Identification of new target molecules PTK2, TGFBR2 and CD9 overexpressed during advanced bone marrow remodelling in primary myelofibrosis


Professor Oliver Bock, MD, Institute of Pathology, Hannover Medical School, Carl-Neuberg-Strasse 1, 30625 Hannover, Germany.


Primary myelofibrosis (PMF) is a myeloproliferative neoplasm characterized by remodelling of the bone marrow, including progressive myelofibrosis and exaggerated angiogenesis. Advanced PMF frequently shows a full-blown fibre meshwork, which avoids aspiration of cells, and the expression profile of genes related to stroma pathology at this stage remains largely undetermined. We investigated bone marrow core biopsies in PMF showing various degrees of myelofibrosis by custom-made low density arrays (LDA) representing target genes with designated roles in synthesis of extracellular matrix, matrix remodelling, cellular adhesion and motility. Among a set of 11 genes up-regulated in advanced stages of PMF (P ≤ 0·01) three candidates, PTK2 protein tyrosine kinase 2 (PTK2), transforming growth factor β type II receptor (TGFBR2) and motility-related protein-1 (CD9 molecule, CD9), were investigated in more detail. PTK2, TGFBR2 and CD9 were significantly overexpressed in larger series of advanced PMF stages (P ≤ 0·01 respectively). Endothelial cells of the increased microvessel network in PMF could be identified as a predominant source for PTK2, TGFBR2 and CD9. CD9 also strongly identified activated fibroblasts in advanced myelofibrosis. We conclude that PTK2, TGFBR2 and CD9 represent new target molecules involved in bone marrow remodelling of PMF and warrant further investigation for potential targeted therapy.

Primary myelofibrosis (PMF) is a Philadelphia-chromosome negative myeloproliferative neoplasm (Ph MPN) evolving to progressive myelofibrosis, exaggerated angiogenesis, and aberrant synthesis of the bone matrix (Tefferi, 2005). Molecular defects, such as mutations in Janus kinase 2 (JAK2V617F) or the myeloproliferative leukaemia virus oncogene (MPLW515L/K), are demonstrable in 50–60% of cases in haematopoietic cells but the expanding stroma cell compartment is thought to be reactive in nature (Tefferi, 2005; Spivak & Silver, 2008). A number of cytokines and growth factors derived from megakaryocytes and monocytes are presumably involved in the proliferation of fibroblasts and endothelial cells in PMF (Lataillade et al, 2008). It appears that cellular decay and passive efflux along with inducible secretion contributes to the bone marrow environment enriched by fibrogenic and angiogenic cytokines (Lataillade et al, 2008). Recent data from our group showed no evidence for increased apoptosis of megakaryocytes in PMF as a potential cause for premature decay and cytokine release (Theophile et al, 2008a). Indeed, relevant pro-apoptotic factors, such as BNIP3, were down-regulated as shown by profiling of apoptosis-related genes. It therefore appears likely that other mechanisms, such as the so-called ‘para-apoptosis’ (Centurione et al, 2004; Thiele et al, 1997), might be involved in increased megakaryocyte turnover because cellular remnants (‘naked nuclei’) are frequently demonstrable in the bone marrow of PMF (Thiele et al, 1997).

The switch from the hypercellular, pre-fibrotic phase to the stage of matrix accumulation in PMF has not yet been defined, either by morphological or by molecular markers. Due to the extended fibre meshwork the access to cellularity in advanced PMF is often hampered by ‘dry tap’. Therefore, core biopsies are the only source for investigation of the dynamics of myelofibrosis in comparison to the cellular stages. By taking advantage of the molecular techniques applicable on fixed and embedded tissues, relevant factors that are involved in stroma remodelling of PMF have been described (Bock et al, 2006; Xu et al, 2005; Bock et al, 2008, 2005). However, the investigation of single candidates with potential importance for stroma pathology is time-consuming and screening approaches by gene expression arrays would be of merit. Unfortunately, the limited concentration of mRNA in a processed core biopsy prevents the application of large-scale array technology. To overcome these technical limitations we recently established a pre-amplification strategy for cDNA followed by low density gene expression arrays (LDA) for investigation of total cellularity and isolated megakaryocytes in PMF (Theophile et al, 2008a,b). The custom made LDA represent between 12 and 384 target genes and is appropriate for a straightforward screening of designated candidate genes (Theophile et al, 2008b; Tothill et al, 2005; Hartmann et al, 2008).

In order to gain insight in stroma-related gene alterations in PMF bone marrow, we have now established a LDA representing 45 genes involved in matrix synthesis, matrix remodelling, angiogenesis, adhesion and motility. We investigated a series of PMF showing different degrees of myelofibrosis and normal haematopoiesis. We also tested cases of essential thrombocythaemia (ET) for potential differences compared to PMF. LDA enabled the identification of 11 genes that are overexpressed in the advanced stages of PMF and one gene that is exclusively up-regulated in ET. From the aberrantly expressed gene set in advanced PMF we selected three hitherto unattended candidates for a more extended analysis: PTK2 protein tyrosine kinase 2 [PTK2, previously known as focal adhesion kinase (FAK)], transforming growth factor β type II receptor (TGFBR2) and CD9.

Materials and methods

Study group

Formalin-fixed and paraffin-embedded (FFPE) bone marrow trephines diagnosed according to the 2008 World Health Organization (WHO) criteria (Tefferi & Vardiman, 2008) were retrieved from the bone marrow registry of the Institute of Pathology, Hannover Medical School. Hypercellular, prefibrotic PMF (n = 8), advanced PMF (n = 4), ET (n = 4), and eight control cases were selected for LDA. Patients’ characteristics are summarized in Table I.

Table I.   Patients clinical characteristics.
 PMF mf 0
n = 8
PMF mf 3
n = 4
n = 4
n = 8
  1. Each sample was investigated separately by LDA (no pooling).

Erythrocytes (1012/l)  4·6 (3·3–6·7)  3·4 (2·1–4·6)  4·8 (4·4–5·0)  4·9 (4·4–5·8)
Haemoglobin (g/l)134 (92–181) 98 (67–127)146 (140–154)133 (98–155)
Haematocrit (%) 41·0 (30·2–56·7) 30·0 (23–39·4) 43·0 (42·7–44·2) 40·2 (38·7–44·0)
Leucocytes (109/l) 13·9 (6·9–18·4)  6·4 (3·0–21·6)  7·9 (6·7–9·5)  5·3 (4·8–8·2)
Platelets (109/l)800 (530–1340)322 (41–1263)739 (630–968)250 (165–269)
Age (years) 72 (54–88) 73 (66–79) 58 (37–74) 58 (19–71)
Sex (male/female)  3/5  3/1  2/2  5/3
Splenomegaly  1  4  1Not documented
JAK2V617F, nMutant allele burden  434–63%  230–33%  229–55%None

For re-evaluation of candidate genes in an independent, larger series two PMF groups were established according to the presence and degree of myelofibrosis (mf) as determined by silver impregnation (Gomori) according to standard procedures (Thiele et al, 2005). Accordingly, PMF cases were assigned to the group of hypercellular, prefibrotic PMF (n = 37) or to the group showing advanced myelofibrosis (n = 33). ET cases were likewise re-evaluated according to the WHO criteria (Tefferi & Vardiman, 2008). Control cases displayed normal haematopoiesis in line with age and were indicated to exclude a haematological disorder in the presence of a clinically transient thrombocytopenia, a mild thrombocytosis or mild leukocytosis (n = 17), Table II.

Table II.   Patients clinical characteristics according to JAK2 mutation.
 Cellular PMFAdvanced PMFControls
  1. Values/numbers represent the median and (range). No MPL mutation (MPLW515L/K) was detected in PMF cases under investigation.

  2. wt, wild-type.

Cases, (n) 18 19 19 14 17
Mutant allele burden (%) 34 (5–61)  42 (17–92)  
Age 71 (34–83) 66 (46–85) 73 (49–89) 67 (40–82) 63 (20–84)
Gender (male/female)m: 9
f: 9
m: 10
f: 9
m: 9
f: 10
m: 9
f: 5
m: 9
f: 8
Erythrocytes (1012/l)  4·6 (3·4–6·6)  3·9 (3·4–8·2)  3·9 (2·0–8·1)  3·5 (2·6–4·3)  4·8 (3·7–5·7)
Haemoglobin (g/l)141 (63–155)110 (81–155)109 (30–168) 97 (63–113)135 (73–155)
Haematocrit (%) 41 (32–52) 34 (25–49) 35 (16–56) 27 (11–37) 42 (26–48)
Leucocytes (109/l)  9·9 (2·9–40·0)  9·6 (3·0–31·8) 12·6 (3·8–48·1)  7·6 (3·9–87·5)  6·2 (3·2–13·9)
Platelets (109/l)804 (167–1340)753 (203–1919)304 (66–1415)437 (100–800)198 (62–431)
 Yes 11  8 16  9  3
 No  2  6  1  2  9
 Unknown  5  5  2  3  5

Archival spleen tissue was retrieved from a PMF patient who underwent splenectomy due to symptomatic splenomegaly and for whom extramedullary haematopoiesis (EMH) was demonstrable. Two representative areas (two paraffin blocks) of EMH in this spleen were selected for further investigation. Control spleen tissue was retrieved from two patients following splenectomy due to traumatic injury or otherwise indicated surgical procedure. Spleens in controls showed no histopathological evidence of neoplastic or otherwise suspicious cellular infiltration.

Following Ficoll-density separation of peripheral blood in three PMF patients and the bone marrow aspirate of one healthy donor, CD34+ cells were purified by immunomagnetic selection according to MACS® (Magnetic cell separation; Miltenyi Biotech GmbH, 51429 Bergisch Gladbach, Germany).

Quantification of potential mutant allele burden of JAK2 (V617F) and MPL (W515L/K) in PMF and ET bone marrow cells was performed using Pyrosequencing® assays as described (Bock et al, 2006).

Low density arrays

A set of 45 genes of interest and two reference genes (POLR2A, GUSB) were chosen for custom-made TaqMan Low Density Arrays (LDA; Applied Biosystems, Foster City, CA, USA; Table SI). Another selection criterion for recruitment of target genes was an amplicon size smaller than 100 bp enabling reliable gene expression analysis by LDA in FFPE bone marrows as described (Theophile et al, 2008a,b). The reference gene GAPDH was not selected by ourselves but declared to be mandatory on LDA according to the distributor. The gene set was spotted eightfold (8 × 48) in a 384-well plate allowing concomitant investigation of eight samples per run. To this end, total RNA was extracted from bone marrow core biopsies as described (Bock et al, 2006). The arrays were loaded with 20 μl complimentary DNA (cDNA) generated from 1 μg total RNA by means of the High Capacity cDNA Reverse Transcription kit (Applied Biosystems), 30 μl high-performance liquid chromatography-H2O (J.T. Baker, Phillipsburg, NJ, USA), and 50 μl Universal polymerase chain reaction (PCR) Master mix (Applied Biosystems). The TaqMan low density arrays were performed as single runs on a 7900HT Fast Real-Time PCR system and recorded by the 7900HT sds 2.3 software (Applied Biosystems).

Real-time RT-PCR

RNA from bone marrow core biopsies, spleen and CD34+ cells was transcribed into the complementary DNA by means of the High Capacity cDNA Reverse Transcription kit. Real-time RT-PCR was performed on an ABI PRISM 7500 Fast Real-time PCR System (Applied Biosystems) using the individual TaqMan Gene Expression assays for PTK2 (hs00178587_m1, 68 bp amplicon), TGFBR2 (transforming growth factor, β receptor II; hs00559661_m1, 74 bp amplicon) and the reference genes GUSB (β-Glucuronidase, hs99999908_m1, 81 bp amplicon) and POLR2A (RNA polymerase 2A: Hs00172187_m1, 61 bp amplicon); all assays were purchased from Applied Biosystems. Re-evaluation of SMAD3 mRNA expression in PMF, ET, and controls was performed likewise by using the individual SMAD3 TaqMan Gene Expression assay (hs00706299_s1, 64 bp amplicon; Applied Biosystems).

For re-evaluation of CD9 mRNA expression, RNA (1 μg), pretreated with RNase free (Rnase) DNase (1 U/μg RNA; RQ1; Promega, Madison, WI, USA), was transcribed into the cDNA using 500 ng random hexamer primer (Amersham Pharmacia, Picattaway, CA, USA) and 200 U of SuperScript II Rnase Reverse Transcriptase (Invitrogen, Karlsruhe, Germany). Home-brewed primers and probe were then applied as follows: CD9-F 5′-ACGCTGAAAGCCATCCACTATG, CD9-R 5′-AAGGTTTCGAGTACGTCCTTCTTGG, CD9 probe 5′-TT GAACTGCTGTGGTTTGGCTGGGG (NM 001769.2, 104 bp amplicon). A likewise in-house TaqMan assay was specifically applied for quantification of CD9 relative to reference gene GUSB in the identical cDNA preparation by using GUSB forward 5′-CTCATTTGGAATTTTGCCGATT-3′, GUSB reverse 5′-CCGAGTGAGATCCCCTTTTTA-3′, and GUSB probe 5′-TGAACAGTCACCGACGAGAGTGCTGG-3′ (NM 000181, 81 bp amplicon).

Linearity of amplification for target and reference genes under investigation could be demonstrated over a broad concentration range in a cDNA dilution series enabling relative quantification in at least two independent runs using the ΔΔCT-method as previously described (Livak & Schmittgen, 2001).


Immunohistochemistry on bone marrow and spleen sections was performed using the ZytoChem-Plus horseradish peroxidase (HRP) Polymer-kit (Zytomed Systems GmbH, Berlin, Germany) and a diaminobenzidine (DAB) Substrate High Contrast kit (Zytomed Systems GmbH). Sections (1–2 μm) were deparaffinized and treated with 3% H2O2 for 10 min. Following pre-treatment in a pressure cooker for retrieval of antigens, sections were incubated for 1 h with antibodies against PTK2, TGFBR2, and CD9 by using a mouse monoclonal anti-human PTK2 antibody raised against amino acids 903–1052 of the mature protein (1:100 dilution, H-1, sc-1688; Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA); TGFBR2 was stained with a mouse monoclonal anti-human TGFBR2 antibody raised against amino acids 1–567 of the mature protein (1:100 dilution, D-2, sc-17799; Santa Cruz Biotechnology Inc.). CD9 was detected by a monoclonal mouse anti-human CD9 antibody (1:100 dilution, Clone 72F6, GTX75373; GeneTex®, Inc., San Antonio, TX, USA). Positive control for PTK2 and TGFBR2 were cells from colon carcinoma metastasized to the liver. Megakaryocytes served as endogenous controls for CD9 immunoreactivity.

Statistical analyses and graphics

In the bone marrow study group, one-way analysis of variance (anova) tests were performed followed by Dunnett’s multiple comparison tests. For potential differences of gene expression in spleen tissues and CD34+ cells, unpaired t tests were applied. Probability values ≤0·05 were considered as statistically significant. Graphics were designed using GraphPadPrism, version 5.0 (GraphPad Software, San Diego, CA, USA).


A set of 11 genes was identified by LDA to be overexpressed in advanced PMF

From 45 genes selected to determine stroma-related gene expression in PMF, 11 genes were identified to be significantly overexpressed in advanced PMF (at least twofold or higher). This set comprised genes directly associated with matrix components (collagen types COL3A1, COL4A2) or collagen processing (LOX), induction of matrix synthesis (TGFB1, TGFBR2, THBS1) and matrix remodelling (BMP2, MMP2, TIMP1). We further identified the CD9 and PTK2 to be overexpressed in advanced PMF (Fig 1).

Figure 1.

 Point plot gallery showing significant overexpression of 11 target genes in advanced PMF as evidenced by LDA analyses. SMAD3 mRNA was exclusively increased in ET when compared to controls. Expression levels were displayed relative to the mean of target gene expression in the control group which was arbitrarily set to 1. A missing symbol stands for a case-specific undetectable target gene. Note that LDA analyses were performed as single runs. P values mean *P < 0·05, **P < 0·01. Total LDA results will be completed by review of Table III.

Other genes with designated or proven roles in matrix remodelling (i.e. BMP4, BMP6, BMP7, MMP1, MMP13, MMP14, PLOD2, SERPINE1, PLAT, GREM1) were also overexpressed in advanced PMF, but lacked statistical significance. An underlying mutation in JAK2 (V617F) was not associated with a particular expression level of a target gene. Figure 1 and Tables II and III give a comprehensive review of the results obtained by LDA expression profiling.

Table III.   Complete results obtained by LDA analyses.
Gene (supplier’s assay ID)PMF mf 0 PMF mf 3 ET Controls
  1. Median expression values relative to mean of controls (endogenous control: POLR2A). */**Significant difference (P < 0·05/P < 0·01) to the control group. (*)/(**)Significant difference (P < 0·05/P < 0·01) to the PMF mf 3 group.

  2. n.d., not detectable.

BMP4-Hs00370078_m10·77  3·97 1·15 1·05
NOG-Hs00271352_s13·64  1·85 9·57 1·67
BMP7-Hs00233476_m14·95  6·59 n.d. 1·47
BMPR1B-Hs00176144_m1n.d. n.d. 5·74 1·00
BMP1-Hs00241807_m14·06  4·20 6·02 1·00
BMP6-Hs01099594_m11·24 11·70 0·40 1·11
BMP2-Hs00154192_m11·00  4·25* 1·51 0·91
BMPR2-Hs00176148_m11·64  2·10 1·23 0·98
SMAD3-Hs00706299_s13·91  5·07 83* 0·99
SMAD4-Hs00232068_m11·38  1·85 0·60 1·29
GAPDH-Hs99999905_m10·57  2·40 1·36 1·02
TGFBR1-Hs00610318_m13·43  5·64 0·21 2·25
SMAD5-Hs00195437_m11·23  4·33 1·38 0·80
SMAD1-Hs00195432_m11·05  3·11 0·97 0·95
TGFBRAP1-Hs00188614_m11·20  2·48 0·96 0·99
MMP1-Hs00899658_m17·32 10·36 8·16 1·00
MMP2-Hs00234422_m10·85(*)  6·48 1·14 1·49
MMP13-Hs00233992_m10·15  7·67 0·96 1·12
MMP9-Hs00234579_m11·16  0·89 0·71 1·10
MMP14-Hs00237119_m11·24  7·84 1·00 1·12
THBS1-Hs00962908_m12·47 10·71** 2·73 1·04
TGFBR2-Hs00234253_m11·33  2·94** 1·02(*) 1·02
TGFB2-Hs00234244_m10·97  7·59 2·45 0·87
TGFB3-Hs00234245_m11·26  4·38 n.d. 1·07
COL1A2-Hs01028971_m10·35 14·43 3·10 1·76
COL4A1-Hs01007469_m12·90  4·20 2·17 1·30
MMP11-Hs00968295_m13·72 11·92 0·68 1·03
COL18A1-Hs00181017_m11·44  1·92 1·77 1·06
LOX-Hs00184700_m10·78(**)  3·58 1·96 1·20
COL3A1-Hs00164103_m11·83 24·16* 1·81 1·62
COL4A2-Hs01098873_m12·95 24·54* 2·96 1·69
COL4A3-Hs01022542_m11·10 n.d. 0·36 1·00
FOXP3-Hs00203958_m12·14  1·73 3·60 1·23
CD9-Hs00233521_m11·35(*) 11·69* 1·65 1·40
PLOD2-Hs00168688_m12·85 10·37 2·53 1·21
TNFRSF11B-Hs00171068_m10·05 n.d. n.d. 3·32
TIMP1-Hs99999139_m11·41 11·71** 1·73 1·19
TIMP2-Hs00234278_m11·20  2·55 0·58 1·07
SERPINE1-Hs01126606_m14·53 37·11 0·17 0·99
PLAT-Hs00263492_m11·05 11·69 1·25 1·06
PLAUR-Hs00182181_m11·37  1·20 0·45 1·05
EDN1-Hs00174961_m11·54  4·49 n.d. 1·21
POLR2A-Hs00172187_m11·00  1·00 1·00 1·00
TGFB1-Hs00171257_m11·91  3·92* 0·95(*) 1·14
GUSB-Hs99999908_m11·33  2·87 1·74 1·20
PTK2-Hs00178587_m12·31  5·98* 1·47 1·16
IL6-Hs00985639_m11·18  1·06 n.d. 1·00
GREM1-Hs00171951_m1n.d.  9·87 n.d. 2·07

Re-evaluation: PTK2, TGFBR2 and CD9 are overexpressed in advanced stages of PMF

Re-evaluation of PTK2, TGFBR2, and CD9 in a larger series of cases confirmed the significantly higher mRNA expression in the advanced stages of PMF. The most significant differences between the advanced cases and both the prefibrotic PMF stage and controls were demonstrable for PTK2 (median 4·1, range 0·6–10·9, P < 0·0001) and TGFBR2 (median 2·4, range 1·1–6·0, P < 0·0001). CD9 was slightly higher expressed, by up to eightfold (median 1·7, range 0·2–8·0, P < 0·05), Fig 2. An underlying JAK2 (V617F) did not correlate with a particular aberrant level of PTK2, TGFBR2 and CD9, respectively (P > 0·05) (Table IV).

Figure 2.

 Re-evaluation of PTK2 (A), TGFBR2 (B) and CD9 (C) confirmed overexpression in advanced PMF stages. Note that mean expression level in controls was set to 1. Horizontal bars represent the median of expression.

Table IV.   Re-evaluation: PTK2, TGFBR2 and CD9 are overexpressed in advanced stages of PMF.
 Cellular PMFAdvanced PMFControl
  1. Target gene expression by bone marrow cells in PMF and control haematopoiesis are shown as x-fold. Median values, (range) and number of cases (n) under investigation are displayed. Note that number of cases under investigation for PTK2, TGFBR2 and CD9 differed due to limited sample volumes.

PTK21·4 (0·05–3·6)
n = 19
4·1 (0·6–10·9)
n = 19
1·0 (0·4–1·8)
n = 10
TGFBR21·0 (0·1–2·1)
n = 19
2·4 (1·1–6·0)
n = 19
1·0 (0·3–2·4)
n = 10
CD91·2 (0·3–4·1)
n = 37
1·7 (0·2–8·0)
n = 33
1·1 (0·3–8·0)
n = 17

Endothelial cells expressed PTK2, TGFBR2 and CD9 in PMF and controls – CD9 labelled fibroblasts exclusively in advanced PMF

PTK2 was expressed by endothelial cells in PMF and in normal haematopoiesis (Fig 3A, B, B insert). Capillaries of smaller size and sinusoids showed only faint staining. Megakaryocytes were also stained for PTK2.

Figure 3.

 Immunohistochemistry for PTK2 in PMF (A) and control (B, insert B) stained endothelial cells and megakaryocytes. TGFBR2 strongly marked the extended sinusoids in PMF (C, insert in C) but also sinusoids in controls (D). Megakaryocytes located in bone marrow sinusoids retained TGFBR2 expression demonstrable in PMF bone marrow (C, white arrow in insert) and controls. CD9 strongly labelled atypically sized and clustered megakaryocytes and endothelial cells in PMF (Fig E and insert E.1). Fibroblasts were prominently stained in advanced PMF (solid black arrow in insert E.2). Osteoblasts in some areas showed CD9 expression as well (insert 2.E, curved black arrow).

TGFBR2 labelled the extended sinusoids and endothelial cells of small capillaries in PMF bone marrows (Fig 3C, insert in C) and in controls (Fig 3D). Megakaryocytes within bone marrow but also within sinusoids in PMF were stained for TGFBR2 (insert in C, white arrow). Besides endothelial cells and megakaryocytes (Fig 3E, black arrow in insert E.1), CD9 additionally stained activated fibroblasts in areas of advanced fibrosis in PMF (Fig 3E, solid black arrow in insert E.2). In PMF osteoblasts lining the bone matrix also showed occasional CD9 staining (Fig 3E, curved black arrow in insert E.1). CD9 strongly labelled normal megakaryocytes in control haematopoiesis (Fig 3F).

Expression of PTK2, TGFBR2 and CD9 in CD34+ cells.  All three factors under investigation were heterogeneously expressed at the mRNA level in CD34+ cells of three PMF patients compared to the control. PTK2 mRNA was expressed with a median of 0·7-fold (range, 0·2- to 1·3-fold, P = 0·5751), TGFBR2 showed a median fold-change of 0·9 (range, 0·6- to 1·1-fold, P = 0·0928) and CD9 was overexpressed by up to 2·8-fold (median 1·0-fold, range 0·7- to 2·8-fold, P = 0·5368) (Fig 4B, D, F respectively).

Figure 4.

 mRNA expression of PTK2, TGFBR2 and CD9 in spleen (panels A, C, E respectively) and CD34+ cells (panels B, D, F respectively) of PMF and controls. Bar charts illustrate the mean level of expression ± standard deviation. Mean expression in controls was set to 1. **P < 0·001.

Expression of PTK2, TGFBR2 and CD9 in PMF spleen showing extramedullary haematopoiesis. PTK2 mRNA was overexpressed by up to 2·5-fold in the PMF spleen (mean 2·0-fold, range 1·6- to 2·5-fold) compared to controls (mean 1·0-fold, range 0·6- to 1·4-fold, P = 0·0553) (Fig 4A). Endothelial cells and megakaryocytes were predominantly labelled for FAK protein, normal spleen showed FAK-positive vessels (not shown).

TGFBR2 mRNA expression showed no difference between the PMF spleen (mean 1·0-fold, range 0·8- to 1·3-fold) and controls (mean 1·0-fold, range 0·7- to 1·4-fold) (Fig 4C). Endothelial TGFBR2 protein was demonstrable in PMF and normal spleen (not shown).

CD9 was significantly overexpressed, by up to 4·4-fold (mean 3·9-fold, range 3·4- to 4·4-fold) compared to control spleens (mean 1·0, range 0·8–1·2, P < 0·001) (Fig 4E). CD9 strongly stained megakaryocytes and endothelial cells in PMF spleen (Fig 5), normal spleen also showed labelling in endothelial cells (not shown).

Figure 5.

 CD9 protein expression in EMH in the spleen of a PMF patient. In addition to endothelial cells and some immature progenitors, megakaryocytes were predominantly labelled (Fig 5, insert).

SMAD3 mRNA expression in ET and PMF

SMAD3 was a candidate gene that potentially discriminated ET from PMF by LDA analysis with a mean of 8·8-fold higher expression compared to hypercellular PMF (mean 3·9-fold) and advanced PMF (mean 5·1-fold), (Fig 1). SMAD3 mRNA expression was re-evaluated by real-time RT-PCR in a series of 10 hypercellular PMF (mf 0), eight ET, and 10 controls. On average, a small tendency to higher SMAD3 mRNA expression was demonstrable in ET (median 1·5-fold, range 0·3- to 4·1-fold) as compared to PMF (median 1·3-fold, range 0·8- to 2·5-fold) and controls (median 1·2-fold, range 0·4- to 1·6-fold), but did not reach statistical significance (Fig S1).


Remodelling of the bone marrow stroma in progressive PMF affects vessel density, extracellular matrix components and bone. A large number of effector molecules, such as growth factors, their receptors and subsequent downstream signalling pathways are involved in these changes of the histopathological phenotype (Lataillade et al, 2008). However, in advanced PMF, the profiling of designated key remodelling factors has been hampered by the dense fibre meshwork, which avoids aspiration and thus access to bone marrow cells.

The gene repertoire on the LDA (‘fibrosis chip’) was assembled according to current knowledge on bone marrow stroma pathology including matrix synthesis (fibrosis), matrix remodelling, cellular adhesion and motility (Tefferi, 2005; Lataillade et al, 2008; Xu et al, 2005; Kuter et al, 2007). Previous findings of our own studies on PMF pathology additionally contributed to the selection of genes (Bock et al, 2005, 2006, 2008).

Among the 11 genes shown to be overexpressed in advanced PMF some important candidates emerged, with important roles in matrix remodelling (MMPs, BMPs), induction of matrix synthesis (TGFB1, THBS1) and genes involved in collagen assembly, such as LOX or different collagen types themselves (Fig 1).

Overexpression of the TGFβ type II receptor, TGFBR2, was not entirely unexpected because one of its ligands, TGFβ-1, is known to be overexpressed in PMF (Tefferi, 2005; Lataillade et al, 2008). More surprising, however, was the fact that, besides its proven role in induction of matrix synthesis, i.e. fibrosis, TGFβ-signalling also might play a role in PMF angiogenesis. This, because endothelial cells were the most prominent cellular lineage labelled for TGFBR2 in PMF (Fig 3C). Angiogenesis-related TGFβ signalling depends on the type of stimulated activin-like kinase (ALK), e.g., ALK5 maintains endothelial cells in a quiescent state whereas signalling via ALK1 induces endothelial cell proliferation (Goumans et al, 2009). TGFβ-1 predominantly activates ALK5 via TGFBR2 and therefore does not directly stimulate angiogenesis. However, it was shown that signalling by TGFBR2 is also important for sufficient ALK1 activation (Goumans et al, 2002). Moreover, direct activation of ALK1 following binding of BMPs to their BMP receptor complexes or activation of its co-receptor endoglin is likely to be involved in PMF angiogenesis. Endoglin (CD105) acts in a pro-angiogenic manner via TGFBR2 and BMP receptors (ten Dijke et al, 2008) and its overexpression was previously shown in PMF (Ponzoni et al, 2004). In addition, BMPs were recently shown to be overexpressed in PMF where BMP expression could be located in endothelial cells (Bock et al, 2008). Thus, a positive feed-back loop appears to be active that comprises increased angiogenesis due to mitogenic stimuli of growth factors, such as vascular endothelial growth factor and basic fibroblast growth factor (Lataillade et al, 2008; Steurer et al, 2007; Gianelli et al, 2007), which then leads to an increased overall TGFBR2 level by endothelial cells. Susceptibility of TGFBR2 for pro-angiogenic BMP and endoglin signalling then further stimulates endothelial cell sprouting (ten Dijke et al, 2008). Because endothelial cells in control haematopoiesis also showed TGFBR2 staining this underlines the physiological role of the TGFβ family in normal vessel formation. Accordingly, proliferating endothelial cells in PMF do not presume clonality rather than a reaction to the pure overload of growth factors that effectively induce angiogenesis. Interestingly, TGFBR2 was previously shown to be reduced in MPN subtypes other than PMF, i.e. chronic myeloid leukaemia, ET and polycythaemia vera (Rooke et al, 1999). Therefore, up-regulated TGFBR2 in PMF angiogenesis might be an interesting candidate for further investigations on TGFBR2 downstream signals and potential therapeutic targeting.

Cellular protein tyrosine kinase 2 PTK2 (alias focal adhesion kinase, FAK) is concentrated in areas of cell–matrix-interaction and promotes signal transduction following cellular integrin binding to matrix components. It is also be activated by cytokine receptors, such CXC-Chemokine receptor 4, after binding of stromal-derived factor 1α (Glodek et al, 2007), an axis previously shown to be involved in aberrant progenitor trafficking in PMF (Migliaccio et al, 2008). However, PTK2 overexpression in PMF can also be related to another aberration induced by TGFβ-signalling. As previously shown in a model of TGFβ-1 stimulated kidney fibrosis, the synthesis of collagen type I was dependent on PTK2 activation (Hayashida et al, 2007). Endothelial cells are unlikely to be involved in notable matrix synthesis in PMF but nevertheless, TGFBR2 activation by TGFβ-1 with subsequent recruitment of PTK2 can induce alterations in endothelial cell phenotypes. Indeed, sprouting and abnormal extension of vessel structures are typical histopathological features in PMF bone marrows (Lundberg et al, 2000) and PTK2 activation is essential for sufficient endothelial cell formation (Carmona et al, 2009). PTK2 is probably another effector molecule acting in concert with other TGFβ-related signalling events in PMF.

CD9 (alias motility-related protein 1) is known to be expressed by various human tissues and its physiological role comprises promotion of cell motility and migration but also cellular proliferation and differentiation (Nakamura et al, 2001). In normal blood cells, expression of CD9 has been demonstrated in eosinophils, basophils, B and T cells and platelets (Clay et al, 2001; Maecker et al, 1997). Given that a previous study showed platelet-derived CD9 to be involved in endothelial cell proliferation (Ko et al, 2006), expression by megakaryocytes and endothelial cells in PMF (Fig 3E) suggests a concerted action. Interestingly, CD9 was also shown to be overexpressed in a recent gene expression profiling study of CD34-positive (CD34+) cells sorted from peripheral blood in PMF patients (Guglielmelli et al, 2007). Based on abnormally high trafficking of CD34+ cells from bone marrow to peripheral blood in PMF (Barosi et al, 2001), CD9 could be involved in the increased mobilisation. In our study, the indistinct CD9 mRNA level in CD34+ cells in PMF is most probably due to the small number of patients and not contradictory to the results obtained by Guglielmelli et al (2007). However, we found a strong CD9 mRNA increase in EMH of PMF spleen (Fig 4E) which could be assigned predominantly to megakaryocytes but also some progenitors and endothelial cells (Fig 5). Overexpression of CD9 in PMF bone marrow is apparently reproducible in EMH of PMF and supplements the existing data on PMF pathology in distant organs, such as spleen (Barosi et al, 2004).

The role of CD9-positive fibroblasts remains a matter for future investigation. In a preliminary experiment, we cultured two primary fibroblast cell lines with TGFβ-1 and bFGF. Both factors similarly increased CD9 mRNA in these cells by twofold (data not shown). If this induction plays any role in matrix-remodelling this needs to be further explored.

Because expression of TGFBR2, PTK2 and CD9 is apparently linked to at least two relevant features in PMF, i.e. angiogenesis, which in turn facilitates progenitor cell trafficking (involving reactive cell populations and the malignant clone itself), we believe that these aberrations justify further attention regarding potential therapeutic targets.


The authors are grateful to Ms Sabine Schröter and Ms Anna-Lena Becker for excellent technical assistance. The authors disclose any conflict of interest.


Deutsche Krebshilfe, Dr Mildred Scheel Stiftung 10-2191 (O.B., H.K.). Deutsche Forschungsgemeinschaft – DFG/Bo 1954/1-1 (O.B., H.K.).