Genetic and phenotypic analysis of B-cell post-transplant lymphoproliferative disorders provides insights into disease biology

B-cell post-transplant lymphoproliferative disorders (PTLD) are classiﬁed as early lesions, polymorphic lymphomas (P-PTLD) and monomorphic lymphomas (M-PTLD). These morphologic categories are thought to reﬂect a biologic continuum, although supporting genetic data are lacking. To gain better insights into PTLD pathogenesis, we characterized the phenotypes, immunoglobulin (Ig) gene alterations and non-Ig gene ( BCL6 , RhoH/TTF , c-MYC , PAX5 , CIITA , BCL7A , PIM1 ) mutations of 21 PTLD, including an IM-like lesion, 8 P-PTLD and 12 M-PTLD. Gene expression proﬁle analysis was also performed in 12 cases. All PTLD with clonal Ig rearrangements showed evidence of germinal centre (GC) transit based on the analysis of Ig and BCL6 gene mutations, and 74% had a non-GC phenotype (BCL6 (cid:1) MUM1 þ ). Although surface Ig abnormalities were seen in 6/19 (32%) PTLD, only three showed ‘crippling’ Ig mutations indicating other etiologies for loss of the B-cell receptor. Aberrant somatic hypermutation (ASHM) was almost exclusively observed in M-PTLD (8/12 vs. 1/8 P-PTLD) and all three recurrent cases analysed showed additional mutations in genes targeted by ASHM. Gene expression analysis showed distinct clustering of PTLD compared to B-cell non-Hodgkin lymphomas (B-NHL) without segregation of P-PTLD from non-GC M-PTLD or EBV þ from EBV (cid:2) PTLD. The gene expression pattern of PTLD appeared more related to that of memory and activated B-cells. Together, our results suggest that PTLD represent a distinct type of B-NHL deriving from an antigen experienced B-cell, whose evolution is associated with accrual of genetic lesions. Copyright


Introduction
Post-transplant lymphoproliferative disorders (PTLD) encompass a heterogeneous group of lymphocytic proliferations, the vast majority of which are of B-cell origin.The current WHO classification categorizes them as early lesions, polymorphic lymphomas (P-PTLD) and monomorphic lymphomas (M-PTLD).Virtually, all early lesions are polyclonal and they appear to lack any known molecular alterations [1].Many P-PTLD and most M-PTLD are clonal [1,2].Compared to P-PTLD, M-PTLD harbour more karyotypic abnormalities, have a higher frequency of alterations of protooncogenes and carry mutations in a number of tumour-suppressor genes [1,3,4].Mutations in genes due to aberrant somatic hypermutation (ASHM), a recently identified B-cell specific mechanism of genetic instability, have been reported to occur exclusively in M-PTLD [5,6].The presence of distinct as well as shared cytogenetic aberrations among P-PTLD and M-PTLD [4] suggests that certain M-PTLD might arise as independent primary neoplasms, while others may represent progression of P-PTLD.The relationship of P-PTLD to M-PTLD, however, remains unknown.
Our understanding of the cellular origin of PTLD also remains unclear.A limited number of studies have shown that the vast majority of cases bear an imprint of germinal centre (GC) transit, as evidenced by the presence of somatically mutated immunoglobulin (Ig) genes, and most have non-GC immunophenotypes (BCL6 AE MUM1þ) [2,7,8].However, more sensitive approaches, such as gene expression profiling, which have significantly advanced our knowledge of B-cell non-Hodgkin lymphoma (B-NHL) biology, have not been systematically applied to the study of PTLD.
The pathogenesis of PTLD is commonly associated with Epstein-Barr virus (EBV) infection, although in more recent studies 23-42% of PTLD have been reported to be EBVÀ [2,9].It has been proposed that EBVÀ PTLD may be similar to B-NHL of immunocompetent hosts [10]; yet, a number of them respond to decreased immunosuppression, suggesting that immune deregulation plays a role in the pathogenesis of at least a subset of EBVÀ PTLD [9].A minority of PTLD also lack surface B-cell receptor (BCR) and it is speculated that EBV-encoded proteins can substitute for BCR signalling and survival of the abnormal clones [11].The relationship of EBVÀ PTLD to EBVþ PTLD or B-NHL occurring in immunocompetent individuals is not known at present.
To gain better insights into the pathogenesis of PTLD, we performed comprehensive phenotypic and genetic characterization of 21 PTLD.Gene expression profiles of a subset of cases were also analysed and compared to other types of B-NHL and cells representing normal B-cell maturational stages.

Case selection and morphologic analysis
We searched our departmental database to identify PTLD for which frozen tissue samples were available.Frozen sections were reviewed prior to and after obtaining tissue for molecular studies, to assess tumour content and ensure morphologic representation, and only cases with >90% lesional composition were used.Sections of formalin-fixed, paraffin-embedded tissue from the corresponding cases were stained with haematoxylin and eosin for morphologic review.PTLD were classified according to the current WHO criteria [12].This study was approved by the Institutional Review Board of Columbia University.

Immunohistochemistry, in situ hybridization and flow cytometry
Immunohistochemical (IHC) stains were performed on paraffin-embedded tissue sections after moist heat-induced antigen retrieval, using an autostainer (Universal staining system, DAKO, Carpinteria, CA) and the primary antibodies listed in Table 2.The Envision Plus kit (DAKO) was used for detection.Staining for BLIMP-1 and activation-induced deaminase (AID) was performed as previously described [13,14].
Cases were scored as positive if >20% cells expressed the analysed antigen, except for LMP1, EBNA2 and BZLF1, where percentage of cellular expression was recorded.
In situ hybridization for EBV-encoded small RNAs (EBER) was performed using the manufacturer's protocol (INFORM EBER, Ventana, Tucson, AZ).
Three or four colour flow cytometric analysis (FACScan; BD Biosciences, San Diego, CA) was performed using a comprehensive panel of antibodies against B-and T-cell antigens, including Ig light (Ig L ) and heavy chains (Ig H ), according to standard methods.Data were analysed using the Cell Quest software (BD Biosciences).

DNA and RNA extraction
Genomic DNA and total RNA were extracted from frozen tissue according to standard methods.In all PTLD and in selected matched non-neoplastic tissue specimens, DNA was also extracted from paraffin-embedded tissue using the QIAamp DNA mini Kit (QIAGEN, Valencia, CA).

Analysis of IgV H genes
PCR amplification of the Ig heavy chain variable region (V H ) genes was performed using DNA extracted from frozen tissue initially with IgV H gene family specific (V H [1][2][3][4][5][6] primers that hybridize to framework region (FR) 1 sequences and an antisense J H primer mix.If no clonal IgV H gene rearrangement was obtained, DNA was amplified with V H 1,3,4 primers that hybridize to the IgV H leader (L) region (sequences for all primers have been previously reported [15,16]).PCR products were purified and sequenced as described [17], and sequences were compared to the most homologous germline V H genes in the IMGT (https://imgt.cines.fr)and IgBLAST (http:// www.ncbi.nlm.nih.gov/igblast/)databases.The presence of IgV H rearrangements was also determined using FR3 primers and DNA extracted from paraffin-embedded tissue as previously described [18].
The ratio of replacement to silent mutations in the complementarity determining regions and FRs was analysed using the multinomial distribution model available at www-stat.stanford.edu/immunoglobulin[19].Determination of antigen selection was based on the analysis of FR mutations since interpretation of complementarity determining region mutations is controversial [20].
Analysis of BCL6, PIM1, PAX5, RhoH/TTF, c-MYC, CIITA and BCL7A The oligonucleotide sequences used for amplification of the BCL6, PIM1, PAX5, RhoH/TTF, c-MYC, CIITA and BCL7A 5 0 sequences and the genomic regions amplified are described in Supplemental Table 1.These regions correspond to the intervals previously shown to harbour >95% of all BCL6 mutations detected in B-NHL [21] and >90% of mutations affecting the remaining six genes in diffuse large B-cell lymphomas (DLBCL) [6] (L.P., unpublished data).Purified amplicons were sequenced directly as described [17] and compared to the corresponding germline sequences (Supplemental Table 1) as previously reported [6].Nucleotide changes due to previously reported polymorphisms or present in paired normal DNA (where available) were excluded from the calculation of mutation frequencies [6].

Generation of gene expression profiles and data analyses
Gene expression profile analysis was performed on 12 PTLD (1 IM-like lesion, 5 P-PTLD and 6 M-PTLD).A case each of IM-like lesion and M-PTLD of GC phenotype were included for analysis; the former, since cellular composition of the interfollicular areas in such lesions is morphologically indistinguishable from P-PTLD, and the latter, to determine any overriding effect of the post-transplant status (effect of immunosuppressive medi-cations on the lymphoproliferations or their microenvironment) on the pattern of segregation of PTLD.Suboptimal RNA quality and insufficient numbers of distinct morphologic types of M-PTLD (e.g.BL) precluded analysis of the remaining cases.Complementary RNA was prepared as previously described and hybridized to HG-U95Av2 microarrays (Affymetrix, Santa Clara, CA) [22].Gene expression values were determined by the GCOS 1.4 Microarray Suite 5.0 software (Affymetrix) using the Global Scaling option.In conjunction with previously generated expression profile data of other B-NHL subsets (derived from B-cells at different maturational stages), including EBVÀ DLBCL (n ¼ 22; 14 centroblastic and 8 immunoblastic), follicular lymphoma (FL, n ¼ 6), Burkitt lymphoma (BL, n ¼ 7; 1 EBVþ and 6 EBVÀ), and hairy cell leukaemia (HCL, n ¼ 4) [23,24], a hierarchical clustering algorithm based on the averagelinkage method was used to generate the dendrogram in Figure 1 [25].Expression values of the selected genes were normalized to have a zero mean value and unit standard deviation, and the distance between two individual samples was calculated by Pearson distance analysis.

Statistical analysis
Fisher's exact test and Student's t-test were performed to evaluate differences between datasets consisting of categorical and numerical variables, respectively, and a p-value < 0.05 was considered significant.Where applicable, the Spearman's rank correlation coefficient was also evaluated.

Morphologic and phenotypic analysis
PTLD cases were classified as IM-like lesion (n ¼ 1), P-PTLD (n ¼ 8) and M-PTLD (n ¼ 12).M-PTLD were further classified as DLBCL (n ¼ 10) and BL (n ¼ 2); the DLBCL were predominantly composed of centroblastlike cells.Morphologic features of the 12 PTLD used for gene expression profiling are shown in Supplemental Figure 1.
The majority of non-GC PTLD were EBVþ (9/14) and the majority of GC PTLD were EBVÀ (4/5); however, the correlation between phenotype and EBV infection did not reach statistical significance ( p ¼ 0.14).There was no significant difference in the mean time from transplant to initial presentation between EBVþ and EBVÀ PTLD (5.2 vs. 7.2 years, p ¼ 0.42).AID expression was observed more frequently in EBVÀ PTLD, although the difference was not statistically significant (7/9 vs. 3/10, p ¼ 0.07).
Potentially functional Ig genes were predicted in 13 (81%) of the 16 clonally rearranged PTLD (5/6 P-PTLD and 8/10 M-PTLD, Supplemental Table 2).The remaining three (19%) cases displayed an original in-frame rearrangement that had acquired crippling mutations: one of these, an EBVþ P-PTLD, had a 295 (bp) deletion involving FR1 and FR2; two other samples, representing clonally related EBVÀ M-PTLD from the same patient, had a 1 bp deletion in the FR3 region, resulting in a frameshift with premature stop codon.All these three cases showed absent sIg L expression (and sIg H in one case evaluated) by flow cytometry.Notably, virtually all mutated functional rearrangements (11/12) showed lower than expected number of replacement FR mutations, indicating selection due to antigen pressure.
No significant differences were observed between P-PTLD and M-PTLD with respect to total number of cases carrying BCL6 mutations (5/8, 63% vs. 9/12, 75%, p ¼ 0.6) and average mutation frequency (0.12/100 bp vs. 0.35/100 bp, p ¼ 0.19).The presence (or frequency) of BCL6 mutations also did not correlate with the type of organ transplanted, immunosuppression, site of PTLD (nodal vs. extranodal), phenotype, EBV infection or AID expression.Contrary to a prior study [3], the presence of BCL6 mutations showed no correlation with clinical outcome.However, patients who died from disease had a significantly higher mutation frequency compared to those who are alive or died from other causes (0.32/100 bp vs. 0.05/100 bp, p ¼ 0.03).

Analysis of genes targeted by ASHM
To examine the role of ASHM in the pathogenesis and progression of PTLD, we sequenced the 5 0 region of six non-Ig genes previously identified as ASHM targets, including two (CIITA and BCL7A) that have not been previously reported (L.P., unpublished observation).No mutations were seen in the IM-like lesion and this case was excluded from further analysis, since a clonal population IM, infectious mononucleosis; P-PTLD, polymorphic post-transplant lymphoproliferative disorder; M-PTLD, monomorphic PTLD; DLBCL, diffuse large B-cell lymphoma; GC, germinal centre; BL, Burkitt lymphoma; w, weak staining; dim, diminished; NA, not applicable; ND, not diluted.ASHM was significantly more frequent in M-PTLD than in P-PTLD (8/12 vs. 1/8 mutated cases, p ¼ 0.028), and correlated with AID expression (8/10 vs. 0/9, p < 0.005).When considering EBV status, ASHM was observed more frequently in EBVÀ PTLD, both with regard to total number of cases (6/10 vs. 3/10, p ¼ 0.4) and mutation load (21 vs. 4 events, p ¼ 0.05).No correlations were seen between ASHM and the type of organ transplanted, immunosuppression, site of PTLD or phenotype.Notably, all three recurrent PTLD acquired additional mutations in at least one gene, compared to the initial sample.Although, not statistically significant, a greater number of patients who died from disease had evidence of ASHM compared to those who died of other causes or are alive (4/6, 66.7% vs. 2/10, 20%, p ¼ 0.1).The former also had a higher number of mutations (20 vs. 5 events, p ¼ 0.07), with similar mean follow up durations for the two groups (31.5 vs. 32.7 months, p ¼ 0.96).

Gene expression profile analysis
To investigate whether PTLD are characterized by a distinct gene expression signature, we compared gene expression profiles of 12 PTLD with 39 B-NHL samples representative of lymphomas of GC/post-GC derivation.Unsupervised hierarchical clustering showed that, with the exception of the M-PTLD with GC phenotype and the IM-like lesion, all PTLD segregated as a separate group (Figure 1).Additionally, PTLD expression profiles were also different from CLL, MCL [23,27] and marginal zone B-cell lymphomas (GB, unpublished data).Thus, non-GC PTLD appear to represent a distinct type of B-NHL.
Supervised analysis, performed to identify genes differentially expressed between PTLD and normal or transformed B-cells, detected four genes whose expression levels were significantly higher in PTLD (PRDM1/ BLIMP-1, XBP-1, IFITM3 and PCMT1) and only one gene (C19ORF6), of unknown function, which was expressed at lower levels (Figure 2).
Both unsupervised and supervised analyses failed to distinguish P-PTLD from non-GC M-PTLD or EBVþ PTLD (n ¼ 8) from EBVÀ PTLD (n ¼ 4) (Figure 1 and data not shown).
Finally, in an attempt to better define the cell of origin of PTLD, we applied a classification procedure that measures the relatedness of PTLD gene expression profiles to those of normal B-cell subsets [27].PTLD expression profiles were first compared to the gene expression signature that distinguishes GC (centroblasts and centrocytes) from non-GC (naı ¨ve and memory) B-cells.This analysis showed that PTLD are significantly more similar to non-GC B-cells (Figure 3A, E).On comparing PTLD with naı ¨ve and memory B-cells, PTLD appeared more related to memory B-cells (Figure 3B, E).Lastly, we evaluated the relationship of PTLD to memory B-cells versus EBV-immortalized lymphoblastoid cell lines (LCL), representative of activated B-cells, or multiple myeloma (MM) cell lines, representative of terminally differentiated plasma cells.In both analyses, the PTLD expression profile appeared significantly more similar to that of LCLs (Figure 3C-E).

Discussion
The present study was aimed at elucidating the molecular histogenesis of PTLD by performing a comprehensive immunophenotypic, genetic and genome-wide expression analysis on a panel of PTLD representative of the polymorphic and monomorphic subtypes.Our findings

Non-GC PTLD are distinct from other B-NHL subtypes
The relationship of PTLD to B-NHL of immunocompetent hosts is unknown.In fact, B-cell M-PTLD are morphologically indistinguishable from DLBCL (or BL), while P-PTLD do not resemble any currently recognizable type of B-NHL.Our comparative gene expression analysis of PTLD with other types of B-NHL demonstrates that non-GC PTLD have a homogeneous expression profile that differs from that of other types of B-NHL, including DLBCL (Figure 1).The histogenesis of GC PTLD could not be addressed by our study; however, the single GC PTLD in our series clustered with other GC type B-NHL of immunocompetent hosts.The observation that, on supervised analysis, PTLD only showed differential upregulation of transcription factors associated with plasma cell differentiation [33] suggests similarities with activated B-cell like DLBCL [34].The DLBCL used in this study were not classified into cell-of-origin-based subgroups using a gene expression signature [34].However, our unsupervised analysis included immunoblastic lymphomas, a subset of which are known to have phenotypes analogous to activated B-cell like DLBCL [35].Thus, the unique clustering of PTLD most likely reflects distinct biologic features of these types of B-NHL.PTLD also differ from DLBCL of immunocompetent hosts with regard to associated genetic lesions, specifically ASHM.The frequency of mutated post-transplant DLBCL was lower compared to DLBCL of immunocompetent hosts for all non-Ig genes except c-MYC and CIITA [6] (L.P. unpublished data), and all genes showed a lower average mutation frequency compared to de novo DLBCL [6].Furthermore, no PIM1 mutations were observed, similar to an earlier report [5].Overall, our data, in conjunction with previous findings in PTLD and HIV-associated B-NHL, suggest that immunosuppression-related lymphomas have lower frequencies of ASHM [5,36].

Non-GC PTLD are related to activated memory B-cells
Virtually, all clonal PTLD in our study appeared to arise from antigen selected B-cells and the majority (74%) had non-GC phenotypes.These data extend the results of prior studies indicating derivation of PTLD from GC experienced B-cells, which have the phenotype of pre-terminally differentiated B-cells, that is BCL6þ/À/MUM1þ/ CD138À [2,7,8].Of note, one P-PTLD lacked IgV H mutations, but had BCL6 mutations.The latter, in the absence of IgV H mutations, have previously been reported in other types of B-NHL, including PTLD [2].This could reflect the stochastic nature of mutations targeting the two different loci by SHM [37], suggesting GC transit in the absence of Ig mutations.When compared to normal B-cell subsets, gene expression profiles of the non-GC PTLD were most similar to those of resting memory B-cells (Figure 3B, E).However, a comparison between PTLD, resting memory B-cells and LCLs showed a closer similarity of PTLD with LCLs (Figure 3C, E), which represent activated B-cells [38] and have expression profiles similar to IgM-activated B-cells [39].Our findings, thus, support the hypothesis that non-GC PTLD, both EBVþ and EBVÀ, may derive from an activated memory B-cell.The suggestion that some PTLD are derived from memory B-cells was made by Timms et al. [8] based on the presence of Ig mutations with evidence of antigen selection.Previous studies had also pointed to similarities in the phenotypes and latent viral gene expression profiles between PTLD and LCLs or tumours developing in SCID mice inoculated with EBV infected human lymphocytes [40,41].Since memory B-cells serve as reservoirs of latent EBV infection [42], it is conceivable that, in the setting of deregulated immune surveillance, activation or acquisition of growth advantage of cells within this compartment could result in a PTLD.

Relationships between different types of PTLD
Our results show that the gene expression programs of P-PTLD and non-GC M-PTLD are indistinguishable, as both unsupervised and supervised gene expression analyse could not segregate them into two distinct groups.This finding was somehow unexpected given the differences in morphology, cellular composition, and genetic abnormalities that characterize these two types of PTLD [1,3].This is also the case with ASHM, which appears to be rare in P-PTLD but occurs in the majority of M-PTLD, as shown by our and a previous study [5].One possible explanation for the inability of gene expression analysis to separate P-PTLD and M-PTLD is the existence of heterogeneity within one or both subtypes investigated.The observation that distinct or different combinations of genes are targeted by ASHM in different M-PTLD, suggests biologic differences among cases.However, an effect of subtle qualitative differences in cellular composition cannot be entirely excluded.If these assumptions are true, then a significantly larger survey will be required to identify distinct subsets within these two morphologic categories.
It has been hypothesized that PTLD progress along a continuum from polyclonal early lesions to monocloclonal M-PTLD [1,43].The identification of cytogenetic aberrations common to both P-PTLD and M-PTLD supports this view [4].However, a direct relationship between P-PTLD and M-PTLD has not been conclusively established.Only a few studies have evaluated clonal relatedness of PTLD, and clonal identity has been confirmed only in a handful of recurrent M-PTLD and one recurrent P-PTLD [1,[43][44][45].Our analysis provides evidence of a clonal relationship between a P-PTLD and a metachronous non-GC M-PTLD for the first time, supporting a precursor-product relationship between some P-PTLD and M-PTLD.
In our study, EBVþ and EBVÀ PTLD did not show significant differences in gene expression profiles, suggesting a similar histogenesis of these PTLD subtypes.This result is in contrast to a recent study by Craig et al. who described differentially expressed genes, including EBV regulated transcripts, among EBVþ and EBVÀ M-PTLD [46].The differing results may be due to only a few cases analysed in both studies, differences in the types of cases evaluated or methodological differences in the biostatistical analyses.We performed a stringent analysis to identify only consistent differences among EBVþ and EBVÀ PTLD, using both P-PTLD and M-PTLD.The gene expression differences among EBVþ and EBVÀ cases identified by Craig et al. require validation by generating a classifier and testing them in an independent panel of cases, to demonstrate robustness of the expression differences.Clearly, larger series of PTLD of similar phenotypes need to be analysed to determine differences or similarities between EBVþ and EBVÀ PTLD.
It also remains to be determined whether EBVþ and EBVÀ PTLD share aberrations of signalling cascades, cell cycle regulators or proto-oncogenes, as has been suggested for classical Hodgkin lymphoma (cHL) [47].The observation of an EBERþ M-PTLD that recurred as a clonally related EBERÀ M-PTLD, raises the possibility that certain EBVÀ PTLD may result from 'hit and run' oncogenesis [48], similar to EBVÀ sporadic BL and cHL [49,50].A comprehensive survey to detect viral genomes in EBVÀ PTLD is required to definitively exclude a role of EBV or other viruses in the pathogenesis of such lymphoproliferations.

Disease progression is associated with accrual of genetic lesions
We could document accumulation of additional mutations by the neoplastic clone in three recurrent PTLD (including an M-PTLD representing progression of a prior P-PTLD).These findings might reflect clonal heterogeneity at inception, with survival of clones harbouring additional advantageous mutations, or may result from ongoing ASHM activity, as has been described for HIV-related B-NHL [36].The correlation between presence of ASHM and AID expression by neoplastic cells makes the latter tenable.Our data, in conjunction with prior studies, suggest that progression of PTLD is associated with accrual of genetic lesions [1,4,5].We speculate that a higher mutation frequency of non-Ig genes may connote recalcitrant disease, similar to observations in B-NHL of immunocompetent hosts [51].However, larger series of recurrent PTLD need to be studied to evaluate the impact of ASHM on disease progression and prognosis.

B-cell receptor abnormalities in PTLD have a multifactorial etiology
Our analysis showed absent or reduced sIg L expression by flow cytometry in 32% of PTLD.While similar to earlier studies [52,53], a more recent report described a higher frequency of sIg L loss (%76%) based on IHC staining [54].These discrepancies may be due to lower sensitivity of IHC-based methods compared to flow cytometry, as has been reported for other B-NHL [55].
Crippling IgV H gene mutations accounted for loss of sIg L in only half of sIg L À PTLD.In contrast to observations in cHL [56], no loss of OCT2 expression was seen in our sIg L À PTLD, in accordance with a recent report [57].Absent PU.1 expression was seen in 4/6 (67%) sIg L À PTLD, confirming the observations of a recent study [57].This may be due to defective PU.1 expression, as proposed [57], or it may reflect activation of a plasma cell differentiation program [58].Consistent with the latter hypothesis, three of our sIg L À PTLD showed phenotypic evidence of transition to plasma cells, as evidenced by CD20 downregulation, cytoplasmic light chain expression, BLIMP-1 protein expression and/or CD43 expression (data not shown).
Capello et al. recently reported crippling mutations in Ig L genes [54].Since we did not analyse mutations in Ig L genes, it is unclear whether cytoplasmic light chain expression, as observed in two of our sIg H þ sIg L À PTLD, is due to Ig L gene mutations or other aberrations.Expression of sIg H in the absence of sIg L has been reported in LCLs and certain types of B-NHL [59][60][61].The etiology of this phenomenon appears multifactorial, some of the proposed mechanisms include truncations in the V H region secondary to abnormal splicing [59], posttranslational defects [62] and abnormal intracellular transport [63].
Irrespective of the mechanism of BCR loss, immunosuppression may allow the survival of BCR-deficient B-cells, as detected in recipients of solid organ allografts [64].In EBVþ PTLD, the virus itself may rescue B-cell clones, which lack evidence of antigen selection or have crippling Ig mutations [2,7,8], a hypothesis supported by recent in vitro experiments [11].However, the finding of sIgÀ EBVÀ PTLD by us and others [54] suggests that as yet unknown molecular alterations of the neoplastic cells may allow survival of EBVÀ BCR-deficient lymphocytes.
In summary, our findings indicate that non-GC PTLD are distinct from other B-NHL and appear closely related to activated memory B-cells.It is likely that distinct morphologic categories of PTLD are biologically heterogeneous.Further studies are required to explore the relationship of PTLD with other types of immunodeficiency-associated B-NHL.
Median age at initial diagnosis was 21 years (mean 27.5 years, range 2-64 years) and the median time from transplant to initial diagnosis was 7.5 years (mean 6.0 years, range 0.3-17 years).P-PTLD occurred significantly earlier than M-PTLD (3.1 vs. 8.7 years, p ¼ 0.02).Six patients are currently alive 1.7-5 years (median 2.6 years, mean 2.8 years) following initial presentation and 11 patients died 1 month-16 years (median 1.2 years, mean 2.9 years) post-diagnosis, 6 of PTLD and 5 of other causes.All three patients whose recurrent PTLD were analysed had variable durations of remissions between recurrences (1.2, 1.7 and 14 years).

Figure 1 .
Figure 1.Unsupervised hierarchical clustering performed on a representative panel of B-NHL, including 22 diffuse large B-cell lymphomas (DLBCL), 6 follicular lymphomas (FL), 7 Burkitt lymphomas (BL), 4 hairy cell leukaemias (HCL) and 12 PTLD; 5 polymorphic (P-PTLD), 6 monomorphic (M-PTLD) and 1 IM-like lesion (case numbers shown correspond to PTLD in Table1).Clustering is based on the expression of 333 genes.(Colour scale identifies relative gene expression changes normalized by the standard deviation-zero representing the mean expression level of a given gene across the panel)

Table 2 .
Immunohistochemical staining profiles of PTLDAntibody Clone, dilution company IM-like lesion Ã

Figure 2 .
Figure 2. Supervised analysis to identify differentially expressed genes between PTLD (5 polymorphic, 6 monomorphic and 1 IM-like lesion) and a representative panel of normal and neoplastic B-cells, which consists of 7 mantle cell lymphomas (MCL), 8 hairy cell leukaemias (HCL), 7 Burkitt lymphomas (BL), 10 chronic lymphocytic leukaemias (CLL), 6 follicular lymphomas (FL), 23 DLBCL (15 centroblastic-CB-DLBCL, 8 immunoblastic-IB-DLBCL), naı ¨ve and memory B-cells (5 each), and 10 germinal centre B-cells.In the matrix, columns represent samples and rows correspond to genes.The scale bar provides visualization of difference in the z-scores (expression difference/standard deviation) relative to the mean.Genes are ranked based on the z-score (mean expression difference of the respective gene between phenotype and control group/standard deviation).Affymetrix probe set ID, gene symbol and title is provided for the differentially expressed genes

Figure 3 .
Figure 3. Supervised analysis to identify differentially expressed genes between two groups of samples: (A) naı ¨ve and memory B-cells (N&M) compared to germinal centre (GC) B-cells; (B) naı ¨ve B-cells compared to memory B-cells; (C) EBV-transformed lymphoblastoid cell lines (LCL) compared to memory B-cells; (D) multiple myeloma cell lines (MM) compared to LCL.Expression of the selected genes investigated in PTLD is depicted on the right side of each matrix.In the matrices, each row represents a gene and each column represents a sample.Genes are ranked according to the z-score (mean expression difference of the gene between the phenotype group and the control group/standard deviation).The scale bar represents differences in the z-score relative to the mean.(E) Cell type classification to measure the relatedness of PTLD with the investigated populations.The grey area marks the 95% of confidence: p-value decreases with increasing distance from the X-axis

Table 3 .
Analyses of IgV