Introduction

In contrast to immortal cell lines, investigation of the authentic primary counterparts, including malignant bone marrow (BM) cells, is limited by the short survival of the latter ex vivo.1 This obstacle reflects the strict dependence of primary BM cells on their niche, which contributes the space, supplies the nutrients and provides the conditions required for cell–cell interactions and soluble factor exchange.1, 2 Newer investigation tools, for example, molecular profiling of cell populations improved the discrimination between different subtypes of the same tumor and provided prognostic classifications to various hematological malignancies.3, 4, 5, 6, 7, 8, 9

As the loss of niche stimuli is one of the limiting factors in growing primary cells, we searched for the target genes in BM cells that respond to the niche stimuli and did it in the setting of BM malignancy. This selection was made in order to assist in designing new targeted therapy directed at critical niche signals required for growth and survival of the primary tumor cells. For this purpose, we analyzed the changes in gene-expression profile (GEP) following interruption of cell–niche interactions due to BM aspiration. The sustained changes in expression disclosed the subset of genes, which are sensitive to the BM microenvironment. The nature of the microenvironment stimuli regulating the BM cells was further specified using isolated culture conditions.

Materials and methods

Sample collections

BM aspiration was collected into heparinized syringe, and two drops (∼100 μl) were transferred immediately into an Eppendorf tube and flash frozen with liquid nitrogen. Subsequent samples were frozen after longer incubation in the syringe pool (left in room temperature). Alternatively, two drops were transferred into tubes containing 1.2 ml RNAlater (Ambion Diagnostics, Austin, TX, USA).

Patients

Newly diagnosed patients with clinical and laboratory manifestations typical for multiple myeloma (MM) with a high tumor load (severe cytopenias, leukemic phase, extreme paraprotein levels, multiple lytic lesions and hypercalcemia) entered the study. Also, patients with acute leukemia presenting with severe cytopenias (grade 4) and no mature leukocytes seen in the peripheral blood were included. The samples from the leukemia cases were analyzed only if cytospine smears from the aspirates confirmed complete ‘replacement’ (∼100%) by tumor cells. Another case had leukemic relapse of diffuse large B-cell lymphoma (DLBCL). The patients provided written informed consent. The study was approved by Assaf Harofeh Institute, Zerifin, Israel, review boards.

Long-term cultures

Residual BM aspirate was placed in a vertical tube for ⩾30 min, and then the fluid layer above the red blood cell sediment was collected and 300 μl fractions (containing 20–100 × 103 cell per μl) were seeded into 24-well plates (BD Falcon 24-well Multiwell Plate, Bactlab Diagnostics, Caesarea, Israel). After >15 min incubation (room temperature, Bactlab Diagnostics, Caesarea, Israel), the cultures were supplemented with 1 ml RPMI1640 containing 2 mM L-glutamine (Sigma, Rehovot, Israel), 50 U/ml penicillin, 50 μg/ml streptomycin (Sigma) and 20% (v/v) heat-inactivated fetal bovine serum (Biological Industries, Kibbutz Beit Haemek, Israel). Cells were maintained at 37 °C in a humidified 5% CO2 incubator. Cultures were fed biweekly by replacing 50% of the medium with fresh supplement. The proportion of tumor cells was determined morphologically using cytospine smears. For RNA extraction, most of the medium was removed (leaving only 200–300 μl), and after vigorous pippetation, the released cells were collected into an Eppendorf tube. Fractions of ∼70 μl were flash frozen at different time points, starting immediately following release. The cells were reincubated between freezing procedures.

RNA extraction:

Frozen samples were lyzed by adding 300 μl lysis buffer to tubes. Samples stored in RNAlater were fractionated and lyzed by adding 10 × volume (v/v) lysis buffer. RNA was isolated by MagNA Pure Compact RNA Isolation procedure using MagNA Pure Compact instrument (Roche). Integrity of RNA was examined by Agilent 2100 Bioanalyzer (Santa Clara, CA, USA).

Gene-expression profile

Biotin-labeled cRNA was generated from 200 ng total RNA, hybridized onto GeneChip Human Gene1.0 ST Array (Affymetrix, Santa Clara, CA,USA) and the data were processed with the Affymetrix GeneChip Scanner 3000 and Affymetrix Expression Console. Normalization was done by the Robust Multi-array Average (RMA) method, and fold-change results were calculated relative to the reference sample of each case (which was fixed without delay). The entire GEP data is accessible from NCBI Gene Expression Omnibus GSE36036 and GSE39184. http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE36036. http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE39184.

Quantitative real-time PCR

cDNA was obtained by Transcriptor First strand cDNA synthesis kit (Roche Diagnostics, Mannheim, Germany), according to the manufacturer’s instruction. Quantitative real-time PCR with various primers and LightCycler 480 SYBR Green I Master kit (Roche Diagnostics) was done using LightCycler 480 Instrument II (Roche Diagnostics) according to the manufacturer’s instructions. PCR primers and thermal cycling parameters were designed using Primer3 software (Supplementary Table 1). Standard curves were generated using 0.2, 0.5, 1, 2 and 5 ng/μl of control GAPDH cDNA.

Results

BM aspirations were obtained from an overall group of 27 patients with MM, 14 patients with acute leukemia and one case with leukemic phase of DLBCL. The aspirates remained in their syringes, and on each time point, a fraction was transferred into an empty tube and fixed (either by flash freezing or using RNAlater). Therefore, each case had its own reference sample fixed immediately following aspiration and many additional samples (6–40 samples) whose fixation was delayed (up to 24 h). GEP was analyzed from whole BM samples of the 10 MM cases, with the highest proportion of plasma cells (60–95%) and the best quality of purified RNA (RIN ⩾6). The latter group of patients presented typical characteristics of symptomatic myeloma in terms of age distribution, gender, clinical and laboratory abnormalities (Table 1a). The relatively high proportion of tumor cells, despite skipping the usual cell-separation procedures, which would prevent the immediate fixation required to record the authentic ‘in vivo’ expression, was the consequence of strict patient selection and using of the initial aspiration fraction while the majority of residual normal cells were granulocytes. GEP was also analyzed from whole BM samples of four acute myeloid leukemia (AML) cases for whom the proportion of tumor cells was nearly 100% by morphology and from a case with leukemic relapse of testicular DLBC with 70% tumor cells (Table 1b).

Table 1a MM patient characteristics
Full size table
Table 1b Other patient characteristics
Full size table

Prompt ‘switch’ in GEP

The first participant had MM (case A), and the BM samples selected for microarray analysis were those frozen immediately following aspiration, at 17 min later and following 120 min spent in the syringe pool. The comparison of the three GEPs accepted revealed prompt induction of numerous genes after 17 min ex vivo, which included upregulation of 363 genes >2-folds, 34 genes >3-folds, 10 genes >4-folds, 7 genes >5-folds and 3 genes >10-folds compared with the reference sample frozen immediately following aspiration. Moreover, after 120 min ex vivo, nearly two-third of the former upregulated genes (at 17 min) remained overexpressed and many of which displayed further increase in expression signals. Surprisingly, a very similar ‘switch’ signature was recorded from the aspiration samples of an AML patient (case A′) after 20 min and 120 min spent out-of-niche compared with his reference GEP (Table 2). In both patients, downregulation was milder and involved fewer genes.

Table 2 Spontaneous induction of numerous genes
Full size table

To trace the earliest onset of gene modulation and view the dynamic of changes, we used quantitative real-time PCR analysis, directed at mRNA levels of three inducible genes. The results revealed an almost immediate induction onset with progressive increase in transcript levels during the first 120 min ex vivo (Table 3).

Table 3 Induction onset (qRT-PCR)*
Full size table

The ‘switch’ in GEP is sustained

The annotation database revealed that many of the promptly induced genes in the aspirates from the first two cases examined are classified as ‘immediate early genes’ (FOS, FOSB, RGS1, DUSP1, JUN, CD69, EGR1, PTGS2 and NR4A2). This class of genes is characterized by rapid induction onset upon stimulation but transient expression pattern.10 To follow the expression of these and related genes at more distant time intervals following aspiration, we analyzed BM samples from additional cases and extended the time-to-fixation delays. It was found that the ‘switch’ in expression of the above genes is still recognizable even after 11 h ex vivo (Figure 1). Furthermore, the majority of the early GEP changes (⩾2-folds) recorded during the first 120 min ex vivo persisted after 260 min (case C′), 420 min (case F), 480 min (case G) and 540 min (case B) spent in the syringe pool, and, frequently, the changes continued to increase (Figure 2). The ‘switch’ signature was also ‘fixed’ in a sense that no late-onset induction or downregulation of genes was evident during all our records, including caspases or other proapoptotic genes expected to be rising. Finally, the tumor type had no visible landmarks on the ‘switch’ signature, and the top modulated genes were similar among different individuals whether diagnosed with MM, AML or DLBCL.

Figure 1
figure 1

Sustained ‘switch’ in GEP following BM aspiration. BM samples from 10 patients with MM, 4 with AML and 1 case with DLBCL (BCL) were fractionated and fixed (either by flash freezing or using RNAlater) at different time points following aspiration. The samples were analyzed by microarray, and the common changes in gene expression (⩾2-folds) are displayed as fold changes relative to the signals recorded from the corresponding reference samples (fixed immediately following aspiration).

Full size image
Figure 2
figure 2

Induction of AREG. The signal intensities are depicted against the time to RNA fixation following BM aspirations obtained from patients with MM (cases A–J), leukemic phase of DLBCL (case B′) and AML (cases A′, C′–E′) or alternatively following mobilization from the bottom well into suspension of cultured BM cells from cases B and C (see details in the next section).

Full size image

The ‘switch’ signature is reversible and contact dependent

In theory, the ‘switch’ in GEP could be secondary to any of the following insults: (1) Stress of aspiration and cell transfer to ambient conditions. (2) Loss of regulatory cues from the BM niche proper. (3) Loss of external inputs from habitual cell–cell interactions among adjacent BM cells due to cell spread, dilution with red blood cells and absence of extracellular matrix to adhere. To search for the real cause, we turned to isolated culture conditions. Our culture model was designed to enable recovery of the usual cell–cell contact relationships via seeding of intact BM samples (excluding red blood cells) on a matrix and adding the medium only after adherence of the whole sample in order to avoid cell dilution. In our experience, such conditions can delay apoptosis of primary MM cells for variable periods of time, although cell proliferation remains limited (as evident from the failure of cellular areas on the bottom well to fill adjacent empty areas).11, 12 Moreover, owing to the short life span of normal granulocytes, the purity of the MM cell populations increased over culture days and, eventually, the cultures became homotypic, excluding scattered macrophages (Supplementary Figure 1). As anticipated, the prolonged viability of the MM cell populations in our culture model was associated with near-complete regression of the changes in GEP, which evolved following aspiration. Specifically, in all analyzed cases: J, B and C, where their aspiration samples were cultured for 2, 7 and 29 days, respectively, the ‘switch’ in GEP fade over culture days except for IL8, which remained overexpressed and its signals even increased further (by 61.6-, 7.6- and 3.9-folds, respectively). Moreover, the ‘switch’ signature recurred after leaving the cultured cells in suspension for additional 60 min (case B) or 120 min (case C), following their release from the bottom well compared with the cells that were frozen immediately following release (Figure 3), and the mobilized cells died within 24 h. These observations indicate that the ‘switch’ phenomenon is reversible and basically contact dependent.

Figure 3
figure 3

The ‘switch’ in GEP is contact dependent. Cultured BM cells (CC) were flash frozen at different time points following their mobilization from the bottom well into suspension. Pleural fluid (PF) cell sediment containing pure MM cell population was split with one part left in the bottom tube and the second was transferred into another tube and maintained in suspension after redilution with pleural fluid. Both parts of the sediment were flash frozen at 270 min after the split. The frozen samples were analyzed by microarray, and the differences in signal intensities are displayed as fold changes relative to the GEPs recorded from the reference samples (fixed immediately following mobilization into suspension). Unannotated genes are denoted by their Affymetrix probe set ID number. The results showed similar changes in GEP following cell mobilization and maintenance in suspension as evolved following BM aspiration, indicating that the switch in GEP is related to the loss of cell–cell–matrix contact relationships.

Full size image

Another source of pure MM cell population was malignant pleural effusion that developed in case C. The pleural fluid obtained formed thick sediment on the bottom tube, which by morphology contained only sheets of MM cells without normal cells. On that occasion, the GEP comparison was between sediment cells that were flash frozen after respreading and maintenance in suspension for additional 270 min versus intact sediment cells flash frozen without any spread. Once again, the spread of the crowding tumor cells into suspension reinduced their ‘contactless’ signature, which already started to regress when the cells were allowed to cluster together on the bottom tube (Figure 3).

‘Contactless switch’ during remission

To clarify if the ‘contactless switch’ is tumor related or actually represents a more generalized phenomenon, we repeated the GEP comparison during remission. The patient selected (case B) entered complete remission following treatment with bortezomib and dexamethasone (with drop in her monoclonal protein from 6.41 g/dl to undetectable levels). It was found that the switch in GEP recorded before treatment recurred following aspiration during remission, although many of the modulated genes were discordant. Consequently, the overall GEP changes could be categorized into three different groups: (1) Genes that were modulated during remission only; in this group, the top modulated gene was CXCL12 (data not shown) which has a major role in homing of myeloid progenitors. (2) Genes modulated in pretreatment samples only (Figure 4, lower field); in this group, the top cited gene in the context of MM was the homing receptor CXCR4.13 (3) Concordant genes which were modulated in both disease stages. The latter category included 22 genes, with many of which being recognizable as immediate early genes (Figure 4, upper field).

Figure 4
figure 4

‘Switch’ in GEP during remission versus before treatment. A comparison between the top post-aspiration changes in GEP (⩾2-folds) evolved in the BM samples from case B during complete remission (CR) (right column) versus the changes recorded before treatment (MM). The results are displayed as fold changes relative to corresponding reference samples (fixed immediately following aspiration). The lower field demonstrates the genes modulated in the pretreatment samples only. The upper field demonstrates overlapping genes modulated in both disease stages. Unannotated genes are denoted by their Affymetrix probe set ID number.

Full size image

Unannotated niche-modulated genes

In addition to known genes, the ‘contactless’ signature consisted of many nameless and unannotated genes (Figure 1). The prominent member in this category was MOP-1, showing remarkable induction in all ‘contactless’ situations. Other genes (that is, Affymetrix probe sets 8017096 and 8015796) were only induced in BM samples that were largely ‘replaced’ by tumor cells. The top unannotated genes induced in the aspirates from the 10 MM cases are presented on Supplementary Table 2. This list likely includes newer immediate early genes waiting for validation.

Niche-modulating genes

Interestingly, the regression of the ‘contactless’ switch during culture was followed by induction of a new set of genes, this time mostly encoding extracellular proteins (Figure 5); among which were angiogenic factors (IL8 and CXCL5) and extracellular matrix proteins (SPP1 and FN1), as well as additional chemokines and growth factors (CXCL5, CCL2, CCL20 and IL6). This adaptive response uncovered inducible genes becoming overexpressed permanently, presumably due to failure of the tumor cells to recruit the stimulated niche components (for example, vascular vessels and extracellular matrix scaffold). Some of the above niche-modulating genes can be considered therefore attractive therapeutic targets, especially SPP1 (or its product osteopontin) and IL8 because of their strongest induction. Likewise, inhibitors to niche-modulated gene products, such as amphiregulin (encoded by AREG) and COX-2 (encoded by PTGS2), may have a role in reducing the tumor expansion in certain BM malignancies. Conversely, suppression of niche-modulated genes like JUN, EGR1 and KLF4 may improve the culture success.

Figure 5
figure 5

Niche-modulating genes. The top induced genes in cultured BM samples from three patients with MM are listed . The changes were recorded after culture for 1, 7 and 29 days and are presented as fold changes (green) relative to the signal intensities recorded from corresponding fresh BM samples flash frozen immediately following aspiration. The genes induced during culture were totally different from the ‘contact-modulated genes’ except for IL8, which was also induced following aspiration.

Full size image

Discussion

This study exhibited pitfalls in gene-expression profiling when RNA fixation is not executed immediately following interruption of homeostasis. In addition, the changes in GEP recorded following aspiration from the whole BM samples of a heterogeneous group of patients with diverse hematological malignancies, and disease stages were integrated into a uniform ‘switch’ signature that was recognizable as early as few minutes following aspiration, persisted for many hours and regressed only after cell–cell–matrix contact was resumed. The latter observation indicates that cell–cell contact per se controls the expression of numerous genes in BM cells, while interruption of the contact relationships relieves cell–to–cell regulation and permits full expression of the ‘default’ genetic program inherent to each individual cell.

The resuming of contact that suppressed the ‘contactless switch’ signature contributed in our culture model to increased viability of the tumor cells. Such protective effect may enhance tumor cell crowding in niche, and the contact may also serve to distribute vital signals from unreachable niche components like osteoclasts, vascular vessels and stromal cells, thereby overcoming the limited access to these regulators. The survival advantage for tumor cells gained by the overwhelming cell density was balanced, however, by stasis and lack of any expansion features, suggesting contact inhibition of cell cycle combined. Indeed, tumor expansion is more likely favored by the ‘contactless’ genetic program with its overexpressed oncogenes (FOS, JUN), homing genes (RGS1, CD69), angiogenic genes (IL8, AREG) and growth-promoting genes (PTGS2), whereas suppression of these genes presumably stimulates cell migration and inhibits proliferation. Nevertheless, the real physiological set up in which switch to the ‘contactless’ signature is allowed remains obscured. One possible scenario is that the dual function of contact, which delays apoptosis of crowding tumor cells but also inhibits their proliferation combined together with contact suppression of homing genes, ultimately, directs cell spreading and tumor expansion toward less populated niche regions where the ‘contactless’ gene signature is supposed to be less suppressed.

In data analysis of the niche-modulated genes based on the Ingenuity software (Redwood, CA, USA), the top function annotation category was “cancer” (25 molecules) followed by “cell death” (23 molecules) and “proliferation of cells” (20 molecules) (Supplementary Table 3). However, “proliferation of vascular tissue” yielded only one molecule (PTGS2), whereas our literature search yielded six molecules. Further analysis of the data from a niche-regulation perspective reasoned classification of the niche-modulated genes (Figure 1) into four functional categories:

Homing

Both RGS1 and CD69 regulate leukocyte retention in niche. RGS1 attenuates migration and retains cells in niche by regulation of G-protein signaling and reduction of B-cell response to chemokines such as CXCL12.14 Overexpression of CD69 retains immature B-cells in the BM and reduces their number in blood.15

Angiogenesis and stromagenesis

Several niche-modulated genes, for example, AREG, PTGS2, RGS1, RGS2, NR4A2 and IL8 support angiogenesis and recruitment of stromal cells.16, 17, 18, 19, 20, 21, 22, 23, 24

Expansion of BM cells

Amphiregulin (encoded by AREG) also supports tumor cells directly and promotes MM cell growth.25, 26 The upregulation of FOS, JUN, NR4A2, TRIB1 and CCL3 and the downregulation of MIR15A can also enhance leukocyte expansion and tumorigenesis.27, 28, 29, 30, 31, 32

Suppression of BM cells

As pointed above, the reduced viability of primary tumor cells out-of-niche was not preceded during all our records by induction of any specific proapoptotic gene like caspase. Still, constitutive overexpression of master regulatory genes like JUN, tumor-suppressor genes like EGR1 and KLF4 and less-defined genes like FOSB, CD69, PPP1R15A, TRIB1 and NFKBIA can be lethal by itself,10, 33, 34, 35, 36, 37, 38, 39 unless contact is resumed on time before translation/protein activation.

In reviewing the literature, we were unable to find any other series with the same GEP design. Still, the study by Chen et al.36 who observed downregulation of EGR1 and JUN in coculture of purified MM cells with osteoclasts compared with culture without osteoclasts is consistent with our data concerning the downregulation of both genes when contact is resumed. Irrespective of this, Chen et al.36 also found correlation between the outcome of MM patients to JUN and EGR1 expression by their tumor cells. The latter observation, however, did not consider the timing of RNA fixation, which in our series strongly influenced JUN and EGR1 expression. Furthermore, in most published GEP data sets analyzing purified MM or AML cells, the expression of the inducible contact-modulated genes was generally high as opposed to the ‘silence’ of these genes when recorded during homeostasis (immediately following aspiration). For example, in GDS531,40 the signal intensities recorded from FOSB, FOS, DUSP1, NR4A2, EGR1 and RGS1 approached the levels of GAPDH (Supplementary Table 4). Therefore, expression analysis may need some standardization in the future.

In conclusion, despite the limited number of cases analyzed in the present study, the serial methodology used and the reproducibility of results substantiated our contact regulation model. The major questions left are: (1) Who is the sensor of contact? and (2) How is the message transmitted and how does it regulate transcription of so many genes simultaneously?