Colorimetric In Situ Hybridization Identifies MYC Gene Signal Clusters Correlating With Increased Copy Number, mRNA, and Protein in Diffuse Large B-cell Lymphoma

Abstract

Abnormalities of the MYC oncogene on chromosome 8 are characteristic of Burkitt lymphoma and other aggressive B-cell lymphomas, including diffuse large B-cell lymphoma (DLBCL). We recently described a colorimetric in situ hybridization (CISH) method for detecting extra copies of the MYC gene in DLBCL and the frequent occurrence of excess copies of discrete MYC signals in the context of diploidy or polyploidy of chromosome 8, which correlated with increased mRNA signals. We further observed enlarged MYC signals, which were counted as a single gene copy but, by their dimension and unusual shape, likely consisted of “clusters” of MYC genes. In this study, we sought to further characterize these clusters of MYC signals by determining whether the presence of these correlated with other genetic features, mRNA levels, protein, and overall survival. We found that MYC clusters correlated with an abnormal MYC locus and with increased mRNA. MYC mRNA correlated with protein levels, and both increased mRNA and protein correlated with poorer overall survival. MYC clusters were seen in both the germinal center and activated B-cell subtypes of DLBCL. Clusters of MYC signals may be an underappreciated, but clinically important, feature of aggressive B-cell lymphomas with potential prognostic and therapeutic relevance.

MYC is a highly regulated gene located on chromosome 8q24 that serves as a transcription factor with diverse roles, including cell growth, biosynthesis, and cell cycle regulation.¹ Translocations of the MYC oncogene on chromosome 8 are characteristic of Burkitt lymphoma and are also frequently encountered in diffuse large B-cell lymphoma (DLBCL), “double-hit” lymphomas, plasmablastic lymphomas, and transformed lymphomas, particularly those arising from antecedent follicular lymphoma.^2,3 Although immunoglobulin heavy or light chain genes are typical translocation partners with MYC in Burkitt lymphoma, translocations with other partner genes, complex translocations, and amplifications are seen in the other types of non-Hodgkin lymphoma. Both MYC translocations and increased copy number have been reported to predict poor prognosis in DLBCL.⁴

We recently described a colorimetric method for detecting extra copies of the MYC gene in DLBCL. We observed frequent occurrence of excess copies of discrete MYC signals in the context of diploidy or polyploidy of chromosome 8, which correlated with increased mRNA signals.⁵ Other authors have since confirmed these results and found that MYC amplification correlated with poor patient outcome, particularly in the presence of del(8p) abnormalities.⁶

Of particular interest, we noted enlarged MYC signals that morphologically resembled multiple individual signals stacked on top of each other, which were counted as a single gene copy. However, their sheer dimension and unusual shape suggested that these enlarged signals were likely to consist of “clusters” of MYC genes similar to what has been observed with other oncogenes. In the current study, we sought to further characterize these “clusters” of MYC signals by determining whether their presence correlated with (1) other genetic features, including diploidy or polyploidy of chromosome 8 or MYC translocations; (2) mRNA levels; (3) protein status by immunohistochemistry; (4) activated (ABC) or germinal center B-cell (GCB) subtypes of DLBCL; and (5) overall survival. In this study, we found that MYC clusters were most commonly found in both DLBCL ABC and GCB subtype cases with extra copies of MYC in the context of the usual 2 copies of chromosome 8. The presence of MYC clusters correlated with higher mRNA and positive immunohistochemistry (IHC), which in turn correlated with poorer overall survival. These findings suggest that clusters of MYC signals are a real phenomenon with implications for tumor biology and patient management.

Materials and Methods

DLBCL Patient Cases

Formalin-fixed, paraffin-embedded (FFPE) biopsy specimens from 202 patients diagnosed with DLBCL and treated with R-CHOP (rituximab-cyclophosphamide, doxorubicin, vincristine, and prednisone) were used for this study. These cases were from the Leukemia/Lymphoma Molecular Profiling Project and are highly annotated with previously collected gene expression profiling (GEP) data.⁷ These cases were represented on 10 tissue microarrays (TMAs) with 25 also available as single whole-tissue sections. TMAs consisted of 1 to 3 core punches at 1 to 2 mm per punch. Approval to perform this study was obtained from the University of Arizona Institutional Review Board for use of these tissues in compliance with the Declaration of Helsinki.

Probe Design

The MYC probe is a polymerase chain reaction (PCR)–generated, ligated, repeat-free probe that covers flanking DNA on both sides of the MYC gene and was designed to detect amplification and not translocation. The MYC gene probe was visualized with blue detection chemistry (MYC DNP probe and ultraView Blue ISH Detection kit, Ventana Medical Systems, Tucson, AZ). Centromere 8 was visualized with red detection chemistry (Chromosome 8 Dig probe and ultraView Red ISH Detection kit, Ventana Medical Systems). The correct localization of the probe was demonstrated by our previous publication, as was the structure of the probe and the staining.⁵ This probe does not identify t(8;14) translocations.

Interpretation of Colorimetric In Situ Hybridization Staining, Evaluation of MYC Copy Number, and Signal Clustering

Each case was evaluated individually for overall staining quality. Cases lacking 2 clear, discrete signals in normal bystander cells (endothelial, stromal, or histiocytic cells) or those with staining deposits outside of cell nuclei were deemed nonevaluable cases. In each case, the total number of blue signals (MYC gene) and red signals (centromere 8) were counted within 20 cells. Events where 2 tightly associated signals were present (appearance of possible double minutes) were counted as 1 signal. The cases were evaluated individually by 2 board-certified pathologists (C.V. and P.B.) and shown in consensus with 2 other pathologists (T.G. and L.R.). The cases were divided into 3 groups based on MYC copy number and CEN8 copy number. In a normal cell, we expected to see 2 copies of MYC and 2 CEN8 signals. Therefore, as in our previous publication, the normal MYC copy number was set at less than or equal to 44 each (<10% of cells show an increased number of signals).⁵ Cases with normal MYC copy number and normal CEN8 were placed into group 1. Cases with more than 44 copies of MYC were subdivided into 2 groups based on CEN8 number. Group 2 was defined as MYC less than 44 and CEN8 less than or equal to 44 (true amplification of MYC). Group 3 was defined as cases with more than 44 MYC and CEN8 (MYC increased with polysomy). Cases with a tightly packed, overlapping/stacked MYC staining pattern in at least 20% of cells counted (≥4/20) were designated as containing “clusters.” In these “clusters,” separate signals could not be individually discerned or counted within the grouped signal. A cutoff of 4 or more cells per case was selected to rigorously exclude artifacts of stain deposition.

MYC Fluorescence In Situ Hybridization Staining and Interpretation

MYC and BCL2 translocations were evaluated using interphase fluorescence in situ hybridization (FISH) in FFPE TMAs. The probes were dual-color “break-apart” probes that are commercially available from Abbott Molecular (Abbott Park, IL) and were used as previously described. Cases that demonstrated break-apart signal in less than 5% of nuclei were interpreted as positive for the presence of the t(8;14) translocation, as previously published.⁸

MYC Messenger RNA Expression

Of the 157 DLBCL patients analyzed in this study, 105 had previously undergone gene expression profiling (GEP) experiments on corresponding snap-frozen tissue blocks. GEP was performed using the Affymetrix U133 Plus 2.0 microarray (Affymetrix, Santa Clara, CA) as previously described.⁷ From these 105 cases, 97 were included in the Lenz et al⁷ publication, and an additional 8 were included for analysis in this study. To compare signals across all 105 patient samples, the Partek Genomics Suite software, version 6.5 (Partek, St Louis, MO), was used to background correct the GEP data according to the robust multiarray analysis (RMA), which included guanine-cytosine (GC) probe content and quantile normalization.⁹ The MYC messenger RNA (mRNA) levels were determined by log2 transformation of the fluorescent signals from probe set 202431_s_at.⁷

Immunohistochemical Staining and Evaluation

The paraffin blocks were cut at 4 μ and placed onto charged slides. The primary antibody was a rabbit monoclonal antibody against the c-MYC N-terminus (clone Y69, catalog 1472-1, Epitomics, Burlingame, CA). The slides were stained on the Ventana Benchmark XT instrument (Ventana Medical Systems) using the CC1 cell conditioning regimen, and 100 μL of a 1:50 dilution of the primary antibody was incubated at 37°C for 36 minutes, detected with the OptiView DAB Kit (cat. 760-700, Ventana Medical Systems), and counterstained with hematoxylin for 4 minutes. For comparison, a completely separate set of 33 DLBCL cases assembled into a tissue microarray were stained with both the Epitomics Y69 antibody and also with the same Y69 antibody clone formulated into a Ventana primary antibody dispenser (product 790-4628, Ventana Medical Systems) with a slightly shorter incubation of 32 minutes as the only protocol change. The slides were evaluated by 2 authors (C.V. and L.R.). Cases were examined individually and evaluated at high power (×400-600). At least 300 cells were counted for each case. An individual cell was called positive if the lymphoma cells demonstrated any visually detectable 3,3′-diaminobenzidine brown staining over the nucleus. As per our previous work, staining in 40% or more of cells was considered “positive.”¹⁰

Survival Analysis

Initially, all patients were divided into 2 groups based on the presence or absence of MYC colorimetric in situ hybridization (CISH) signal clusters and evaluated for differences in survival. Subsequently, the patients were divided into the 3 MYC and CEN8 copy number groups as previously mentioned, and each group was subdivided according to the presence or absence of MYC CISH signal clusters. Survival curves were also generated for patients stratified by DLBCL subtype classification, ABC or GCB, and MYC mRNA expression to confirm the cohort used in this study exhibited similar survival rates as previously published.^7,11 For the MYC mRNA expression stratification, the 80th percentile was used as the cutoff value to designate high vs low expression.^11,12 Although many cut-points have been suggested, our previous work has suggested that this 80th percentile is most significant. Analyses using MYC protein expression were performed at the 40% or more nuclei positive level as in our previous work with IHC, as well as at an 80% or more positive nuclei level as established in our variable cut-point analysis using mRNA.^10,12 The overall survival (event = death from any cause) was estimated using the Kaplan-Meier method, and the log-rank test was used to detect any significant differences and the hazard ratio.¹³ The 95% confidence intervals were calculated with the Mantel-Haenszel method. All survival analyses were performed using GraphPad Prism software, version 4.0 (GraphPad Software, La Jolla, CA).

Statistical Analysis

All statistical tests for significance were performed using the GraphPad Prism software, version 4.0. A 1-way analysis of variance (ANOVA) was used to determine significant differences among 3 groups of data sets. The Mann-Whitney 2-tailed t test was used to determine significance in 2-group comparisons with significantly different variances; otherwise, an unpaired t test was used.

Results

Analyzable Cases

Of 202 DLBCL cases, 177 (87.6%) were successfully stained using CISH. Cases were excluded if they failed to meet the staining criteria (see Materials and Methods). If a case was represented on a TMA as well as a full section, the data from the full section were used in the analysis. This was done to include more total cases and since the findings were very similar for those cases with both TMA and whole-tissue section results. This resulted in a total of 157 unique cases, of which 19 were excluded because the biopsy specimen was not from a de novo pretreatment sample. Therefore, the remaining 138 patients with de novo, untreated DLBCL were used for all further staining and subsequent analyses.

CISH Demonstrates Frequent Presence of Increased MYC Copies, Most Commonly in the Presence of Diploidy for Chromosome 8

MYC copy number in the unique cases in all groups combined ranged from 34 to 100 copies/20 cells (average of 61 copies/20 cells or 3.1/cell). CEN8 copy number in all groups combined ranged from 15 to 100 total copies/20 cells (average 42.5 copies/20 cells or 2.1/cell).

On the basis of the MYC and CEN8 copy number data, the cases were divided into 3 groups as described in Materials and Methods. Group 1, with the normal expected 2 copies of MYC (MYC ≤44 copies), contained 24 cases (17.4%). The range of MYC copy number in group 1 was 34 to 44 copies/20 cells (average of 39.5 copies/20 cells or 2 copies/ cell as expected). In some cases, the number of MYC signals fell below 40 in 20 cells, presumably due to tissue sectioning excluding some portion of the nucleus. Group 2 (MYC >44 copies, CEN8 ≤44) consisted of 63 total cases (45.6%) with a range of 45 to 95 MYC copies/20 cells (average of 53.6/20 cells or 2.7/cell) indicating a low level of amplification or more frequent MYC copies than chromosome 8 centromeres. Group 2 CEN8 copy numbers ranged from 19 to 44 copies/20 cells with an average of 36.4/20 cells (1.8/ cell as expected). Group 3 included cases with polysomy for chromosome 8 correlating with increased MYC signals (MYC >44, CEN8 >44) consisting of 51 total cases (40%). The MYC copy number range in this group was 46 to 100 copies/20 cells (average of 67.8 copies/20 cells or 3.4/cell). For group 3, the total CEN8 number ranged from 45 to 100 copies/20 cells (average of 55.9 copies/20 cells or 2.8/cell) indicating a low level of polysomy 8 with an extra copy of MYC passively carried on the extra chromosome 8 rather than amplification. As summarized in Table 1, overall, group 2 was the largest with 45.6% of cases, followed by group 3 with 40% and group 1 with 17.4%. Therefore, the majority of cases had an extra copy of MYC either through amplification or polysomy 8.

Table 1.

MYC and CEN8 Copy Number

	Group 1	Group 2	Group 3
	MYC ≤44	MYC >44	MYC >44
	CEN8 ≤44	CEN8 ≤44	CEN8 >44
n/Total N (%)	24/138 (17.4)	63/138 (45.6)	51/138 (40.0)
MYC copies/cell, average (range)	2.0 (1.7-2.2)	2.7 (2.3-4.8)	3.4 (2.3-5.0)
MYC total copies/20 cells, average (range)	39.5 (34-44)	53.6 (45-95)	67.8 (46-100)
CEN8 copies/cell, average (range)	1.5 (0.75-2.0)	1.8 (1.0-2.2)	2.8 (2.3-5.0)
CEN8 copies/20 cells, average (range)	30.0 (15-41)	36.4 (19-44)	55.9 (45-100)
Cases with clusters, No. (%)	1/24 (4.2)	31/63 (49.2)	14/51 (27.5)
Clustered cases: number of clustered cells, average (range)	4 (4)	6.9 (4-15)	6.1 (4-14)
Cases without clusters, No. (%)	23/24 (95.8)	32/63 (50.8)	37/51 (72.5)

The visual display of these MYC and CEN8 signals by CISH was very apparent and easy to distinguish using a light microscope by the contrasting colors linked to the 2 different probes in the nuclei and the ability to visualize the morphology of the cells with the hematoxylin counterstain. Image 1 demonstrates representative examples of MYC staining for CISH and IHC. Images 1A-1C show normal CISH staining (group 1) as well as abnormal staining in groups 2 and 3. Images 1D-1G show representative examples of clustered signals that likely represent more than 1 copy of the MYC locus.

Image 1.

MYC immunohistochemistry (IHC), colorimetric in situ hybridization (CISH), and CISH clusters. A-C, MYC CISH: MYC = blue, CEN8 = red A. Diffuse large B-cell lymphoma (DLBCL) MYC CISH ×1,000. This case is a representative example of group 1 (MYC <44 copies/20 cells; CEN8 <44 copies/20 cells). B, DLBCL MYC CISH ×1,000. An example of group 2 (MYC ≥44 copies/20 cells; CEN8 ≤44 copies/20 cells). The large cell (arrow) was interpreted as 6 blue signals and not as a clustered signal. C, DLBCL MYC CISH × 1,000. An example of group 3 (MYC ≥44 copies/20 cells; CEN8 >44 copies/20 cells). D-G, MYC CISH signal clusters: DLBCL MYC CISH × 1,000. The images show various patterns of clustering found in multiple cases examined. Frequently, the signal patterns wrapped around the CEN8 (red) signal, as exemplified in Image 1F (arrow). H-J, MYC IHC. H, Tonsil control, ×400. The image shows expected staining of the proliferative layer of epithelium but not the more superficial layers of epithelium (lower left). The lymphatic endothelial cells are negative, while scattered positivity is seen in the underlying lymphocytes (upper right). I, Tonsil control, ×400. Close-up view of the germinal center showing scattered positive centroblasts, negative follicular dendritic cells (arrow), and negative histiocytes. J, DLBCL, ×400. Strong positive staining in nearly all nuclei in a t(8;14)-positive case.

Greater Than 2 MYC Signals Per Cell Are More Common Than Translocations in DLBCL

Increased MYC copy numbers (MYC >44, including both groups 2 and 3) were seen in the majority of cases (n = 114, 82.6%). In contrast, the translocations involving 8q24 using a break-apart fluorescence in situ hybridization (FISH) probe were evaluated in 121 cases for which data were available for both clustering and translocation status. Similar to what has been previously described, translocations involving the MYC locus were detected in 11 of 121 cases (9%).¹⁴ These 11 cases were distributed between groups 1, 2, and 3 Table 2. No correlation between the presence of the translocation and clustering was seen, indicating that these may represent a distinct phenomenon at a molecular genetic level (data not shown).

Table 2.

t(8;14) Translocation vs Signal Cluster Comparison^a

	Group 1	Group 2	Group 3
	MYC ≤44	MYC >44	MYC >44
	CEN8 ≤44	CEN8 ≤44	CEN8 >44
n/Total N (%)	19/121 (15.7)	57/121 (47.1)	45/121 (37.2)
Translocation positive cases	1/19 (5.2)	3/57 (5.3)	7/45 (15.6)
Cases with signal clusters	1/19 (5.2)	27/57 (47.4)	13/45 (28.9)
Cases with both translocation and signal clusters	1/19 (5.2)	1/57 (1.8)	1/45 (2.2)

^aData are given as number/total number (%).

Clusters of MYC Signals Were Frequent and Most Often Seen in Amplified Cases

Of the 138 total cases, 46 (33.3%) demonstrated MYC signal clustering in at least 4 cells within a 20-cell count, which was the predefined threshold for considering a case positive for clustering. Signal clusters were not seen in normal tissues and were never present in normal bystander cells such as endothelial cells or normal lymphocytes. There was a significant enrichment of clustering in group 2, which consisted of 49.2% (31/63 cases) clustered cases in comparison to groups 1 and 3 with 4.2% (1/24 cases) and 27.5% (14/51 cases), respectively (Table 1). As shown in Figure 1, the number of cells with MYC signal clustering significantly differed among the 3 MYC copy number groups (P < .0001). The presence of clusters also correlated with total MYC copy number Figure 2A but did not correlate with CEN8 number (P = .3215; data not shown). Specifically, clustered cases (n = 46) had a significantly (P = .0021) higher mean total MYC copy number (66.2) than nonclustered cases (n = 92; 57.9) (Figure 2A). In other words, cases with MYC clustering most often had a higher MYC copy number with no increase in CEN8 number, indicating a correlation with amplification rather than with polysomy. As expected, total MYC copy number was highest in cases with abnormal MYC (mean, 64.8) by any mechanism (amplification, polysomy, clusters, or trans-locations) compared with cases with normal diploid copies of MYC (mean, 40.0, P < .0001) Figure 2B. It is important to note that we may have underestimated the gene copy number of MYC within the clusters due to the inability to adequately distinguish the individual MYC signals in these large signals. Nonetheless, the presence of clustered signals was enriched in group 2 (cases with amplified MYC signal), supporting their association with increased MYC copy number. We next determined if the presence of clusters correlated with MYC mRNA in the 105 DLBCL cases with GEP data. However, no correlation was found between mRNA level and the presence of MYC signal clusters Figure 2C. In addition, MYC mRNA levels in cases with abnormal MYC gene status (by amplification, polysomy, clusters, or translocations) did not significantly differ from mRNA levels in cases with normal MYC gene status Figure 2D.

Figure 1.

MYC clustering in diffuse large B-cell lymphoma (DLBCL) cases with MYC gene amplification. The DLBCL tumor sections were divided into 3 groups according to MYC and centromere 8 (CEN8) copy number. Group 1 consisted of cases with MYC and CEN8 copy number equal or less than 44 (normal MYC), group 2 with MYC copy number greater than 44 and CEN8 less than 44 (MYC amplification), and group 3 with MYC copy number and CEN8 greater than 44 (chromosome 8 polysomy). Lines represent the mean. The number of cells that contained clustering of MYC colorimetric in situ hybridization signal was counted for a total of 20 cells per section. A 1-way analysis of variance was used to determine significance (P < .0001).

Figure 2.

Total MYC copy number and messenger RNA (mRNA) levels in cases with normal or abnormal MYC and in the presence or absence of MYC clustering. The total MYC copy number (A, B) and log2-transformed MYC mRNA levels (C, D) were compared between diffuse large B-cell lymphoma patient cases with normal (not increased) and abnormal (increased MYC and/or CEN8 copy number or MYC translocation positive) MYC (B, D) and in the absence or presence of MYC colorimetric in situ hybridization signal clustering (A, C). MYC mRNA levels were obtained from the Affymetrix U133 Plus 2.0 Chip (Affymetrix, Santa Clara, CA). Lines represent the mean. The total MYC copy number was counted from 20 cells per case. The Mann-Whitney 2-tailed t test was used to determine significance when the 2 groups differed in variances; otherwise, an unpaired t test was used.

MYC Protein Expression Is Frequently Detected and Correlates With Increased MYC Copy Number, mRNA, and Poor Overall Survival

MYC IHC was performed using a new monoclonal antibody yielding crisp nuclear staining in 132 de novo DLBCL cases. Images 1H-1J show nuclear staining with IHC for MYC protein. In a normal tonsil control, MYC staining was in the expected distribution in the basal layer of the epidermis. Nondividing cells such as the upper stratified layers of the epithelium and lymphatic endothelium were completely negative, whereas scattered positive lymphocytes were noted in the interfollicular regions (Image 1H). Within the germinal centers, there were more frequent scattered positive lymphocytes, whereas the follicular dendritic cells and histiocytes were negative (Image 1I). In addition, comparison of the 2 different MYC Y69 clone antibody sources (Epitomics and Ventana Medical Systems) yielded identical results in an independent set of 33 DLBCL cases. In our DLBCL cases, 59 (44.7%) demonstrated at least 40% positive nuclei in a 300-cell count of the cytologically malignant-appearing cells (Image 1J). Of the 132 cases stained with MYC antibody, 112 also had successful CISH staining. Protein expression was most often seen in group 2 (20/46; 43.5% positive) and group 3 (24/43; 55.8% positive) with significantly (P = .024) less staining in group 1 (7/23; 30.4% positive) Figure 3A. Cases positive for MYC protein (>40% positive nuclei) exhibited significantly (P = .0445) higher MYC mRNA levels (mean, 8.6; range, 6.5-11.5) as determined by GEP in comparison to cases that were scored MYC protein negative (mean, 8.0; range, 5.9-9.3) Figure 3B. Of interest, in IHC-negative cases in group 2, the average percentage of positive nuclei was 25.9%, whereas in group 3, the average was 28.3%. In contrast, there was a comparatively higher average of 35.6% nuclei positive by MYC IHC in the cases with MYC signal clusters compared with IHC-negative, cluster-negative cases (27.7%; P = .10).

Figure 3.

MYC protein expression in cases that differ in MYC and chromosome 8 copy number and correlated to MYC messenger RNA (mRNA) and patient overall survival. The percent MYC protein positive cells as determined by immunohistochemistry was compared among the 3 different MYC and chromosome 8 copy number groups using a 1-way analysis of variance (A). MYC protein positive and negative cases were evaluated for level of MYC mRNA (log2 transformed) (B). Lines represent the mean. A case was defined as positive for MYC protein if at least 40% of cells were stained within a 300-cell count. The Mann-Whitney 2-tailed t test was used to determine significance.

The original MYC IHC scoring included faint, moderate, and strong staining patterns; therefore, we also analyzed the IHC data using just those cases with the strongest staining. Under this more stringent cutoff for MYC protein positivity, only 16 of the 132 cases were considered MYC strongly positive. These strongly positive cases still occurred more often in groups 2 (n = 5) and 3 (n = 8) but otherwise did not change results.

GCB- and ABC-like DLBCL Tumors Are Similar in MYC Copy Number, Clusters, and mRNA

The well-known biological and clinical differences between GCB-like and ABC-like DLBCL tumors prompted us to evaluate the prevalence of MYC clusters, copy number, and mRNA according to cell-of-origin subtype. In the 53 GCB and 51 ABC cases examined, there were no significant (P = .1032) differences in MYC signal clustering Figure 4A. However, there was a trend toward an increased number of cells with MYC clusters in the GCB cases (mean, 3.6 cells; range, 0-14) in comparison to the ABC cases (mean, 2.5 cells; range, 0-13). The total MYC copy number was similar (P = .6797) for both ABC (mean, 62.6; range, 35-100) and GCB (mean, 62.4; range, 35-96) tumor tissues Figure 4B. From the previous GEP data, there were mRNA expression values for 48 of the 53 GCB and 44 of the 51 ABC cases. A trend (P = .1135) was observed for higher MYC mRNA levels in the ABC (mean, 8.4; range, 7.3-10.3) tumor samples compared with the GCB (mean, 8.1; range, 5.9-11.5) samples Figure 4C. The lack of significant differences in MYC between the GCB and ABC DLBCL subtypes indicates that abnormalities of MYC may be a more global phenomenon in DLBCL and not restricted to particular cell-of-origin subtypes.

Figure 4.

MYC clustering, copy number, and messenger RNA (mRNA) levels in germinal center B-cell (GCB) and activated B-cell (ABC) diffuse large B-cell lymphoma (DLBCL) subtypes. The DLBCL cases were previously stratified into GCB-like or ABC-like subtypes based on gene expression profiling and analyzed for MYC clustering (A), total MYC copy number (B), and MYC mRNA levels (C). Lines represent the mean. The Mann-Whitney 2-tailed t test was used to determine significance when the 2 groups differed in variances; otherwise, an unpaired t test was used.

MYC mRNA and Protein Expression Significantly Correlate With Overall Survival

As previously reported in the literature, our 51 ABC cases had poorer overall survival as compared with the 53 GCB cases (P = .0173; data not shown).^15,16 MYC mRNA expression divided at the 80th percentile (as per our previous mRNA variable cut-point analysis work) significantly correlated with poorer overall survival (P = .0095; hazard ratio [HR], 0.5042; 95% CI, 0.2226-0.8117) Figure 5A. This survival analysis was performed on the larger de novo DLBCL cohort of 200 cases that included our CISH-stained cases.⁷ A subset analysis within group 2 (MYC amplified cases) of those cases with clusters of MYC signals demonstrated a slight trend (P = .1701; HR, 0.5467; 95% CI, 0.2312-1.295) but no significant difference in overall survival compared with cases with no clustering of MYC Figure 5B. To determine the effects of MYC protein expression on overall patient survival, we used the same cut-point as established for mRNA, the 80th percentile of the percent positive nuclei detected by IHC.^11,12 There were long-term follow-up and survival data after therapy with R-CHOP for 113 of the 132 patients successfully stained using MYC IHC. Patients with their percentage of positive nuclei above the 80th percentile (n = 21) had a significantly (P = .0029; HR, 0.3646; 95% CI, 0.0905-0.6100) worse overall survival outcome with respect to patients with positive nuclei staining that fell into the lower 80th percentile (n = 92) Figure 5C. However, when the 113 patients were dichotomized into positive (n = 52) vs negative (n = 61) MYC IHC defined at the ≤40% cutoff, there was no direct correlation between positive IHC protein staining and survival (data not shown). Although we observed no direct correlation between MYC clusters and worse overall patient survival, the presence of clusters did correlate with increased copy number as well as any MYC abnormality (amplification, polysomy, or translocation), which in turn correlated with increased mRNA. Both increased MYC mRNA and protein levels correlated with poorer overall survival.

Figure 5.

Diffuse large B-cell lymphoma (DLBCL) patient survival following R-CHOP (rituximab-cyclophosphamide, doxorubicin, vincristine, and prednisone). Kaplan-Meier overall survival curves of DLBCL patients according to MYC messenger RNA levels as detected by the Affymetrix U133 Plus 2.0 DNA microarray (Affymetrix, Santa Clara, CA) for the entire cohort (A), MYC colorimetric in situ hybridization signal clustering for group 2 cases (B), and MYC protein of patients dichotomized based on the 80th percentile (C). A, P = .0095; hazard ratio (HR), 0.5042; 95% CI, 0.2226-0.8117. B, P = .1701; HR, 0.5467; 95% CI, 0.2312-1.295. C, P = .0029; HR, 0.3646; 95% CI, 0.09051-0.6100.

Discussion

In this study, we describe a frequent pattern of MYC signals detected with a novel CISH assay that consisted of multiple, tightly packed signals that we designated as “clusters.” This signal clustering phenomenon has not been described previously in the literature, and thus the clinical significance of this pattern is unknown. The clustering of signals appeared to represent a molecular phenomenon rather than an artifact of the stain since clusters of MYC signals were never identified in normal tissues or in normal bystander cells within lymphoma tissues. Furthermore, cases containing clusters significantly fell into group 2 (increased MYC with normal CEN8). The presence of MYC clusters also significantly correlated with abnormal MYC locus by any mechanism (amplification, polysomy 8, or translocation). An abnormal MYC locus correlated with increased MYC copy number. Increased MYC copy number correlated well with protein expression, which, in turn, correlated with mRNA and poor outcome. However, a direct correlation between clusters and outcome could not be demonstrated. It was difficult to identify individual MYC signals in these clusters; therefore, it is reasonable to assume that many more copies of MYC are present in these cases than we have estimated here. These clusters may in fact have similarities to “homogeneous staining regions” or “double minutes” noted by FISH evaluation of MYC in leukemia and carcinoma.^17-19 Similar to our findings, other investigators have recently reported finding extra copies/amplification of MYC in 20 of 33 (61%) of DLBCL cases using FISH methodology.²⁰ Another group of investigators recently found that there was strong expression of MYC protein by IHC more frequently than can be accounted for by translocations, concluding that mechanisms other than MYC translocation must be responsible for MYC protein expression in a large fraction of cases.²¹ In the current study, there was a trend toward a comparatively higher average percentage of nuclei positive by MYC IHC in the IHC-negative cases with MYC signal clusters compared with no clusters, furthering the point that the clustering of the MYC signal likely has biological significance. MYC amplification may be one of a number of mechanisms resulting in dysregulated MYC and an underdetected phenomenon with clinicopathologic implications. Further studies using 3-dimensional image analysis techniques to more accurately classify MYC copy number may help to further investigate this issue.

Currently, karyotype analysis and FISH are the routine methods for detection of MYC chromosomal abnormalities. Karyotyping is a complex process that requires fresh tissue and technical expertise to perform and is labor intensive. FISH has many advantages over karyotyping, including easier assessment of gene copy number and determining the presence of the translocation as well as the ability to use FFPE tissue. However, the rapid signal decay, requirement of a fluorescent microscope, and the relative inability to visualize tissue architecture limit the usefulness of FISH as a diagnostic tool. To overcome these disadvantages, the CISH technique was developed, which provides a direct, permanent visualization of chromosomal staining concurrently with tumor morphology using bright-field microscopy. This particular probe was designed to detect amplification and not translocation. We found that the CISH probe performed well, was easy to evaluate, and may provide additional information not readily detectable by other methods. This CISH technique was also recently applied to HER2/neu detection in breast cancer, ALK translocations in lung cancer, and MALT1 translocations in MALT lymphomas.^22-25 CISH is analogous to FISH in that it uses a probe for a specific region of a chromosome and maintains tissue architecture. However, the probe is linked to a chromogen or colormetric signal, which conveys a number of advantages over FISH. These advantages include the ability to use a light microscope, a signal that has a long half-life, ability to evaluate tissue morphology, use of routinely processed FFPE tissue sections, and long-term archival storage. Interobserver comparisons between FISH and CISH and intraobserver comparisons using CISH have been reported in excess of 95%.²² Furthermore, we speculate that the smaller discrete colorimetric signals resulting from CISH may allow resolution of clusters of gene signals not easily identified with fluorescent signals. Finally, given the variability between individual cells within the same tumor, extra copies may not be observable by array comparative genomic hybridization or other techniques that rely on homogenization of tissues, particularly in the face of tumor heterogeneity.

The new MYC antibody performed as expected, yielding a nuclear staining pattern in highly proliferative cells. The protein expression correlated with mRNA levels and was present at a higher rate in cases with excess MYC copies (groups 2 and 3) and also correlated with worse overall survival. Other recent reports have described similar staining patterns and outcome correlations, indicating that the MYC IHC test may be helpful in identifying MYC genetic abnormalities.^10,26

The presence of MYC translocations in the context of abnormalities in BLC2 and/or BCL6 (so-called double hits or triple hits) has been associated with more aggressive disease. Currently, for new DLBCL cases at the University of Arizona, complete karyotyping is performed if sufficient fresh tissue is available or, alternatively, FISH assays for MYC, BCL2, and BCL6. Given the current work and our previous publications, we suggest that MYC IHC may be a suitable screening tool for MYC abnormalities and therefore may be used in conjunction with BCL2 IHC to detect cases with the poorest outcome.¹⁰ MYC CISH will need further analysis and experience to become routinely useful, but it has potential for widespread application as a substitute for MYC FISH.

In this work, we demonstrate for the first time the frequency of MYC clusters, which correlate with high numbers of MYC signals and in turn correlate with abnormal MYC locus and mRNA. We found, as have others, that both increased mRNA and protein correlated with patient outcome. Amplification of MYC signals may be an underreported but clinically pertinent feature of DLBCL with potential prognostic implications. In the current era of targeted therapeutics including strategies to interrupt MYC signaling, the careful evaluation of MYC status prior to therapy may become more important.