Reduced lymphotoxin-beta production by tumour cells is associated with loss of follicular dendritic cell phenotype and diffuse growth in follicular lymphoma.

Cytokine production is essential for follicular dendritic cell (FDC) maintenance and organization of germinal centres. In follicular lymphoma, FDCs are often disarrayed and may lack antigens indicative of terminal differentiation. We investigated the in situ distribution of cells producing lymphotoxin-beta (LTB), lymphotoxin-alpha (LTA), and tumour necrosis factor-alpha (TNFA) transcripts in human reactive lymph nodes and in follicular lymphomas with follicular or diffuse growth pattern. LTB was the cytokine most abundantly produced in germinal centres. LTB was present in nearly 90% of germinal centre cells whereas LTA and TNFA were detected in 30 and 50%, respectively. Moreover, the amount of LTB expressed in reactive germinal centre cells was 80-fold higher than that of LTA and 20-fold higher than that of TNFA. LTB-positive cells were more numerous in the germinal centre dark zone, whereas expression of the FDC proteins CD21, CD23, VCAM, and CXCL13 was more intense in the light zone. Tumour cells of follicular lymphomas produced less LTB than reactive germinal centre cells. The results of the in situ study were confirmed by RT-PCR; LTB was significantly more abundant in reactive lymph nodes than in follicular lymphoma, with the lowest values detected in predominantly diffuse follicular lymphoma. In neoplastic follicles, low production of LTB by tumour B cells was associated with weaker expression of CD21+/CD23+ by FDCs. Our findings detail for the first time the distribution of LTA-, LTB-, and TNFA-producing cells in human reactive germinal centres and in follicular lymphoma. They suggest the possibility that impaired tumour-cell LTB production may represent a determinant of FDC phenotype loss and for defective follicular organization in follicular lymphoma.

Introduction

Organization of B cell follicles requires a mutually dependent collaboration of B cells and follicular dendritic cells (FDCs). While FDCs provide signals to sequester and maintain B cells within B cell follicles (CXCL13), B cells are essential for FDC maintenance by providing stimulation with tumour necrosis factor‐alpha (TNFA) and lymphotoxin (LT) 1. Mature FDCs derive from perivascular mural cells expressing platelet‐derived growth factor receptor‐beta and alpha smooth muscle actin. Perivascular mural cells also give rise to fibroblastic reticular cells (FRCs) and marginal reticular cells (MRCs) 2. FDCs, FRCs, and MRCs have distinct morphologies and functions, but share common markers, and are probably strongly correlated 3.

Receptors for LT and TNF (LTβR and TNFR1) are highly expressed on FDC‐precursors. Mice deficient in LTβR, TNFR1, or their ligands suffer from complex pathological phenotypes of lymphoid organs which may be devoid of FDCs 4, 5, 6, 7, 8, 9, 10, 11. It is well‐established that LT and/or TNFα play a crucial role for maintenance of most FDC traits 12, 13; they consist of CXCL13 production 14, 15, 16, expression of ICAM1, VCAM1, and MadCAM1 17, 18, expression of complement receptors 1 and 2 (CR1 and CR2), and expression of Fc receptors for IgG, IgE, IgA, and IgM 18. Inhibition of LT leads to the disappearance of multiple markers on FDCs. Inhibition of the TNFα pathway is also effective, but only in the absence of a strong antigenic response.

Most of the information concerning interactions between cytokines and FDCs were obtained in murine models or in in vitro studies. Until recently, visualization of cytokine‐producing cells in tissue sections was extremely difficult. The development of RNA in situ hybridization (ISH) with the RNAscope technology has provided a major advance 19. In fact, this technology is highly specific, and allows identification of cytokine‐producing cells in tissue sections; moreover, the number of cytoplasmic dots per cell represents an approximate quantitative indication of the amount of cytokine RNA.

In the present study, we have investigated the tissue distribution of cells producing lymphotoxin‐alpha (LTA), lymphotoxin‐beta (LTB), and TNFA RNA in human reactive B cell follicles and in follicular B cell lymphomas (FL). Cytokine production was compared with expression of molecules indicative of FDC differentiation (CD21, CD23, VCAM, and CXCL13). Our findings indicate that there is a strict correlation between LTB production and FDC differentiation in reactive follicles and also in FL.

Materials and methods

Patients

Twenty‐six lymph nodes, removed for diagnostic purpose at the Sant'Andrea Hospital of Rome, were investigated. Eleven cases (8M:3F; mean age= 58 years) were diagnosed as reactive lymphadenitis (RL) with follicular hyperplasia. Lymph node site was: cervical (two), axillary (four), mediastinal (three), inguinal (one), and supraclavicular (one); the mean size of the lymph nodes was 1.75 cm. Eight cases were diagnosed as follicular lymphoma with predominantly follicular growth pattern (5M:3F), age‐range 51–82 years (mean age = 66 years), size range 1.8–3.8 cm (mean size = 2.6 cm). Grading: G1/G2 (n = 8). Lymph node site: inguinal (n = 4), axillary (n = 2), mediastinal (n = 1), and cervical (n = 1). Seven cases were classified as predominantly diffuse FL (3M:4F) 49–68 years (mean age = 58 years), lymph node size 2.5–4.5 cm (mean = 3.43 cm). Grading: G1/G2 (n = 7). Lymph node site: inguinal (n = 6), cervical (n = 1). The study was performed in accordance with the Helsinki Declaration. Institutional Review Board approval was obtained (EC n° 168/SA/2003).

Tissue samples and immunostaining

Lymph nodes were formalin‐fixed and paraffin‐embedded (FFPE). Paraffin sections were immunostained for CD21 (clone 1F8), CD23 (clone MHM6), Bcl‐2 (clone 124), CD10 (clone 56C6), Ki‐67 (clone Mib‐1) (Dako, Denmark), Stathmin (clone SP49; Spring Bioscience, Pleasanton, CA USA), VCAM (Clone VCAM1/843; Scytek Laboratories, Logan, UT, USA), and CXCL13 (Polyclonal Goat; R&D Systems, Minneapolis, MA, USA), using an automated immunostainer (Dako). The staining of follicular stroma and interfollicular fibroblastic stroma was graded quantitatively as previously described 20: 0, absent; 1+ focal; 2+ extensive; 3+ diffuse.

Cytokine production

Cytokine mRNA production was investigated using two different techniques: RNA ISH with RNAscope technology (Advanced Cell Diagnostics, Milano, Italy), and real‐time PCR (RT‐PCR). RNAscope is an in situ enzymatic technique that also provides quantitative information; in fact, the number of dots per cell is directly proportional to the number of specific RNA molecules.

RNAscope

The RNAscope assay was applied to tissue paraffin sections using probes for LTA, LTB, and TNFA, as previously described 19. In brief, FFPE tissue sections 2 μm thick were deparaffinized in xylene and then hydrated in an ethanol series. Hybridization was with target probes: Probe‐Hs‐LTΑ, Probe‐Hs‐LTΒ, and Probe‐Hs‐TNFA. The preamplifier, amplifier, label probe, and chromogenic detection procedures were performed according to the manufacturer's instructions (RNAscope 2.0 HD Reagent Kit, Advanced Cell Diagnostics, Hayward, CA, USA).

RNAscope‐stained tissue sections were digitalized at ×40 magnification using Aperio Scan Scope. Digital slides were used for determining percentage of positive cells and pixel of reactivity. For each lymph node, five different areas, measuring 17 000 μm2 each, were selected within different regions (i.e. GC, mantle, interfollicular). The percentage of positive cells per area was then calculated by manually counting the total number of cells and the number of stained cells directly on the screen. The number of positive pixels was determined using the Aperio software Positive Pixel Count v9 Algorithm.

RNA isolation and real‐time PCR

Total RNA was extracted from paraffin sections of 25 lymph nodes (11 reactive, 8 nodular follicular lymphoma, and 6 diffuse follicular lymphoma) using High Pure miRNA Isolation kit (Roche Diagnostics, Monza, Italy). The quantity and quality of the RNA was determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Waltham, MA USA). One microgram of purified RNA was reverse transcribed into cDNA using iScript Select cDNA Synthesis Kit (BIO‐RAD, Milan, Italy) according to the manufacturer's protocol. For cytokine mRNA expression analysis, real‐time PCR was performed using QuantiFast SYBR Green PCR Kit (Qiagen, Hilden, Germany) and the expression levels of cytokine were normalized to the housekeeping gene β‐actin. Real‐time PCR was performed on a 7500 Fast Real‐Time PCR System (AB). Primer sequences used for real‐time PCR analysis were: LTB fwd (5′‐ GAG GAC TGG TAA CGG AGA CG −3′); LTB rev (5′‐GGG CTG AGA TCT GTT TCT GG‐ 3′); ACTB fwd (5′‐CGG TTC CGC TGC CCT GAG‐3′); ACTB rev (5′‐TGG AGT TGA AGG TAG TTT CGT GGA T‐3′). Results of RT‐PCR performed in FFPE tissue were expressed as relative levels of LTB mRNA in reactive and neoplastic lymph nodes with reference to LTB mRNA of a reactive lymph node (RL1) that was chosen to represent 1x expression. Experiments were performed in triplicate.

Combined ISH and immunohistochemistry

RNAscope assay for LTΒ RNA was carried out on a FFPE section from a reactive lymph node as described above. After hybridization slides were immunostained for CD79a (clone JCB117, Dako) using Envision G/2 System/AP, Rabbit/Mouse (Permanent Red) (Agilent, CA, United States) following the manufacturer's instructions. Slides were mounted using Glycergel Mounting Medium (Dako).

Fluorescence in situ hybridization (FISH) analyses

Interphase FISH was performed on 2 μm‐thick FFPE sections of lymph nodes with a diagnosis of predominantly diffuse follicular lymphoma using dual colour break‐apart BCL‐2 probe (Kreatech, Leica Biosystems, Italy), dual colour LSI 1p36/LSI 1q25 probe (Vysis, Abbott, Chicago, USA), and the pretreatment kit (Tissue digestion kit, Kreatech, Leica Biosystems) following the manufacturer's instructions. Sections were then viewed under a NIKON fluorescent microscope with appropriate filters (NIKON Instrument, Italy). The hybridisation signals for each probe were evaluated in at least 100 interphase nuclei.

1p36 loss of heterozygosity (LOH)

1p36 LOH was assessed using five dinucleotide repeats (D1S2734, D1S199, D15508, D1S243, and D1S468). One of the primers in each pair was fluorescently labelled. Data analysis was performed using GeneScan software on a genetic analyzer ABI3I00 (Applied Biosystems, Foster City, CA, USA). Samples were regarded as uninformative if the normal tissue was homozygous or if instability was present in neoplastic tissue. LOH of a locus was determined using the calculation of Ganzian et al 21. Ratios less than 0.6 or greater than 1.67 were regarded as loss of the major or minor allele, respectively.

Statistical analysis

The expression levels of LTB, LTA, and TNFA mRNA assessed by RNAscope or by RT‐PCR were compared among the samples using Student's t‐test.

Results

LTB RNA is produced in B cell follicles of reactive lymph nodes

Cytokines known to be active in FDC differentiation (LTA, LTB, and TNFA) were investigated in paraffin sections of reactive lymph nodes. Cells producing cytokine RNAs were demonstrated with the RNAscope technology. Our findings provide evidence that LTB RNA is present mostly in reactive B cell follicles (Figure 1). At that site, cells positive for LTA RNA and TNFA RNA were significantly less numerous and contained fewer dots of reactivity (Figure 2). A more accurate estimate of the number of cytokine‐positive cells was performed on digital slides (Table 1). It was found that almost 90% of GC cells were positive for LTB RNA, whereas cells producing LTA and/or TNFA RNA were significantly less numerous (30 and 50% reduction, respectively; p ≤ 0.001). This difference was more evident when cytokine RNAs were evaluated as pixels per area; in fact, LTB RNA produced by GC cells was 80‐fold higher than that of LTA and 20‐fold higher than that of TNFA (p < 0.001). These findings indicate that a large proportion of GC B cells are capable of producing LTA, LTB, and TNFA; nevertheless, the amount of LTB RNA produced by each single cell is much higher as compared with LTA and TNFA.

Figure 1

In situ hybridization for LTB RNA in a reactive lymph node using RNAscope technology. Reactivity, consisting of cytoplasmic brown dots, is present in the cytoplasm of RNA‐producing cells, and is directly proportional to the number of RNA molecules. LTB RNA was mainly associated with cortical B cell follicles. GCs and mantle zones were both stained with higher levels in GCs.

Figure 2

In situ hybridization for cytokines in a GC using RNAscope technology. LTB RNA (A), LTA RNA (B), and TNFA RNA (C) were all present, but LTB reactivity was much more pronounced. Large cells with clear cytoplasm and tingible bodies (GC macrophages) did not showed reactivity for TNFA RNA (C).

Table 1

LTA RNA, LTB RNA, and TNFA RNA in B cell follicles and in interfollicular areas of reactive lymph nodes

Case No.	LTA+ cells (%)	LTA+ pixels	LTB+ cells (%)	LTB+ pixels	TNFA+ cells (%)	TNFA+ pixels
Germinal centre
1	23	857 ± 455	83	123 621 ± 50 123	48	7071 ± 3129
2	33	1503 ± 613	93	137 052 ± 64 044	34	2871 ± 1171
3	37	2209 ± 1166	95	140 043 ± 32 107	66	7978 ± 1041
4	30	1849 ± 685	80	116 592 ± 62 562	55	8613 ± 3882
Mean± SD **	31 ± 6	1605 ± 576	88 ± 7	129 327 ± 11 094	51 ± 13	6633 ± 2587
Mantle zone
1	10	370 ± 376	90	120 847 ± 9834	52	4531 ± 1407
2	16	524 ± 144	90	116 931 ± 43 960	24	1007 ± 328
3	17	863 ± 479	72	51 598 ± 8175	54	4460 ± 1535
4	15	685 ± 333	50	35 149 ± 13 533	33	3227 ± 2394
Mean±SD **	15 ± 3	611 ± 212	76 ± 19	81 131 ± 44 142	41 ± 15	3306 ± 1646
Interfollicular area
1	14	356 ± 173	75	31 710 ± 10 500	19	2212 ± 774
2	6	519 ± 738	51	17 563 ± 10 677	10	868 ± 269
3	7	945 ± 324	30	5256 ± 4431	13	1156 ± 306
4	13	801 ± 541	41	17 370 ± 12 551	29	4059 ± 3570
Mean±SD **	10 ± 4	655 ± 266	49 ± 19	17 975 ± 10 816	18 ± 8	2051 ± 1442

Statistical analysis (Student :

Germinal centre: LTB+ cells versus LTA+ cells p < 0.001; LTB+ pixels versus LTA+ pixels p < 0.001; LTB+ cells versus TNFA+ cells p = 0.001; LTB+ pixels versus TNFA+ pixels p < 0.001; LTA+ cells versus TNFA+ cells p = 0.017; LTA+ pixels versus TNFA+ pixels p = 0.005.

Mantle zone: LTB+ cells versus LTA+ cells p < 0.001; LTB+ pixels versus LTA+ pixels p = 0.005; LTB+ cells versus TNFA+ cells p = 0.01; LTB+ pixels versus TNFA+ pixels p = 0.006; LTA+ cells versus TNFA+ cells p = 0.006; LTA+ pixels versus TNFA+ pixels p = 0.009.

Interfollicular area: LTB+ cells versus LTA+ cells p = 0.004; LTB+ pixels versus LTA+ pixels p = 0.009; LTB+ cells versus TNFA+ cells p = 0.01; LTB+ pixels versus TNFA+ pixels p = 0.01; LTA+ cells versus TNFA+ cells p = 0.07; LTA+ pixels versus TNFA+ pixels p = 0.05.

Germinal Centre Mantle Zone: LTA+ cells p = 0.001; LTA+ pixels p = 0.009; LTB+ cells p = 0.137, LTB+ pixels p = 0.039; TNFA+ cells p = 0.176, TNFA+ pixels p = 0.037.

Germinal Centre Interfollicular Area: LTA+ cells p < 0.001; LTA+ pixels p = 0.012; LTB+ cells p = 0.005, LTB+ pixels p < 0.001; TNFA+ cells p = 0.003, TNFA+ pixels p = 0.01.

Mantle Zone versus Interfollicular Area: LTA+ cells p = 0.065; LTA+ pixels p = 0.40; LTB+ cells p = 0.05, LTB+ pixels p = 0.016; TNFA+ cells p = 0.017, TNFA+ pixels p = 0.15.

*Paraffin sections of four reactive lymph nodes were stained for LTA, LTB, TNFA RNA using RNA scope technology. Stained sections were digitalized using Aperio ScanScope. Percentage of stained cells and number of pixels (Aperio Positive Pixel Count v9 algorithm) were determined in areas measuring 17 000 μm2 each. The value reported in the Table is the mean ± SD of five different measurements made in different follicles or in different regions of the same lymph node.

**Mean ± SD of the four investigated cases.

Most RNA‐positive cells had the morphology of GC B cells; GC macrophages were negative for LTA, LTB, and TNFA. When GCs were compared with mantle zones, it was found that mantle cells were less efficient than GC cells in producing LTB RNA (Table 1). In fact, mantle cells stained for LTB RNA were less numerous (76 versus 88%; p = 0.137) and produced 40% less LTB (81 131 pixels versus 129 372 pixels p = 0.039) than GC cells. Mantle cells were also less effective in producing LTA RNA (611 pixels versus 1605 pixels p = 0.009) and TNFA RNA (3306 pixels versus 6633 pixels p = 0.037).

The presence of GC cells positive for LTB RNA correlated with expression of CD21, CD23, VCAM, and CXCL13 by FDCs (Table 2). As expected, FDCs were diffusely and intensely positive for CD21/CD23, and were focally positive for VCAM and CXCL13. The percentage of GC cells positive for LTB RNA varied from 55 to 95% (mean 80 ± 13), and the number of pixels per cell varied from 442 to 1000 (743 ± 231). An interesting observation, made possible by the in situ study, was that cells positive for LTB RNA were not distributed homogenously throughout the GC. In fact, cells positive for LTB RNA were more numerous in the basal dark zone of polarized GCs, where cell proliferation takes place (Figure 3). In contrast, expression of LTB induced molecules, such as CD21, CD23, VCAM, and CXCL13 was more prominent in the light zone, proximal to the sub‐capsular sinus (Figure 3).

Table 2

Immunophenotype of FDCs and LTB RNA in germinal centres of reactive lymph nodes

Case No.	Age/sex	FDC immunophenotype				LTB RNA
		CD21	CD23	VCAM	CXCL13	+ cells (%)	+ pixels (n)	pixels/cell (n)
1	77/M	3+°	3+	3+	1+	83 *	123 621 ± 50 123 *	902 **
2	43/M	3+	3+	1+	2+	78	61 841 ± 14 070	476
3	73/M	2+	2+	1+	1+	93	137 052 ± 64 044	1000
4	32/F	3+	3+	1+	1+	55	38 540 ± 24 996	602
5	81/M	3+	3+	2+	ND	ND	ND	ND
6	74/M	3+	2+	2+	1+	83	61 929 ± 22 160	442
7	63/F	3+	3+	1+	2+	ND	ND	ND
8	77/F	3+	3+	2+	2+	71	63 679 ± 16 271	624
9	71/M	3+	3+	2+	3+	ND	ND	ND
10	8/M	3+	3+	2+	2+	95	140 043 ± 32 107	909
11	34/M	3+	3+	2+	3+	80	116 592 ± 62 562	988
Mean±SD		2.9 ± 0.3	2.8 ± 0.4	1.7 ± 0.6	1.8 ± 0.8	80 ± 13	92 912 ± 40 373	743 ± 231

°The immunostaining of FDCs was graded as previously described [20]: absent (0); focal (1+); extensive (2+); diffuse (3+).

*Mean percentage of LTB+ cells and mean number of LTB+ pixels (Aperio Positive Pixel Count v9 algorithm) determined in five different GC areas measuring 17 000 μm2 each.

**Pixels per cell were calculated as total positive pixels/total number of stained cells.

Figure 3

Polarized germinal centre in a reactive lymph node. Cell proliferation (Ki67) and dots of LTB RNA (ISH with RNAscope technology) were more evident in the basal dark zone. Combined in situ hybridization for LTB RNA (brown) and immunohistochemistry for CD79a (red) showed that most LTB reactivity is associated with CD79a+ B cells of the mantle and GC dark zone. Very few LTB+ cells were present in the CD79a‐negative interfollicular area. Immunostainings for CD21, CD23, CXCL13, and VCAM‐1 were more intense in the upper part of the GC facing the sub‐capsular sinus.

Production of LTB RNA in T‐dependent interfollicular areas was significantly lower (p < 0.0001) as compared with B cell follicles (Table 1). At that site, LTB‐positive pixels (17 975 ± 10 816) were 19% of those present in GCs. In Figure 3, a lymph node section was double‐stained for LTB RNA (brown) and CD79a protein (red). It was confirmed that most LTB+ cells were of B cell origin (CD79a+), that they were polarized in GC, and that LTB+ cells were much less numerous in CD79a‐negative interfollicular T cell areas. Stromal reticular cells of interfollicular areas were negative for CD21/CD23, and were occasionally positive for VCAM/CXCL13.

LTB RNA production in B cell follicular lymphomas

Follicular B cell lymphomas are GC‐derived malignant tumours, which may exhibit a follicular or a diffuse pattern of growth. The so‐called ‘classical‐type’ FL carries the t(14;18) translocation, expresses BCL2 protein in most cases, and generally has a predominantly follicular pattern of growth. Classical FL with a diffuse pattern of growth is extremely rare. More recently, a distinct subtype of FL has been described, the so‐called ‘inguinal‐type’ 22. The latter is characterized by a predominantly diffuse pattern of growth, is usually BCL2‐negative, and often carries the 1p36 deletion. In Table 3, we have investigated LTB RNA in follicular versus diffuse FL. It was found that tumour cells of predominantly follicular FL were less effective in producing LTB RNA than GC cells, but were more efficient than tumour cells of predominantly diffuse FL. In fact, the percentage of LTB+ cells in predominantly follicular FL (62%) and the number of LTB+ pixels were significantly lower than those of reactive GCs (p = 0.004 and p = 0.001). In predominantly diffuse FL, the levels of LTB RNA were the lowest; in fact, LTB+ cells were 14% and LTB+ pixels were 3.540 (96% less than GC and 91% less than predominantly follicular FL; p < 0.001) (Figure 4). In our series, no correlation was found between LTB production and tumour grade (G1/G2) or cell proliferation (% of Ki67+ cells).

Table 3

LTB RNA and FDCs in predominantly follicular and in predominantly diffuse follicular B cell lymphoma*

Case N.	Age/sex	Lymph node site	Size (cm)	Grade	Ki67 (%)	BCL2#	1p36 loss°	LTB+ cells (%)	LTB+ pixels	LTB pixels/cell
Predominantly follicular FL
1	51/F	Inguinal	1.8	G2	10	+	ND	54	21 886 ± 11 138	257
2	53/M	Axillary	2.5	G2	40	+	ND	64	33 159 ± 22 146	299
3	67/F	Inguinal	2.5	G1	20	+	ND	75	60 813 ± 13 500	760
4	80/M	Inguinal	3.8	G2	30	‐	ND	68	62 476 ± 40 999	625
5	54/M	Cervical	3.0	G2	20	+	ND	40	33 437 ± 9052	423
6	82/M	Mediastinal	2.2	G1	<10	+	ND	68	35 017 ± 11 228	294
7	68/F	Axillary	2.5	G2	10	+	ND	65	28 890 ± 14 113	251
8	70/M	Inguinal	2.5	G2	20	+	ND	58	30 260 ± 8256	309
Mean ± SD								62 ± 11	38 242 ± 14 999	402 ± 190
Predominantly diffuse FL
1	53/M	Inguinal	3.0	G1	<10	–	+	27	8044 ± 5832	223
2	68/F	Inguinal	3.5	G2	<10	–	+	5	1512 ± 814	252
3	64/M	Inguinal	3.5	G2	20	–	+	5	730 ± 370	146
4	56/M	Cervical	2.5	G2	30	+	ND	5	750 ± 250	125
5	49/F	Inguinal	2.5	G2	10	–	‐	33	9072 ± 5412	171
6	51/F	Inguinal	4.5	G2	30	–	+	9	2086 ± 840	116
7	68/F	Inguinal	4.5	G2	<10	–	+	13	2587 ± 456	108
Mean ± SD								14 ± 12	3540 ± 3505	163 ± 56
Reactive lymph nodes (n = 8) ∧
Mean ± SD								80 ± 13	92 912 ± 40 373	743 ± 231

Statistical analysis (Student t‐Test): RL follicular FL: LTB+ cells p = 0.004; LTB+ pixels p = 0.001; LTB+ pixels/cell p = 0.003; RL diffuse FL: LTB+ cells p < 0.001; LTB+ pixels p < 0.001; LTB+ pixels/cell p < 0.001; follicular FL diffuse FL: LTB+ cells p < 0.001; LTB+ pixels p < 0.001; LTB+ pixels/cell p = 0.003.

*Paraffin sections were stained for LTB RNA using RNA scope technology. Stained sections were digitalized using Aperio ScanScope. The percentage of LTB+ cells and the mean number of LTB pixels were determined in five squared areas measuring 17 000 μm2 each. Pixels per cell were calculated as total positive pixels/total positive number of stained cells.

#BCL2 expression was investigated by immunohistochemistry, using Dako clone 124. BCL2‐negative cases were re‐investigated by FISH which confirmed the absence of t(14;18).

°Chromosomal 1p36 loss was investigated by FISH and by LOH analyses.

∧Cases are detailed in Table 2.

Figure 4

In situ hybridization for LTB RNA using RNAscope technology in reactive GC, predominantly follicular FL, and predominantly diffuse FL. In reactive follicles, most GC B cells and mantle zone cells are positive for LTB RNA. In follicular FL, most tumour cells are positive. In diffuse FL, only rare cells are stained.

Predominantly diffuse FL may still have limited follicular areas. Production of cytokine RNAs was investigated in follicular and diffuse areas of four cases of FL ‘inguinal‐type’ (Table 4). It was found that production of LTB RNA was only three‐fold higher in follicles as compared to diffuse areas. It is of interest that the small tumour‐associated follicles were positive for CD10, BCL‐6, and stathmin, but were negative or weakly positive for CD21 and CD23 (Figure 5). These findings seem to indicate that terminal differentiation of FDCs in the follicles of diffuse FL is to some extent impaired.

Table 4

LTB RNA and CD21/CD23 expression in follicular and diffuse areas of follicular lymphoma

Case No.	CD21	CD23	LTB+ pixels	CD21	CD23	LTB+ pixels
	Predominantly follicular FL
	Follicles			Interfollicular areas
1	1+	2+	21 886 ± 11 138 **	0	0	22 342 ± 8276 **
2	3+	2+	33 159 ± 22 146	0	0	2207 ± 1170
3	3+	NE	60 813 ± 13 500	0	NE	7096 ± 6411
4	3+	0	62 476 ± 40 999	NE	1+	17 159 ± 4409
5	2+	NE	33 437 ± 9052	0	NE	16 017 ± 3563
6	3+	1+	35 017 ± 11 228	1+	1+	2748 ± 1543
7	1+	0	28 890 ± 14 113	0	1+	11 305 ± 2497
8	3+	2+	30 260 ± 8256	0	0	5148 ± 2012
Mean ± SD	2.37 ± 0.92	1.17 ± 0.98	38 242 ± 14 999	0.14 ± 0.38	0.50 ± 0.55	10 503 ± 7414
	Predominantly diffuse FL
	Follicles			Diffuse areas
1	1+	0	26 539 ± 7794 **	1+	NE	8044 ± 5832 **
2	1+	1+	7894 ± 14 189	0	0	1512 ± 814
3	0	0	2100 ± 1374	0	1+	730 ± 370
4	1+	0	ND	1+	0	750 ± 250°
5	0	0	25157 ± 3924	0	NE	9072 ± 5412
6	1+	0	ND	1+	NE	2086 ± 840°
7	0	NE	ND	0	NE	2587 ± 456°
Mean ± SD	0.57 ± 0.53	0.17 ± 0.41	15 422 ± 12 281	0.43 ± 0.53	0.33 ± 0.58	3540 ± 3505

Statistical analysis (Student :

Follicles versus interfollicular area of predominantly follicular FL: CD21 p < 0.001; CD23 p = 0.09; LTB+ pixels p < 0.001; Follicles versus diffuse areas of predominantly diffuse FL: CD21 p = 0.31; CD23 p= 0.31; LTB+ pixels p = 0.08; Follicles of predominantly follicular FL versus follicles of predominantly diffuse FL: CD21 p < 0.001; CD23 p = 0.02; LTB+ pixels p = 0.01; Interfollicular area of predominantly follicular FL versus interfollicular area of predominantly diffuse FL: CD21 p = 0.13; CD23 p = 0.34; LTB+ pixels p = 0.02.

NE, not evaluable because CD21 or CD23 were expressed by the lymphoid component of the tumour as well; ND, not determined.

*Paraffin sections were immunostained for CD21 and CD23. The staining intensity of FDCs was graded quantitatively as previously described [20]: absent (0); focal (1+); extensive (2+); diffuse (3+).

**Paraffin sections were stained for LTB RNA using RNA scope technology. Stained sections were digitalized using Aperio ScanScope. The mean numbers of LTB+ pixels were determined in five squared areas measuring 17 000 μm2 each.

°These values were excluded when follicles and diffuse areas were compared in predominantly diffuse FL.

Figure 5

Predominantly diffuse CD23+ B cell FL. Small CD10+/Stathmin+ B cell follicles are poorly stained for CD21/CD23. A positive control is provided by CD23 staining of tumour cells.

The data obtained with RNAscope technology indicate that there are significant differences in the levels of LTB RNA produced in reactive GCs, predominantly follicular FL and in predominantly diffuse FL. This observation was confirmed by the use of RT‐PCR. cDNAs obtained from total RNA extracted from tissue sections of 11 reactive lymph nodes, 8 follicular FLs, and 6 diffuse FLs was tested for the presence of LTB by RT‐PCR. The data shown in Figure 6 confirm that LTB RNA is more abundant in reactive lymph nodes than in predominantly follicular FL (p = 0.006). The lowest values of LTB RNA were observed in predominantly diffuse FL (RL versus diffuse FL p< 0.001), thus confirming the data obtained with ISH.

Figure 6

LTB RNA expression determined by RT‐PCR. Total RNA was extracted from FFPE lymph nodes involved by RL (n = 11), predominantly follicular FL (n = 8), and predominantly diffuse FL (n = 6). Statistical analysis (Student's t‐test): RL versus follicular FL p = 0.006; RL versus diffuse FL p < 0.001; follicular FL versus diffuse FL p = 0.06.

Discussion

This is the first report describing the tissue distribution of cells producing LT and TNFA RNA in human lymphoid tissues. Our results show that cells producing LTB are mainly located in B cell follicles, and are more numerous than those producing LTA and TNFA. These observations are consistent with the notion that production of LTB RNA is constitutive in B‐cells 23, 24, whereas production of LTA and TNFA RNA is inducible 25. It has to be emphasized that quantitative differences in cytokine RNAs do not necessarily reflect protein synthesis and release. In fact, it has been reported that TNFA RNA accumulates rapidly, but has a brief half‐life. In contrast, LTA RNA accumulates more slowly, but persists much longer with a half‐life longer than that of TNFA RNA 26. Thus, RNA half‐time and protein translation represent further regulatory checkpoints which might profoundly alter the levels of cytokine production in lymphoid tissues.

We provide evidence that proliferating B cells located in the basal dark zone of GCs contain large amounts of LTB RNA, and that light zone FDCs express high levels of molecules involved in FDC function. Indeed, differences in FDCs populating dark and light zone have already been reported. It was described that FDCs in the dark zone were half as dense as FDCs in the light zone 27, and that dark zone FDCs produce large amounts of CXCL12 28. These findings and our observations raise the possibility that the topographical organization of GCs in dark and light zones also affects FDC differentiation and function. In fact, B cell proliferation in the dark zone represents a signal that GC cells are efficiently stimulated and that productive antigen presentation is occurring. In this situation, it is necessary to optimize FDC function through induction of molecules involved in antigen trapping (CD21, CD23), cell‐cell adhesion (VCAM), and B cell recruitment (CXCL13). Release of high levels of LTB by proliferating B cells of the dark zone might have this role. In agreement with this view, Mackay et al 7, 12 demonstrated that inhibition of LTA/B in tissue cultures caused disappearance of multiple markers on FDCs within 1 day, and that inhibition of the TNF pathway was much less effective. The possibility that TNF and LT exert different actions on stromal cells is supported by an in vitro study 29 showing that TNF alone was able to induce a strong increase of adhesion molecules, but not of meshwork formation, whereas LT had the opposite effect. Thus, it seems likely that LT and TNF are both necessary to support FDC function, but with different roles.

Tumour B cells of predominantly follicular FL show a strict topographical and functional relationship with FDCs 30; in fact, the latter are crucial for supporting tumour growth and survival 31, 32. Chang et al 20 investigated the immunophenotype of FDCs present in FL and found different patterns of antigen expression, depending on the pattern of growth (follicular versus diffuse). They reported that stromal reticular cells of diffuse FL showed only minimal immunophenotypic evidence of FDC‐like differentiation, and that FDCs of follicular FL were characterized by reduced expression of FDC antigens as compared to reactive GC. Similar findings were reported in a more recent study, where it was shown that FDCs found in different types of lymphoma show reduced expression of several FDC antigens as compared with normal GCs 33. We have confirmed and extended these observations. Moreover, the new information provided by our study is that there is a correlation between amounts of LTB RNA produced by tumour cells and levels of FDC differentiation of tumour‐associated stromal reticular cells.

A purely diffuse pattern is rare in FL. More recently, a peculiar type of diffuse FL was proposed as a separate entity 22, 34, 35. This tumour, known as ‘inguinal‐type’, has distinctive traits consisting of frequent involvement of inguinal lymph nodes, absence of BCL2 translocation, frequent occurrence of 1p36 deletion, frequent expression of CD23 by tumour B cells, and a better prognosis 22, 34, 35. Five of the seven cases of diffuse FL investigated in the present study had features of the ‘inguinal‐type’ variant. We have found that these tumours produce low amounts of LTB RNA as compared with ‘classical’ predominantly follicular FL. It can be speculated that poor production of LTB by tumour cells of predominantly diffuse FL is responsible for defective differentiation of stromal reticular cells into FDCs, and hence for defective nodular organization.

Author contributions statement

LR and ADN conceived experiments and designed the study. GP, CC, and SS carried out experiments. ADN, GP, LR, and EP analysed data. LR, ADN, and GP interpreted results and wrote the paper. All authors had final approval of the submitted and published versions.