Molecular analysis of immunoglobulin genes in diffuse large B-cell lymphomas

Tumor tissues were chosen randomly from the available fresh frozen biopsy specimens of lymph nodes or extralymphatic lymphoid tissue from untreated patients with a pathologic diagnosis of primary DLBCL according to the REAL classification.24 Table1 summarizes the biopsy sites from which tissue samples were obtained. The tissue samples were embedded in Tissue-Tek Optimal Cutting Temperature (OCT) compound 4583 (Miles Inc., Elkhart, IN) and preserved at −80°. Twenty-seven samples were obtained from the University of Nebraska Medical Center and 44 samples from the Stanford University Medical Center. Clinical outcome data were available for all samples.

The randomly chosen frozen tissue specimens were reviewed to confirm the DLBCL diagnosis according to the REAL classification24before RNA extraction. Samples derived from Stanford University Medical Center were stained with anti-Ig heavy and light chain antibodies (Becton Dickinson, San Jose, CA) and with the 9G4 antibody (generous gift from Dr. F. Stevenson, Southhampton, UK), a rat IgG 2a antibody directed to FR1 of V_H 4-34.14 

Frozen sections, 5 μ thick, were cut onto slides coated with poly-l-lysin (Sigma, St. Louis, MO). Frozen sections for staining with anti-Ig heavy and light chain antibodies were fixed in cold (4°C) acetone for 10 minutes, air dried, and incubated with mouse antihuman Ig μ, κ, or λ antibodies at 25°C for 30 minutes. Sections were then rinsed in 25°C phosphate-buffered saline (PBS) and incubated with biotin-conjugated goat antimouse antibody (Jackson Immuno Research, West Grove, PA) for 40 minutes at 25°C. Frozen sections for staining with the 9G4 antibody were fixed in cold formalin for 10 minutes, washed in 25°C PBS, and stained with the 9G4 antibody at 25°C for 30 minutes. The 9G4 antibody was detected with biotin-conjugated goat antirat antibody (Jackson Immuno Research) for 40 minutes at 25°C. Secondary antibody-stained sections were rinsed in 25°C PBS and incubated with streptavidin-conjugated horseradish peroxidase (Jackson Immuno Research) for 40 minutes at 25°C. Sections were then rinsed twice in PBS and reacted with 30 mg/mLl of diaminobenzidene (Sigma) in 0.01% H₂O₂ in PBS for 5 minutes at 25°C. Sections were rinsed in PBS and in water and incubated with CuSO₄(0.5% in 1 N NaCl) for 5 minutes at 25°C. Sections were then rinsed once more in water, counterstained in 2% methylene blue for 20 minutes, and finally rinsed, dehydrated, and examined. Morphologically identified lymphoma cells were considered positive if a clear membrane immunoperoxidase reaction product was seen.

Total cellular RNA or messenger RNA was isolated from the cryopreserved DLBCL specimens using the RNeasy kit (Qiagen, Valencia, CA) or FastTrack 2.0 kit (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The RNA was reverse transcribed with 15 u AMV Reverse Transcriptase (Promega, Madison, WI) per 1μg of RNA at 45°C for 45 minutes in a 30-μL volume in a buffer containing 5 mM MgCl₂, 10 mM Tris-HCl pH 8.8, 50 mM KCl, 0.1% Triton X-100, 1 mM of each dNTP, 1 U/μL Recombinant Rnasin Ribonuclease Inhibitor, 0.5 μg Oligo(dT)₁₅ per 1 μg RNA and 1.5 μg RNA. One thirtieth of a complementary DNA (cDNA) sample was amplified by Taq DNA polymerase with a specific 5′ primer corresponding to 1 of the 6 human variable H chain family leaders (VH₁through VH₆) and a 3′ antisense JH consensus primer.25 In our hands the VH₁ leader primer also amplifies sequences from the closely related VH₇family. PCR was performed in a final volume of 50 μL containing 0.5 μM of each primer, 20 mmol/L Tris-HCl (pH 8.4), 50 mmol/L KCl, 1.5 mmol/L MgCl₂, 200 μmol/L of each dNTP and 2.5 U Taq DNA polymerase (Gibco BRL, Grand Island, NY). The PCR conditions were: 96°C for 5 minutes, 55°C for 1 minute, 72°C for 3 minutes, 1 cycle; 94°C for 30 seconds, 55°C for 30 seconds, 72°C for 30 seconds, 30 to 35 cycles; and 72°C for 5 minutes, 1 cycle. Beta-2-microglobulin (β2 M) was amplified using specific primers (5′ β2M: ATCCAGCGTACTCCAAAGATT and 3′β2M: CATGTCTCGATCCCACTTAAC) and served as a control of RNA-cDNA integrity. For each PCR, a control with no added template was used to check for contamination. To control for potential PCR error, all patient samples were evaluated by 2 independent PCR reactions and sequencing, as described later. PCR products were analyzed by 2% agarose gel electrophoresis and stained with ethidium bromide. Bands of appropriate size were excised from the gels and purified by adsorption to a silica matrix (QIAquick columns, Qiagen).

Direct DNA sequencing of PCR amplicons was performed on an 373 automatic DNA sequencer (Applied Biosystems, Foster City, CA) using ABI Prism Big Dye Terminator Kit (Perkin Elmer, Foster City, CA) as recommended by the manufacturer. The same primers used for the PCR were used for sequencing. The sequence was defined as clonal if identical CDR3 sequences were obtained from 2 independent PCR reactions. If direct DNA sequencing attempt of the PCR amplicon failed to recover an unambiguous sequence, the PCR amplicons were cloned into a TA-PCR cloning vector (Invitrogen). After the transformation of competent Escherichia coli (1 Shot INV αF′, Invitrogen) and plating on selective agar (50 μg/mL of kanamycin, 40 μL of the 40 μg/mL of the X-gal), 10 to 12 white colonies were picked per sample and used in a second round of PCR. Identity of the products was established by restriction digest and 3 to 6 amplicons were selected for sequencing.

To determine the RT-Taq polymerase error rate of our experimental design, 26 clones of β2 mol/L were sequenced. These clones were generated according to the same RT-PCR and cloning procedures as used for the VH genes. Our RT-Taq error rate thus established is 0.09%, which amounts to 0.36 mutations per VH clone.

To evaluate for the presence of ongoing mutations in primary DLBCL, 8 specimens were examined by repeated cloning and sequencing of at least 15 molecular clones from each specimen. For evaluation of intraclonal heterogeneity, the following definitions were used: unconfirmed mutation—a substitution mutation observed in only 1 of the VH gene molecular clones from the same tumor specimen; confirmed mutation—a mutation observed more than once in the VH gene molecular clones from the same tumor specimen.

Only the confirmed mutations were considered as evidence of intraclonal heterogeneity; the unconfirmed mutations were disregarded because they can be caused by Taq polymerase error.

Sequence analysis was done using programs MacVector and Assembly Lign (Oxford Molecular Group, Campbell, CA). Sequences were aligned with germline sequences derived from Vbase database and DNA plot on the Internet. The V_H gene sequences were compared with the germline genes with the highest homology and, accordingly, the number of somatic mutations was determined. Mutations at the last nucleotide position of the sequenced fragment were excluded from the mutational analysis because they might result from nucleotide deletion at the joining sites. Percent of sequence identity was calculated from the aligned sequences from the beginning of FR1 to the end of FR3.

The probability that an excess or scarcity of replacement (R) mutations in V_H CDRs or FRs occurred by chance was calculated by a multinominal distribution model. The total number of mutations in each V_H gene is denoted by n = r₁ +s₁+r₂ + s₂, in which r₁ and r₂ are replacement mutations and s₁ and s₂ are silent mutations in FR and CDR regions, respectively. The theoretical probabilities for r₁, s₁, r₂, and s₂ mutations are denoted by p₁, q₁, p₂, and q₂, respectively. These probabilities are calculated by the following equations: p₁ = Rf_FR × Lr_FR; q₁ = (1-Rf_FR) × Lr_FR; p₂ = Rf_CDR × (1-Lr_FR) and q₂ = (1-Rf_CDR) × (1-Lr_FR) in which Lr_FR is a relative size of the FR and Rf_FR and Rf_CDR are the inherent susceptibility to R mutations of the FRs and CDRs, respectively. Rf_FR and Rf_CDR were calculated for each of the identified germline genes and are based on the chance of the occurrence in each codon of an amino acid replacement given any single nucleotide change not resulting in a termination codon.

The probability of observing r₁ or fewer replacement mutations in FR regions is then given by the multinominal tail probability:

The sum is taken over values of k ranging from 0 to r₁ and all combinations of S₁, R₂, S₂ such that k+ S₁ + R₂ + S₂ = n. To compute the P value of an observed number r₁, it is customary to split the probability at r₁: Pvalue = P (R₁ < r₁) + 0.5 × P (R₁ = r₁). It should be noted that P (R₁ = r₁) = P(R₁ ≤ r₁)-P(R₁ ≤ r₁-1), and P(R₁< r₁) = P(R₁ < r₁)-P(R₁ = r₁).

The probability of observing r₂ or more replacement mutations in CDR regions is similarly computed by the following equation:

And the P value is computed by Pvalue = P(R₂ > r₂) + 0.5 × P(R₂ = r₂). For both FR and CDR, 1-sidedP values were used.

The Chang and Casali equation26 was not applied because it considers a binominal distribution of mutations while the distribution is multinominal and it does not consider all the combinatorial possibilities of mutations and results in less stringent analysis.

Reverse transcriptase-PCR using VH leader and JH consensus primers was performed in 71 patients with primary untreated DLBCL. In 3 (4.2%) patients an Ig V_H gene PCR product could not be detected despite successful amplification of β2M that served as a control for RNA-cDNA integrity. In 6 (8.4%) patients, multiple nonclonal bands, as determined by subcloning and sequencing, were detected. Most probably these bands derive from polyclonal reactive B cells infiltrating these tumors. The failure to detect monoclonal V_H gene sequences in these 9 cases may result from somatic mutations in the region to which PCR primers used in this study are designed to hybridize, thus leading to lower amplification efficiency and possible false-negative results. Alternatively, absence of Ig rearrangement, as was previously reported in rare DLBCL cases,16 27 may explain the absence of monoclonal V_H gene sequences in some of these cases.

Seventy clonal V_H gene sequences were detected in 62 patients (2 clonal sequences in 6 patients and 3 clonal sequences in 1 patient). In 43 (69%) cases the PCR product could be sequenced directly, whereas in 19 (31%) cases PCR amplicons had to be subcloned to identify the V_H gene. In 4 of these cases, the clonal product amplified by the VH3 leader primer had 200 bp or larger substitution inserts located between the leader and the JH regions, substituting part of the natural V_H gene sequence and resulting in open reading frame sequences in 2 samples. The presence of these inserts was verified by repeated PCR and cloning. A search in the GenBank data library could not identify similar sequences. These V_H genes were excluded from analysis.

Among the remaining 66 clonal V_H gene sequences, 58 were potentially functional and 8 were rendered nonfunctional by out-of-frame rearrangement (5 sequences) or by somatic mutations leading to the introduction of stop codons in the V region (3 sequences). The nonfunctional sequences were derived from the following V_H genes: VH1-08, VH1-46, VH2-05, VH3-23, VH3-07, VH3-09, VH4-61, and VH4-30-4. Three of these nonfunctional sequences derived from tumors that had an additional potentially functional V_H gene sequence, whereas in 5 cases the nonfunctional sequence was the solitary V_H gene isolate found in the tumor cells.

Fifty-eight potentially functional V_H gene sequences were detected in 53 DLBCL cases (Table2). In 5 cases, 2 potentially functional V_H genes were found (in one of these, an additional nonfunctional clonal V_H gene was also detected). The multiplicity of potentially functional V_H genes in the same tumor may be attributed to: (1) lack of allelic exclusion, as was previously reported in the B-cell chronic lymphocytic leukemia (CLL)28; (2) numerical chromosomal aberrations that are reported in 91% of DLBCL29; gain of an additional chromosome 14 may result in the presence of 3 rearranged Ig genes in the tumor cells; or (3) biclonal B cell tumor with 2 different clonal VH genes.30 

The 58 potentially functional V_H genes used by these 53 DLBCL cases were derived from 6 of the 7 human V_H gene families in the following distribution: VH1,12.1%; VH2,10.3%; VH3, 44.8%; VH4, 25.9%; VH5, 5.2%; and VH7, 1.7% (Table3). This V_H family distribution is comparable with the relative complexity of functional germline V_H genes within each family and to the use of V_H families in peripheral and lymph node lymphocytes in healthy donors, as was previously established by various techniques (see Table 3).2 31-34 Thus, V_H gene family use by untreated primary DLBCL is random and unbiased.

The most frequently encountered genes were VH3-23 (n = 6), VH3-33 (n = 5), VH4-34 (n = 5), VH4-39 (n = 5), VH3-48 (n = 4), and VH2-05 (n = 4), which represent 10.3%, 8.6%, 8.6%, 8.6%, 6.9%, and 6.9%, respectively, of all the potentially functional V_H genes identified in this study. Some of these genes, including VH3-23, VH4-34, and VH4-39, are also found at higher than expected frequency in normal individuals.33 35-37 

No correlation between V_H gene family use and the IPI or patient's survival was detected (data not shown).

Forty-four tumor samples derived from Stanford University Medical Center were stained with anti-Ig and 9G4 antibodies. The staining results were interpreted blindly, without the knowledge of sequencing results. Staining with 9G4 antibody, directed to the VH4-34 gene, was positive in 5 cases, identifying all the VH4-34 DLBCL cases determined by molecular analysis without false-positive or false-negative results.

Staining with anti-Ig antibodies was positive in 19 (42.3%) cases and negative in 25 (56.8%). Both lower (17%)38 and higher (> 60%)39 prevalence of anti-Ig-stained negative DLBCL cohorts are reported in the literature. All the tumors that reacted with anti-Ig antibodies had potentially functional Ig V_Hgene PCR product. Among the 25 tumors that lacked staining with anti-Ig antibodies, 8 had solitary nonfunctional V_H gene sequences, including the 4 cases with the insertions described above and 4 cases in which monoclonal V_H genes were not found. However, in 13 cases that did not stain with anti-Ig antibodies, a potentially functional monoclonal V_H gene sequence was detected. The discrepancy between the PCR sequencing and the staining results may stem from a posttranscriptional effect on Ig expression by the requirement of proper Ig protein folding, that could be impaired in these cases by replacement somatic mutations. Moreover, the light chain sequences, which are required for Ig stabilization and expression, were not analyzed. The presence of nonfunctional rearranged light chain may explain the observed disconcordance between the staining and the PCR sequencing results. Noticeably, all the cases that did not stain with anti-Ig antibodies while harboring a potentially functional monoclonal V_H gene sequence also did not stain with anti-κ or anti-λ chain antibodies.

A similar V_H gene family use was found in the cases that stained or did not stain with anti-Ig antibodies (data not shown).

Intraclonal variation was assessed by extensive molecular cloning in 8 potentially functional V_H gene isolates from 8 tumor specimens obtained from various biopsy sites (Table4). These specimens were selected randomly and included 7 samples whose clonal Ig sequence was established by direct sequencing of the PCR product and 1 sample in which the V_H gene sequence was established after molecular cloning from the PCR product. In 2 of the tested samples, the extensively mutated clonal V_H gene isolates did not show intraclonal heterogeneity. In 6 samples intraclonal heterogeneity was detected. In these cases molecular clones harboring confirmed mutations not observed in the most abundant clonal V_H gene sequence were found. The extent of the intraclonal heterogeneity varied between DLBCL samples, 3 specimens (DLBCL no. 4, 34, and 41) demonstrated a limited number of additional mutations, whereas others (DLBCL no. 15, 20, and 43) exhibited extensive variations between tumor V_Hgene subclones, similar in prevalence to that typically observed in FL.40-42 

Only 3 V_H gene sequences had less than 2% difference from the most similar germline gene and an additional 3 V_Hgene isolates had a germline unmutated sequence. More than 2% difference from the most similar germline sequence was detected in the remaining 52 potentially functional and the 8 nonfunctional V_H gene sequences. No significant difference in the mutation level was observed between the nonfunctional and potentially functional V_H gene isolates, thus suggesting that the mutational process was active on both allele products. Forty (68.9%) of the potentially functional V_H gene isolates differed by more than 5% from the most similar germline counterpart and 24 (41.4%) differed by more than 10%. In the 5 cases with 2 potentially functional V_H gene isolates, 3 had a similar level of mutation in both identified V_H genes, whereas in 2, only 1 isolate was mutated while the second was unmutated. No correlation between the level of mutation and V_H gene family use was observed.

Analysis of the distribution of the R versus S mutations demonstrated that 33 of the 55 potentially functional mutated V_H genes contained a significantly lower number of R mutations in the FRs (see Table 2) than would be expected if mutations had occurred by chance alone without selective forces. Thus, in these genes a selective force to conserve FR sequences and maintain binding to an antigen can be inferred. In most of the mutated sequences the CDRs R/S ratio values were higher than those within FRs of the corresponding V_Hgenes. However, in only 13 potentially functional sequences was the presence of the positive selective forces observed, as demonstrated by an excess of R mutations in the CDRs, exceeding that expected to occur by chance (see Table 2). In 12 DLBCL cases concomitant scarcity of R mutations in FRs and excess of R mutations in CDRs were observed. We performed a search for similar R mutations within the CDRs of isolates belonging to the same V_H genes. Recurrent amino acid changes were not observed in the tumor samples tested in this study.

The present study was undertaken to clarify the issues of V_H gene use and somatic mutation in DLBCL. For this purpose, molecular analysis of the Ig VH region was performed in 71 untreated primary DLBCL patients—the largest cohort studied to date.

This study demonstrates an unbiased V_H gene use in DLBCL. The VH3, the largest family, was used most often, followed by VH4, VH1, VH2, VH5, and VH7, an ordering that is similar to the number of functional genes within each family.2,3 The V_Hgene use in DLBCL is also similar to the previously reported use of V_H families in peripheral and lymph node lymphocytes in normal individuals (see Table 3).2,31-34 VH4-34 gene was used in 5 (8.6%) potentially functional V_H sequences similar to its use in normal individuals.36 37 

Previous studies evaluating V_H gene use in DLBCL reported disconcordant results. Daley et al12 evaluated V_H gene use in 10 DLBCL cell lines and reported random V_H family use that mirrors V_H family complexity and V_H family use in normal B cells. Rosenquist et al16 studied 35 DLBCL samples by PCR and found an unbiased V_H family use. However, sequencing or cloning analysis of V_H genes to verify clonality and to exclude amplification of Ig genes derived from infiltrating reactive B cells was not performed in that study. Kuppers et al13 examined 19 DLBCL tumors and reported unbiased V_H family use.

In contrast, biased V_H family use with overrepresentation of VH4-34 gene was reported in other DLBCL studies. Stevenson et al14 evaluated the use of VH4-34 gene in NHL by immunostaining with 9G4 antibody, which recognizes the A-V-Y amino acid sequence within FR1 of VH4-34. Five of the 28 intermediate and high-grade NHL used VH4-34; however, it was not clear which of these cases were DLBCL. Funkhouser and Warnke15 found VH4-34 gene use in 6 of 20 DLBCL cases by using the same antibody. In none of these studies was the clinical status of evaluated cases reported, making it unclear if these cases were evaluated at presentation, before therapy, or at the time of lymphoma relapse. In a previous study performed in our laboratory,11 biased V_H family use in DLBCL was found, with 14 of 17 cases using VH4 family genes, 11 of which were VH4-34. This study group mostly included samples acquired from patients at the time of lymphoma relapse. None of these cases was included in the present study. In addition, we have reexamined these cases by our current methods and reconfirmed the previously published results. The reason for the discrepancy in the reported use of V_Hfamilies in DLBCL is unclear, but it may be caused by the inclusion of untreated and relapsed DLBCL cases in different studies and the relatively small size of the previous study cohorts. Analysis of a large group of untreated DLBCL specimens, as was done in the present study, should put to rest the issue of V_H gene use in this tumor.

Intraclonal heterogeneity is a consistent finding in FL, which represents a germinal center tumor prototype. However, contradicting results regarding the presence of intraclonal heterogeneity in DLBCL were reported in the literature. Intraclonal heterogeneity was shown in several untreated DLBCL cases,21,22 primary testicular lymphoma,23 and in some but not all the specimens of primary DLBCL of the brain43 and stomach.44Absence of intraclonal variation in DLBCL was previously reported in relapsed DLBCL cases11 and in DLBCL of the skin19 and in patients with acquired immunodeficiency syndrome.18 The extent of the molecular cloning may account for the observed discrepancy. Evaluation of insufficient number of molecular clones may fail to discriminate between real mutations and Taq errors and fail to detect a low level of intraclonal heterogeneity. Indeed, our initial examination of a small number of clones in 19 specimens in this study failed to reveal intraclonal heterogeneity, whereas evaluation of a larger number of molecular clones disclosed its presence. The present study extensively cloned 8 untreated DLBCL specimens from different tissues. The results suggest the presence of 2 DLBCL subgroups based on the presence of intraclonal heterogeneity: a large DLBCL subgroup showing evidence of ongoing mutations, which is a hallmark of GC microenvironment, and a smaller subgroup without intraclonal heterogeneity. Whether these subgroups represent DLBCL cases originating from GC and post-GC cells, respectively, or whether the transforming events may render the cell independent from the influence of GC microenvironment is presently unclear. Further studies elucidating these issues and correlating the presence of intraclonal heterogeneity with tumor immunophenotype are necessary.

Analysis of mutations in V_H genes can provide insights regarding the role of antigen prior to or during DLBCL clonal outgrowth.6 7 In 24% of the potentially functional mutated V_H genes there was evidence for positive selection as demonstrated by analysis of R mutations in CDRs. Scarcity of R mutations in FRs was observed in 60% of mutated sequences, thus suggesting selection for functional Ig in these tumors. It is possible that in some DLBCL CDRs of germline V_H genes possess sufficient antigen-binding affinity and the antigen provides negative selective pressure by selecting against FR R mutations to maintain Ig function. However, presently the identity of possible antigens involved in DLBCL clonal selection is unknown. Alternatively, it is possible that superantigens that bind Ig receptor via FR regions may be involved in stimulation of lymphoma cells, thus providing negative selection pressure on FR region without positive selection pressure on CDR regions. No evidence of antigen selective pressure was evident in 35% of mutated DLBCL cases, thus suggesting antigen independent tumor evolution. Furthermore, in contrast to normal B cells and most cases of FL, which usually express surface antigen receptors, DLBCL cells may propagate without surface antigen receptor expression, as demonstrated by negative Ig immunostaining and finding of solitary nonfunctional V_H genes in a significant proportion of DLBCL tumors.

In conclusion, the present study results revealed a random use of V_H genes by DLBCL cells. In a substantial number of DLBCL cases a clonal V_H gene sequence was not detected. Functional V_H genes were commonly mutated and only a small number of unmutated germline V_H gene isolates were found. Ongoing somatic mutations were observed in the majority of cases. Positive and or negative antigen selection pressure was observed in 65% of mutated DLBCL cases. These findings indicate a heterogeneous pathobiology of DLBCL.