CD38/CD31 interactions activate genetic pathways leading to proliferation and migration in chronic lymphocytic leukemia cells

Abstract

Human CD38 is a pleiotropic glycoprotein belonging to a family of genes coding for enzymes/receptors involved in the catabolism of extracellular nucleotides. CD38 receptor activities are regulated through binding to the non-substrate ligand CD31. CD38 expression above a critical threshold represents a negative prognostic marker for CLL patients. Activation of CD38 by means of agonistic monoclonal antibodies or the CD31 ligand induces proliferation and immunoblast differentiation of CLL cells. Here we define the genetic signature that follows long-term in vitro interactions between CD38⁺ CLL lymphocytes and CD31⁺ cells. The emerging profile confirms that the CD31/CD38 axis activates genetic programs relevant for proliferative responses. It also indicates a contribution of this pathway to the processes mediating migration and homing. These results further support the notion that the CD31/CD38 axis is part of a network of accessory signals that modify the microenvironment, favoring localization of leukemic cells to growth-permissive sites.

Introduction

Human CD38 is a pleiotropic glycoprotein belonging to a complex family of genes coding for enzymes of the cell surface involved in the catabolism of extracellular nucleotides (1). During evolution, CD38 acquired the ability to mediate cell-cell interactions, acting as a receptor and binding the non-substrate ligand CD31. CD31/CD38 interactions drive activation and proliferation of distinct lymphocyte populations (2).

Along with the absence of mutations in the immunoglobulin variable region heavy chain gene (IgVH) (3) and expression of the cytoplasmic kinase ZAP-70 (4), CD38 expression above a critical threshold represents a reliable negative prognostic marker for chronic lymphocytic leukemia (CLL) patients (5). The working hypothesis of our group is that CD38 is not a mere a marker in CLL, but a cell surface receptor and adhesion molecule directly involved in the delivery of growth signals (6). Several pieces of evidence support this view. First, CD38 expression is higher within the bone marrow and the lymph nodes, where the proliferative core of the disease resides (7). Moreover, within each CLL clone, cells expressing CD38 are enriched in expression of Ki-67, suggesting that CD38⁺ cells represent a cycling subset (8). The CD38⁺ subset is also characterized by a specific genetic profile, showing up-regulation of proliferation/survival pathways (9).

We showed that activation of CD38 induces proliferation and immunoblast differentiation of CLL cells (10). These signals are regulated by physical localization of the CD38 receptor within lipid micro-domains and act in synergy with the B cell receptor (BCR)/CD19 complex (11). The initial observations were obtained using agonistic monoclonal antibodies (mAbs) and then extended to a model closely resembling the physiological environment where CLL cells grow and expand. Indeed, CD38⁺ CLL cells can bind to murine fibroblasts transfected with the CD31 ligand with resulting increased growth and survival (12). In vivo, several CD31⁺ cells can be found in lymphoid organs, often in close contact with CD38⁺ CLL cells (13).

This work was undertaken in order to obtain a genome-wide signature of the transcriptional events that follow activation of the CD31/CD38 axis. It also provides some answers to the issues raised by a recent report, where the relevance of CD31/CD38 interactions in the context of CLL was called into question (14). The emerging genetic profile indicates that CD31/CD38 interactions are indeed followed by activation of genetic programs relevant not only for proliferative responses, as could be anticipated on the basis of in vitro experiments, but also for migration and homing. This latter point deserves further attention and points to a potential role for CD38 in directing CLL cells to specific microenvironments.

Materials and Methods

Patients and cells

10 samples from CLL patients presenting with typical morphology and immunophenotype were obtained after written informed consent in accordance with Institutional Guidelines and Declaration of Helsinki. Analyses included CD38 and ZAP-70 stainings [cut-off positivity values of ≥20% and ≥10%, respectively (15)], IgVH mutational status (unmutated if ≥98% homology to the germline gene) and fluorescence in situ hybridization (FISH) for chromosomes 11, 12, 13 and 17 (Table 1). B cells were purified from PBMC by negative selection (15).

Table 1.

Clinical and biological features of the CLL patients

Patient No.	Age (yr)	Sex	Stage	YD	TX	IgVH (%)*	CD38 (%)	ZAP-70 (%)	FISH
1	60	F	0/A	10	T	100	54	2	11q-
2	65	F	2/A	7	T	100	35	12	11q-,13q-
3	72	F	0/A	7	NT	-	70	11	+12,17p-
4	47	M	2/A	21	T	91.2	75	0	ND
5	57	M	4/C	7	T	97	79	9	ND
6	83	M	2/A	5	T	99.7	66	23	+12
7	75	F	1/B	6	NT	99	42	12	17p-
8	61	M	4/A	8	NT	99.3	78	21	ND
9	48	M	2/B	8	T	99.1	34	35	13q-
10	67	M	2/A	13	T	99	58	14	+12

Stage: defined according to Rai and Binet diagnostic criteria. YD: years since diagnosis. TX: therapy; NT: untreated.

^*Shown as percentage similarity to the closest germline gene.

-: missing data;

ND: none detected.

Culture conditions

Mouse L-cells transfected with human CD31 (L-CD31⁺) or with the empty plasmid (L-mock) were obtained as described (16). Freshly purified CLL cells (1.5x10⁶) were plated on mitomycin C-inactivated (0.5 mg/ml, 20 min at 37 °C) L-CD31⁺ or L-mock fibroblasts (2x10⁵/well in a 24w plate) for 5 days.

Gene profiling

1. Data collection. RNA was obtained from freshly purified CLL cells (“basal”) or after culture on L-CD31⁺ (“L-CD31⁺” profile) or L-mock fibroblasts (L-mock” profile). Amplified and labeled specimens from each sample (15) were combined with the Human Universal Reference Total RNA (Clontech, Saint-Germain-en-Laye, France), fragmented and hybridized to oligonucleotide glass arrays representing 41,000 human unique genes and transcripts (Human Whole Genome Oligo Microarray Array, Agilent Technologies, Cernusco, Italy). A dye-swap replicate was performed for each sample.

2. Data analysis. Data were pre-processed using the Rosetta Resolver SE software (Rosetta Biosoftware, Seattle, WA) (15). Error-weighted Student two-sample paired t-test with Benjamini and Hochberg p-value adjustment was applied to obtain differentially expressed sequences in the “L-CD31⁺ vs L-mock, “L-CD31⁺ vs basal” and “L-mock vs basal” class comparisons, after a sequence pre-filtering step (average expression ratio p-value respect to the Universal Reference less than 0.01 in at least 50% of the patients of each class comparison). Venn diagrams were then created to visualize intersections between class comparison results and to select the sequences of interest.

3. Pathway analysis. Significant differential expression at the level of a collection of pre-defined pathways from the KEGG database was identified using PLAGE (http://dulci.org/pathways/) (17).

4. Hierarchical clustering. Results were represented using Cluster 3.0 and TreeView (http://rana.lbl.gov/EisenSoftware.htm) and applying hierarchical clustering to the pathway matrix output given by PLAGE.

The data are deposited in the NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) and are accessible through the GEO series GSE14063.

Results and Discussion

CD38 receptor activities are regulated through binding to CD31, a non-substrate ligand surrogated by ligation with agonistic mAbs [reviewed in (1)]. In vitro experiments showed that these interactions induce marked proliferation and immunoblast transformation of a fraction of the leukemic clone. We asked whether these results could find a confirmation using a different technological approach, needed also because the relevance of the CD31/CD38 axis in CLL has recently been called into question (14). This issue was answered by obtaining a genome-wide signature of the events taking place after CD31/CD38 interactions. To this aim, CLL cells from 10 CD38⁺ patients were used immediately (“basal” profile) or cultured for 5 days in the presence of L-CD31⁺ transfectants (“L-CD31⁺” profile) or of mock-transfected cells (“L-mock” profile), used as controls. The choice of the 5-days time point was based on previous results indicating that proliferation and survival are highest after this incubation time (12). Moreover, this approach rests on physiological considerations in that CD31/CD38 static interactions are likely to take place within lymphoid organs, where leukemic cells lie in close and prolonged contacts with CD31⁺ residential elements (13).

Sample and Reference RNAs were co-hybridized to arrays representing 41,000 human unique genes and transcripts. The strategy selected to identify the signature in response to CD31/CD38 interactions relied upon i) identification of genes differentially expressed in the “L-CD31⁺ vs L-mock” comparison, and ii) subsequent exclusion of genes modulated in the “L-mock vs basal” comparison. Results are represented in the Venn diagram of Figure 1. Two p-value thresholds were adopted to i) exclude genes that changed expression simply as a result of co-culture (“L-mock” vs “basal”, p-value 0.01), and to ii) select sequences with a more significant level of modulation (“L-CD31⁺” vs “L-mock” and “L-CD31⁺” vs “basal”, p-value 0.001) among the ones differentially expressed as a direct consequence of co-culture on L-CD31⁺ transfectants.

Figure 1. CD38 engagement by the CD31 ligand triggers genetic pathways leading to cell proliferation and movement.

Venn diagram showing the strategy selected to identify the signature in response to CD31/CD38 interactions. Results were obtained by applying error-weighted Student two-sample paired t-test with Benjamini and Hochberg p-value adjustment to “L-CD31⁺ vs L-mock”, “L-CD31⁺ vs basal” and “L-mock vs basal” class comparisons. Circle radii are proportional to the number of transcripts differentially expressed in each condition. The grey area defines the CD31/CD38 signature. A cut-off of 0.01 for the adjusted p-value was applied to exclude genes that changed expression simply as a result of co-culture (“L-mock vs basal”), while a more stringent threshold (adjusted P<0.001) was applied to the other two comparisons. The resulting 1,645 sequences are involved in 85 differentially regulated pathways identified by PLAGE. 30% of these pathways are implicated in lymphocyte adhesion/movement and/or signalling.

The result was the identification of 1,645 sequences modulated after CD31/CD38 interactions. The list of genes contained into the pathways of the Kyoto Encyclopedia of Genes and Genomes (KEGG) with significant enrichment score (adjusted p-value <0.01) is available as Supplementary Table 1. Pathway Level Analysis of Gene Expression (PLAGE) of the sequences identified 85 differentially regulated pathways (Figure 2 and Supplementary Table 2). Of these, 14 pathways (16%) are involved in lymphocyte signaling (black arrows in Figure 2), while 12 pathways (14%) are implicated in cell adhesion/movement phenomena (red arrows in Figure 2). Relevant examples of up-modulated genes regulating activation/proliferation pertain to the BCR, the MAPK and the TLR signaling pathways (Figure 2 and Supplementary Table 2). Specific attention was dedicated to the panel of 30 apoptosis-regulating genes found unmodulated by Tonino and colleagues (14). In our series, 7/30 genes are differentially regulated upon CD31/CD38 interactions, including P19, Bax, Bcl-RAMBO, Bim, PUMA, Harakiri and Mcl-1 (Figure 3). All of these genes, with the exception of Mcl-1, are up-regulated upon CD31/CD38 interactions, suggesting that de-regulation of apoptosis is a component of this signature. The apparent contrast of these results with those of Tonino et al. could be explained on the basis of experimental differences (3T3 cells versus L-fibroblasts, 72 versus 120 hrs incubation) and/or on the basis of characteristics intrinsic to the patients selected.

Figure 2. Representation of the expression pattern of 85 pathways differentially expressed as a consequence of CD31/CD38 interactions.

Hierarchical clustering was applied to the matrix of the activity levels of the 85 KEGG pathways found as differentially expressed in the L-CD31⁺ samples as compared to the basal and L-mock ones by PLAGE analysis.

Figure 3. Representation of the 7 apoptosis-related genes differentially expressed upon CD31/CD38 engagement.

Expression pattern of 7 genes of the 1,645 gene signature that are also part of the 30 apoptosis-regulating genes reported in (14). Red squares refer to up-regulation in the sample with respect to the Universal Reference, while green squares to down-regulation. Expression ratios are represented in a log10 scale. All the genes but Mcl-1 increase their expression as a consequence of CD31/CD38 interactions. Two different Agilent probes specific for Bax are over-expressed in the L-CD31⁺ condition.

In the context of the control of cell movement, relevant examples are up-modulation of genes ruling leukocyte transendothelial migration, actin cytoskeleton and focal adhesion on the one hand, and down-regulation of genes influencing interactions with the extracellular matrix and coding for cell adhesion molecules on the other. This genetic signature implies that long-term CD31/CD38 interactions modulate the competence of CLL cells to grow and progress in response to microenvironmental signals and conditions. Independent immunohistochemical studies on lymph node sections revealed a direct association between the number of endothelial cells (CD31⁺) and the level of CD38 expression by CLL cells (13), indirectly confirming that CD31/CD38 static interactions occur in vivo. Moreover, the percentage of CD38⁺ CLL cells (8, 18) and the density of tumor vessels (19, 20) correlate with neoplastic proliferation and disease aggressiveness.

In conclusion, these results provide further support to the hypothesis that CLL maintenance and progression derive not only from the essential drive provided by the antigen but also from a number of accessory signals that modify the microenvironment favoring localization of neoplastic cells to growth-permissive sites. The CD31/CD38 axis is part of this network.