2 and hMLH1 proteins and similar levels of the other three proteins.

These results were explored further by immunohistochemistry using hMSH2 Ab to localize the cells expressing the hMSH2 protein in cryosections of tonsil. As expected from the immunoblot analysis, GC centroblasts are stained preferentially with the anti-hMSH2 Abs; the stained cells are localized prominently to the dark zone (mainly centroblasts) in which Ig somatic hypermutation takes place (Fig. 3). Next, adjacent serial sections were stained with an Ab against the proliferation-associated molecule Ki67 (data not shown), confirming that hMSH2 Ab-stained cells in the dark zone were rapidly proliferating centroblasts. When these observations are taken together, we conclude that the human MMR proteins involved in strand-specific DNA MMR are all expressed in GC centroblasts at levels similar to those in MMR-proficient HeLa cells. Among the DNA MMR proteins examined here, hMSH2 and hMLH1 proteins are expressed preferentially in GC centroblasts compared with FM B cells. Preferential expression of these two MMR proteins in GC centroblasts is not surprising, because both MSH2 and MLH1 are known to interact with proliferating cell nuclear Ag, which is involved in the cell cycle in yeast (27, 28). Hence, these two proteins would be expected to be expressed at a higher level in proliferating cells. It is of interest to note that hPMS2 is expressed at a similar level in each of the cell types.

The human MSH2/GTBP heterodimer that is present in HeLa cells binds to G-T mispairs with high affinity in vitro (18, 23); hMSH2 is reportedly central to all mismatch recognition (24). Based on our immunoblot analysis, we tested whether the mismatch-binding proteins present in GC centroblasts are capable of binding to a mismatched base pair using gel mobility shift assays (Fig. 4). The radiolabeled oligoduplex probe containing a single G-T mismatch yields a single bound complex with NEs prepared from both GC centroblasts and HeLa cells. The unlabeled oligonucleotide duplex containing the G-T mispair abolishes this DNA-protein interaction. When the A-T homoduplex was used as a probe in the same assay, some bound complex was detected; however, the signal was much lower. This result suggests that the mismatch-binding proteins in GC centroblasts recognize mismatched DNA substrates and initiate repair.

To determine whether the MMR system of GC centroblasts is functionally intact, we used an in vitro MMR assay that was developed in Dr. Paul Modrich’s laboratory. We examined whether NE from human GC centroblasts contains MMR activity that is capable of correcting mispairs in a strand-specific manner. The DNA MMR efficiency of GC centroblasts was compared with those of other NEs using diagnostic restriction enzymes that cut only when the mismatch was repaired on the nicked strand. Repair products were detected in both HeLa cells and in the Burkitt’s lymphoma B cell line CA46 (Fig. 5). HCT116, a colon carcinoma cell line that is defective in the expression of hMLH1 protein, was included as a negative control. To determine the sensitivity of this assay system in detecting the repair activity of primary human B cells, we also measured the repair activity of stimulated human PBMCs and resting human PBMCs. A comparable level of repair activity was detected in stimulated PBMCs, whereas no such activity is detected in resting PBMCs (Fig. 5). We detected MMR products in the NE prepared from isolated human GC centroblasts using the same assay system. Our findings indicate that the NE of GC centroblasts is functionally proficient in correcting mismatches in a naked heteroduplex DNA in a strand-specific manner in vitro.

Recent studies by Cascalho et al. (14) imply that the MMR protein PMS2 has a role in Ig somatic hypermutation. In homozygous PMS2 knockout mice, these researchers found a much lower than normal rate of Ig hypermutation (14), even though these mice had an elevated rate of mutation in their non-Ig genes (29). Cascalho et al. proposed that the PMS2 protein is involved in enhancing Ig gene hypermutation rather than preventing it. However, recent studies have shown that the frequency of mutation in mice deficient in a number of DNA repair genes is similar to that of normal mice, even though the pattern of mutation was changed in some cases (30, 31, 32). These investigators concluded that the DNA repair systems that they examined are not required for the induction of hypermutation. In our study, we have directly examined the expression and function of MMR in the centroblasts and found the system to be normal. Our results from this complementary approach support the conclusion that MMR does not play a significant role in generating hypermutation.

Somatic hypermutation within a restricted region of the Ig genes in GC centroblasts in the presence of functional MMR components suggests that MMR is not responsible for inducing Ig hypermutation. It is possible that a specific chromatin structure at the Ig loci or the presence of a mutator factor might be involved in introducing mutations into Ig V genes. The mechanism of Ig somatic hypermutation is still unknown, but several lines of evidence suggest that it is linked to transcription (33, 34, 35). Among the important factors involved in the transcription of Ig genes, the involvement of the Ig promoter and the polymerase II system can be ruled out. The Ig promoter can be replaced by a heterologous promoter such as the β-globin promoter (36), and the polymerase II system is universal. On the other hand, Ig enhancers and the nuclear matrix attachment region (MAR) are essential for a full Ig hypermutation (36), although the mechanism of their actions in this process is unknown. A recent experiment demonstrated that the targeting of Ig hypermutation in the Igκ gene is dependent on both the nuclear MAR that flanks the intronic enhancer and the core and flanking sequences of the 3′ enhancer element (35). Ig enhancers and nuclear MARs with their associated cluster of binding sites for topoisomerase II may initiate hypermutation by promoting the formation of single-strand nicks, as suggested by Neuberger and Milstein (6).

In conclusion, we have examined the function of the MMR system in GC centroblasts. Our finding of the presence of functional MMR activity in isolated human tonsillar GC centroblasts provides the first evidence that hypermutating B cells do not have a general suppression of MMR.

We thank Drs. Paul Modrich, Carl G. Figdor, and Moon H. Nahm for their generous gifts of f1 MR1 and f1 MR3 phage stocks, anti-CD44 mAbs (clones NKI-P1 and NKI-P2), and HJ9 hybridoma, respectively. We also thank Drs. Mark S. Schlissel and James T. Drummond for their comments on the manuscript and Jennifer Macke for substantive editing. We owe special thanks to Drs. James T. Drummond and Garnett Kelsoe for help in setting up the in vitro mismatch repair assay and in starting cell separation of GC and FM B cells at the initial stage of this study, respectively. We also thank the staff members of the Department of Otorhinolaryngology-Head and Neck Surgery at Samsung Medical Center for providing tonsil specimens.