TBLR1 and CREBBP as potential novel prognostic immunohistochemical biomarkers in diffuse large B-cell lymphoma

Abstract

Recent studies have identified prognostic mutational clusters for diffuse large B-cell lymphoma (DLBCL) patients, both within and outside the original cell-of-origin (COO) classification. For many of these mutations, there is limited information regarding the corresponding protein expression. With the aim to determine the relationship of protein expression and intensity to COO and prognosis, we used digital image analysis to quantitate immunohistochemical staining of CREBBP, IRF8, EZH2, and TBLR1 in 209 DLBCL patients. We found that patients with strong nuclear expression of TBLR1 had inferior progression-free survival (PFS) and overall survival (OS) in univariable analysis and inferior PFS in multivariable analysis. Patients with higher proportion of intermediate to strong nuclear CREBBP expression had a worse PFS and OS in univariable analysis. CREBBP was expressed with stronger intensity in non-GCB patients and the prognostic impact was restricted to this subgroup. These findings suggest that high nuclear protein expression of TBLR1 and CREBBP is negatively associated with prognosis in DLBCL.

Keywords:

• DLBCL
• CREBBP
• TBLR1
• prognosis
• digital image analysis

Introduction

Determining the cell-of-origin (COO) is now mandatory in the diagnostic workup of diffuse large B-cell lymphoma (DLBCL) [1]. Although the COO concept is well established, DLBCL has a heterogenic genomic profile with a number of recurrent mutations resulting in interindividual patient variation [2]. Recently published comprehensive genetic studies have also identified additional molecular clusters both within and outside of the COO classification, which could have prognostic significance [3–5]. Nevertheless, genomic changes can only affect a phenotype if they impact protein translation. For many of the gene mutations in DLBCL, there is no information about the corresponding protein expression, or its relation to prognosis. Immunohistochemistry (IHC) remains an important tool for the study of tumor biomarkers. However, consistent manual quantitative evaluation of membranous or cytoplasmic IHC staining is notoriously difficult. Instead, the use of digital software may offer a more objective intensity quantitation of IHC stainings and is increasingly being used for lymphomas, including DLBCL [6–11]. Furthermore, as cell nuclei are physically separated into distinct objects, quantitation of IHC staining of nuclear antigens may be more amenable to automated digital analysis.

Thus, we screened the recent genomic studies [3–5] for recurrent gene mutations where the potentially encoded protein had a predominant nuclear expression. The most important genes fulfilling this criterion are the epigenetic regulators, cAMP-response element binding protein (CREBBP) and enhancer of zeste homolog 2 (EZH2), the transcription factor interferon regulatory factor 8 (IRF8), the transcriptional regulator nuclear receptor corepressor 1 (NCOR1), and the F-box-like/WD repeat-containing protein (TBL1XR1) (for a full description of these genes, including possible prognostic impact, see Table 1).

Table 1. Overview of genes mutated in DLBCL, considered for an explorative study with immunohistochemistry.

Gene (protein)	Reddy et al. [3] COO	Prognostic information	Chapuy et al. [4] COO	Mol.	Prognostic information	Schmitz et al. [5]. COO	Mol.	Prognostic information
CREBBP (cAMP-response element binding protein)	GCB	Favorable^a	GCB	C3	Unfavorable	GCB	EZB	Favorable
EZH2 (enhancer of zeste homolog 2)	GCB	Favorable	GCB	C3	Unfavorable	GCB	EZB	Favorable
IRF8 (interferon regulatory factor 8)	GCB	–	GCB	C3	Unfavorable	GCB	EZB	Favorable
TBL1XR1 (F-box-like/WD repeat-containing protein TBL1XR1; TBLR1)	ABC	–	ABC	C5	Unfavorable	ABC	MCD	Unfavorable
NCoR1 (nuclear receptor corepressor 1)	GCB	Unfavorable	–		–	–		–

COO: cell of origin, in which the mutated gene was most commonly encountered; GCB: germinal center B-cell like; ABC: activated B-cell like; Mol.: molecular cluster within the study.

a In ABC patients.

For these genes, only a few and small studies using IHC to evaluate expression of the encoded proteins in R-CHOP treated DLBCL patients have been published. A correlation between elevated IHC expression of IRF8 in DLBCL patients (n = 67), as measured by IHC, and shorter survival has been suggested [12]. Deng et al. found that R-CHOP-treated patients (n = 61) with high EZH2 expression experienced a shorter progression-free survival (PFS) [13] but a most recent study by Neves Filho et al. found no association between EZH2 expression and clinical outcome (n = 139) [14]. To our knowledge, no IHC studies have been published that describe CREBBP, TBL1XR1, or NCoR1 expression in DLBCL patients.

Hence, with the aim to explore the relationship of protein expression and intensity to COO and prognosis, we performed IHC staining of CREBBP, IRF8, EZH2, TBLR1 (the protein encoded by TBL1XR1), and NCoR1 in samples from a large retrospective cohort (n = 209) of R-CHOP-treated DLBCL patients, using digital image analysis (DIA).

Materials and methods

Patients

Two-hundred and eight DLBCL patients treated with R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone), identified from the Swedish Lymphoma Registry as diagnosed in western Sweden 2006–2016 and with sufficient archived formalin-fixed, paraffin-embedded tissue material, were included. Primary mediastinal large B-cell lymphoma, primary CNS lymphoma, HIV-related lymphoma, EBV-positive lymphoma, and transformed lymphoma were excluded. Ethical approval was obtained from the Regional Ethics Review Board, Gothenburg, Sweden (no. 140-10, T-831-15 and 2019-01412).

Methods

TMA-blocks

Duplicate 0.5 mm cores from 74 patients were available in two TMA-blocks from a previous study [15]. For the remaining patients, duplicate 1 mm diameter cores were embedded in four additional TMA-blocks.

IHC stainings

Detailed information on the antibodies used for IHC is provided in Supplementary Table 1. Sections (4 µm) of the TMA blocks were stained for CREBBP, EZH2, IRF8, NCOR1, TBLR1, and MYC with immunohistochemical antibodies. In a minor proportion of the cases, routine diagnostic IHC was completed by staining for BCL2 and COO-classification according to Hans et al. [16] (BCL6, CD10, and MUM1). Antibodies for the determination of COO, MYC, and BCL2 were evaluated as per routine diagnostic workup. Cutoff values for BCL2 and MYC were ≥50% and ≥40% of the lymphoma cells, respectively [1,17] and patients expressing both BCL2 and MYC were designated as double expressers (DEs).

Digital image analysis of IHC stainings

Sections stained with CREBBP, EZH2, IRF8, NCOR1, and TBLR1 were scanned at ×40 magnification and digitalized with a NanoZoomer S210 (Hamamatsu Photonics, Hamamatsu City, Japan). DIA was performed using the Visiopharm 2019.02 software (Hörsholm, Denmark). Detailed information on the interpretation of digitalized IHC slides is included in the Supplementary information. In short, individual analysis protocol packages (APPs) were created for each antibody to analyze the proportion of positive nuclei and nuclear staining intensity (subclassified into four categories: negative, weak, intermediate, and strong).

The analyses were performed in regions of interest (ROI), containing a high proportion of lymphoma cells. For NCOR1, the internal variability between different TMA sections was too large to enable a reliable adjustment of the threshold values between different APPs, thus no further analyses were performed. IHC staining was interpretable for CREBBP, IRF8, EZH2, and TBLR1 for 190–195 patients. Individual staining was designated non-interpretable if a specific core biopsy in one or more sections was missing or if no staining was observed in one of the duplicate core biopsies with the other core exhibiting strong intensity.

Statistical analysis

Progression-free survival was defined as the time from diagnosis until the date of progression, relapse, death, or the last follow-up. Overall survival (OS) was defined as the time from diagnosis until the date of death or the last follow-up. For a more stringent prognostic evaluation of the immunohistochemical biomarkers, the survival analyses were performed with a maximum follow-up time set to 60 months. At this time, patients are considered to be clinically cured from their lymphoma with very few relapses and a majority of deaths from other causes may occur thereafter. The Kaplan–Meier method and log-rank test were used to calculate PFS and OS. The Cox proportional hazard regression model was used for uni- and multivariable analyses. The percentage of (a) strong and (b) strong plus intermediate nuclei staining for each of the immunohistochemical markers was first tested as a continuous variable in a Cox regression and categorization was performed only if a significant prognostic impact was found. Pearson’s chi-squared test and the Mann–Whitney U-test were used to compare the different clinical characteristics and IHC expression pattern between the COO patient groups. The statistical analyses were performed either by SPSS version 25 (SPSS Inc., Chicago, IL) or Stata for Macintosh, version 13.1 software (StataCorp, College Station, TX).

Results

Patients

Detailed patient characteristics are provided in Table 1. Briefly, the median age in the whole group was 67 years and the proportions of male patients, female patients, and COO groups were equal.

Immunohistochemistry

IHC expression patterns in all patients

BCL2 was positive in 74.0% of the cases, 16.8% were MYC-positive, and 12.0% of the cases were DE (Table 2). The median and mean percentage of nuclei within the staining intensity categories weak, intermediate, and strong as well as the median percentage of positive and negative nuclei were determined (Supplementary Table 2). For TBLR1, nearly all patients had positive nuclear staining but with different intensities, while CREBBP staining showed more variation in the percentage of positive nuclei and a very few patients had a strong nuclear staining intensity. Typical immunohistochemical staining for TBLR1 and CREBBP, with corresponding pseudo-colored digital software analyses images are shown in Figure 1(A–J).

Figure 1. Immunohistochemical staining of formalin-fixed, paraffin-embedded tumor tissues (magnification ×40) with TBLR1 and CREBBP antibodies with corresponding pseudo-colored images from the digital image analyses (blue: negative; green: positive with weak intensity; red: intermediate intensity; yellow: strong intensity). TBLR1: (A, B) low % of strong nuclear staining intensity, equal amounts of weak and intermediate intensity, (C, D) higher % of strong staining intensity with lower % of weak intensity and with intermediate intensity the same as A and B, (E, F) very high % of strong nuclear staining intensity and low % of intermediate and weak intensity. CREBBP: (G, H) low % of nuclei with intermediate/strong nuclei staining intensity, and (I, J) high % of intermediate/strong nuclear staining intensity.

Table 2. Characteristics of patients in this study.

Patients characteristics, N = 208	N (%)
Age (years)
Median	67
Range	28–87
Gender
Male	111 (53.4)
Female	97 (46.6)
aaIPI
0	49 (23.5)
1	67 (32.2)
2	64 (30.8)
3	26 (12.5)
Missing data	2 (1.0)
COO
GCB	103 (49.5)
Non-GCB	105 (50.5)
BCL2
≥50%	154 (74.0)
<50%	53 (25.5)
Missing data	1 (0.5)
MYC
≥40%	35 (16.8)
<40%	173 (83.2)
Double expresser
Yes	25 (12.0)
No	182 (87.5)
Missing data	1 (0.5)

aaIPI: age-adjusted international prognostic index; COO: cell-of-origin; GCB: germinal center B-cell like.

Different IHC expression patterns in GCB vs. non-GCB

For TBLR1, non-GCB patient samples exhibited a higher percentage of positive nuclei with strong intensity (Figure 2(A)) and fewer with weak intensity (Supplementary Table 2). For CREBBP, non-GCB patient samples exhibited a larger proportion of both positive nuclei, intermediate and strong staining intensity, yet the total percentage of strong nuclei staining was very low in both groups (Figure 2(B–D)). For IRF8, GCB cases had a larger proportion of positive nuclei than non-GCB cases.

Figure 2. The distribution and median values for: (A) TBLR1 – strong nuclei staining intensity, (B) CREBBP – all positive nuclei, (C) CREBBP – intermediate nuclei staining intensity, and (D) CREBBP – strong nuclei staining intensity, in GCB patients compared with non-GCB.

Prognostic analysis

The follow-up time varied from 1 to 166 months for the whole cohort and after the maximum follow-up time was set to 5 years, the median follow-up for living patients was 60 months. For CREBBP, intermediate to strong nuclei were evaluated together since the percentage of nuclei with strong staining intensity, was very low and the percentage of total positive nuclei was therefore also evaluated. For TBLR1, as the opposite was found, we chose to evaluate the percentage of nuclei with strong staining intensity.

Univariable analyses

Background variables

PFS and OS for the whole patient cohort at 3 and 5 years were 75% and 81%, and 70% and 76%, respectively. Both patients with a high age-adjusted International Prognostic index (aaIPI) score and those who were DE exhibited a lower PFS and OS: (p=.003, p=.010) and (p<.001, p=.001), respectively (Supplementary Figure 1(A–D)). No difference in PFS (p=.21) or in OS (p=.28) was observed between non-GCB and GCB patients.

Immunohistochemical markers

For TBLR1, a univariable Cox regression analysis showed a significant association with worse PFS and OS for patients with a higher percentage of nuclei with strong staining intensity (hazard ratio (HR) 3.19; 95% confidence interval (CI) 1.35–7.52, p=.008 and HR 2.94; 95% CI 1.14–7.56, p=.025, respectively). In addition, patients with an above median (28.3%) percentage of nuclei with strong TBLR1 intensity had an inferior PFS and OS (p=.006 and .049) (Figure 3(A,B)).

Figure 3. (A) Progression-free and (B) overall survival for patients with a percentage of nuclei staining with strong TBLR1 intensity above/under the median. (C) Progression-free and (D) overall survival for patients with a percentage of CREBBP nuclei staining with intermediate to strong intensity above/under the median. aaIPI: age-adjusted international prognostic index; COO: cell-of-origin; DE: double expresser.

For CREBBP, a higher percentage of positive cells was associated with inferior PFS (HR 3.86; 95% CI 1.06–14.0, p=.040) and with a trend for inferior OS (HR 3.51; 95% CI 0.85–14.4, p=.081). Furthermore, a higher percentage of nuclei with intermediate to strong staining intensity was associated with worse PFS and OS (HR 3.53; 95% CI 1.49–8.35, p=.004 and HR 3.52; 95% CI 1.36–9.14, p=.010). Patients with a percentage of intermediate to strong nuclei above the median (>23.0%) had an inferior PFS (p=.043) while there was no difference in OS (Figure 3(C,D)). For IRF8 or EZH2, we found no prognostic association.

Association with PFS and OS for IHC expression related to COO

For strong TBLR1 nuclei staining, a univariable Cox regression analysis showed a trend for an association with inferior PFS for both GCB and non-GCB patients (p=.090 and p=.073, respectively). GCB patients with a high percentage of TBLR1 strong nuclei intensity had a worse OS (p=.034) while this was not observed in non-GCB patients (p=.29).

There was a significant association of CREBBP expression with inferior PFS and OS for non-GCB patients with respect to the percentage of nuclei with intermediate to strong staining intensity (p=.004 and p=.001), while this was not observed in GCB patients (p=.42 and p=.94, respectively). In addition, non-GCB patients with a percentage of intermediate to strong CREBBP nuclei staining above the median had an inferior PFS (p=.051) and OS (p=.011) while no differences were observed in GCB patients (p=.51 and p=.72, respectively) (Figure 4(A–D)). Indeed, the non-GCB group patients with CREBBP intermediate to strong > median had 25 events (19 relapses, four PD and two deaths of other reasons) while patients with CREBBP intermediate to strong < median had seven events (three relapses and four PD). Causes of death in the non-GCB group were for patients with CREBBP intermediate to strong > median lymphoma-related in 21/23 patients, while for patients with CREBBP intermediate to strong < median only four patients died of lymphoma. We found no prognostic association between COO and EZH2 or IRF8.

Figure 4. (A, B) Progression-free and (C, D) overall survival for non-GCB and GCB patients, respectively, with a percentage of intermediate to strong CREBBP nuclei staining intensity above/under the median.

Multivariable analyses

For TBLR1, a percentage of nuclei with strong intensity above the median, high aaIPI and DE were all independently associated with a worse PFS, while there was no association with PFS for COO (Table 3). With respect to OS, DE remained an independent risk factor for inferior survival while TBLR1 and high aaIPI showed a border-significance for inferior survival.

Table 3. Multivariable Cox regression analysis for strong TBLR1 nuclei staining and intermediate to strong CREBBP nuclei staining.

	Progression-free survival		Overall survival
Variables	HR (95% CI)	p Value	HR (95% CI)	p Value
TBLR1 strong: >median vs. <median	1.96 (1.13–3.40)	.017	1.65 (0.91–3.01)	.10
aaIPI: 2–3 vs. 0–1	2.01 (1.17–3.43)	.011	1.79 (0.99–3.24)	.053
COO: non-GCB vs. GCB	1.06 (0.62–1.82)	.82	1.00 (0.55–1.82)	.99
DE: yes vs. no	2.43 (1.30–4.53)	.005	2.52 (1.27–4.97)	.008
CREBBP intermediate/strong: >median vs. <median	1.54 (0.89–2.67)	.13
aaIPI: 2–3 vs. 0–1	1.81 (1.06–3.10)	.029
COO: non-GCB vs. GCB	0.93 (0.54–1.60)	.80
DE: yes vs. no	2.71 (1.47–5.00)	.001

aaIPI: age-adjusted international prognostic index; COO: cell-of-origin; GCB: germinal center B-cell like; DE: double expresser.

Regarding CREBBP, both aaIPI and DE were independently associated with inferior PFS while CREBBP intermediate to strong nuclei staining intensity above the median or COO were not.

Discussion

A number of driver genes and prognostic mutations in DLBCL have been identified in recent years, but the expression of their encoded proteins and possible correlation with patient outcome, is largely unexplored. In this explorative study, we used a DIA approach, to show that the protein expression pattern of two genes commonly mutated in DLBCL (CREBBP and TBLR1) has prognostic impact. With respect to TBLR1, a higher percentage of strong nuclear staining was associated with worse PFS and OS, and the effect on PFS was also independent of other risk factors including aaIPI, COO, and DE status. For CREBBP, a high percentage of nuclei with intermediate/strong staining intensity was also associated with inferior PFS and OS, however, not independent of aaIPI and DE.

Interestingly, when comparing the outcome in the COO groups we found that the prognostic significance of CREBBP expression appeared to be limited to non-GCB patients. Mutations in the CREBBP gene, one of the most common genetic aberrations in DLBCL patients [18], occur more often in GCB patients [3,19]. This mutation mainly has an inactivating or silencing effect [20], which conceivably could lead to a lower protein expression of CREBBP. Similarly, we found that GCB patients had both a lower median percentage of total positive cells and cells with intermediate to strong CREBBP nuclear intensity compared with non-GCB patients. As ABC patients with a CREBBP mutation were found to have a better prognosis [3], this could imply that an ABC patient with such a mutation could exhibit lower CREBBP protein expression, which in our study was associated with better survival for non-GCB patients. Still, in the GCB patient group, there was no difference in PFS related to CREBBP nuclear intensity. CREBBP is a histone modifier that forms a complex with the histone acetyltransferase EP300, and functions as a transcriptional coactivator for a large number of DNA-binding transcription factors. It is involved in multiple signaling and developmental pathways, by modifying lysine residues on both histone and non-histone nuclear proteins [21,22]. Furthermore, CREBBP and EP300 enhance transcription through multiple mechanisms, including acetylation-mediated inactivation of transcriptional repressors, such as the DLBCL-associated BCL6 oncogene [23]. BCL6 also represses genes involved in terminal differentiation including IRF4 and PRDM1 [24]. The effect of IRF4 in ABC DLBCL is not exactly understood but it is considered to be an important survival factor for lymphoma cells [25]. Thus, mutational silencing of the CREBBP gene in ABC DLBCL, leading to reduced inhibition of BCL6 and subsequent interference with IRF4 could possibly be one explanation for the more favorable prognosis observed in ABC DLBCL. Furthermore, high CREBBP protein expression appears to be associated with worse prognosis or increased tumor aggressiveness in other tumor types. CREBBP overexpression has been associated with poor outcome in acute myeloid leukemia [26], small cell lung cancer [27] and it has also been associated with increased recurrence of bladder cancer [28]. On the other hand, a low CREBBP expression is associated with an adverse outcome in pediatric acute lymphoblastic leukemia [29] which together with our findings in the GCB patient group, indicate that CREBBP could have a context-related effect. However, CREBBP appears mainly to act as an oncoprotein and for the ABC DLBCL subtype such an effect could potentially translate high IHC nuclei expression of CREBBP into a potential negative prognostic biomarker.

Although non-GCB patients had a higher percentage of cells with a strong TBLR1 nuclear staining intensity, we found no differences between the COO groups regarding PFS or OS. TBLR1 is a subunit of the silencing mediator of retinoic acid and thyroid hormone receptors (SMRT)/NCoR/histone deacetylase 3 (HDAC3) complex, which functions as a transcriptional repressor while also activating NF-κB signaling [30,31]. In the study by Chapuy et al. [4], mutations in TBL1XR1 were found enriched in cluster 5, and were associated with an unfavorable outcome. These results are consistent with those in the study by Schmitz et al. [5]. Furthermore, up-regulated TBLR1, as measured by IHC, was associated with poor outcome in serous ovarian carcinoma [32], gastric cancer [33], and cervical cancer [34]. TBLR1 also promotes migration and invasion in ovarian and breast cancer cells [35]. ABC DLBCL mainly relies on chronic active B-cell signaling and activation of the NF-κB pathway [36,37]. The latter mechanism may explain the increased aggressiveness of the lymphoma, observed in non-GCB patients with high TBLR1 expression. Conversely, in GCB DLBCL, translocation of BCL2 and disrupted epigenetic regulation are important drivers [38,39], suggesting a different mechanism for the actions of TBLR1 on lymphoma behavior in these cases.

The use of DIA for intensity quantitation of immunohistochemical staining for nuclear proteins has been shown to be at least as reliable as manual quantitation [11,40]. Indeed, the strict consistency provided by the automated intensity quantitation appears much more difficult to obtain if performed manually. Moreover, technical variations in global staining results between the different TMA blocks could be adjusted by manually changing the software settings, thus lowering the risk of misinterpretation. For TBLR1 and CREBBP, a majority of cells were positive in nearly every case, though with varying intensities. Thus, it seems as if it is the staining intensity, rather than the percentage of positive nuclei, which is of prognostic importance.

There are admittedly some limitations in this study. First, the embedding of control tissue for all analyzed antibodies, into all six TMA-blocks, containing nuclei with all four categories of nuclear staining intensity, would have further strengthened the adjustment of the settings in the software applications used for the different TMA sections. Second, a parallel manual interpretation for the same staining intensity categories could have provided more insight on the usefulness of digital analysis for immunohistochemical staining. Third, validating the results in an independent cohort of R-CHOP treated DLBCL patients would also have been beneficial. Fourth, the use of DIA in routine diagnostics is, although increasing, still rather limited which presently hampers the use of a procedure like the one in this study for routine prognostic evaluation of diagnostic samples. Lastly, as our patient cohort appears to have a rather good overall prognosis, this could potentially disguise smaller differences between IHC patterns and outcome regarding EZH2 and IRF8.

In conclusion, we found that high nuclear expression of CREBBP and TBLR1 in DLBCL patients was associated with decreased survival. Although these findings need to be further evaluated in other cohorts, they also indicate that potential prognostic information from immunohistochemical staining patterns may be facilitated by DIA.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgements

The authors thank Shahin De Lara, Department of Pathology, Sahlgrenska University Hospital, for the performance of TMA sectioning, optimizing and performance of IHC stainings. Erik Holmberg for expert statistical assistance. Ylva Magnusson and Göran Landberg, Department of Pathology and Genetics, Sahlgrenska Academy, Gothenburg University and Kristina Lövgren, Department of Oncology and Pathology, Lund University Cancer Center, Lund University, for the production of TMA blocks. We thank Edanz Group for editing a draft of this manuscript.

Disclosure statement

P.O.A. joined the speakers’ bureau of Roche, Gilead, and Janssen and has been a consultant for Abbvie, Gilead, Janssen, and Roche. H.N.E. joined the speakers’ bureau of Roche and Janssen and has been a consultant for Roche. The other authors declare no conflicts of interest regarding the publication of this paper.

Data availability statement

Additional information

Funding

This study was supported by funding from Borås Cancer Foundation, from the Swedish State under the Agreement between the Swedish Government and the Country Councils; the ALF-agreement [79606], Lions Cancerfond Väst and the Swedish Cancer Society.