All of the authors contributed equally to the design of this review as well as to its final writing and factual content.
David J. Kuter, MD, DPhil, Hematology Division, Massachusetts General Hospital, Yawkey 7940, 55 Fruit Street, Boston, MA 02114, USA. E-mail: firstname.lastname@example.org
In bone marrow biopsies, stromal structural fibres are detected by reticulin and trichrome stains, routine stains performed on bone marrow biopsy specimens in diagnostic laboratories. Increased reticulin staining (reticulin fibrosis) is associated with many benign and malignant conditions while increased trichrome staining (collagen fibrosis) is particularly prominent in late stages of severe myeloproliferative diseases or following tumour metastasis to the bone marrow. Recent evidence has shown that the amount of bone marrow reticulin staining often exhibits no correlation to disease severity, while the presence of type 1 collagen, as detected by trichrome staining, is often associated with more severe disease and a poorer prognosis. It was originally thought that increases in bone marrow stromal fibres themselves contributed to the haematopoietic abnormalities seen in certain diseases, but recent studies suggest that these increases are a result of underlying cellular abnormalities rather than a cause. A growing body of evidence suggests that increased deposition of bone marrow stromal fibres is mediated by transforming growth factor-β and other factors elaborated by megakaryocytes, but it is likely that other cells, cytokines and growth factors are also involved. This suggests new avenues for investigation into the pathogenesis of various disorders associated with increased bone marrow stromal fibres.
A wide variety of benign and malignant disorders are associated with an increase in bone marrow stromal fibres. These fibres are commonly composed solely of reticulin fibres but may also include collagen fibres (Bain et al, 2001). The underlying aetiology and clinical relevance of an increase in stromal fibres is not yet well understood, but evidence suggests that the clinical implications of increased reticulin (reticulin fibrosis) may be quite different from those of increased collagen (collagen fibrosis). Several studies have shown that the amount of bone marrow reticulin shows little correlation with blood counts or the severity of the underlying disease (Hann et al, 1978; Thiele et al, 1991; O'Malley et al, 2005), while the presence and amount of collagen fibres shows a stronger correlation with abnormal blood counts and severity of the underlying disease (Bain et al, 2001; Thiele et al, 2005). Moreover, in conditions that are responsive to treatment, reticulin fibrosis is often reversible, while collagen fibrosis is less likely to be so.
The concern that increases in bone marrow reticulin and collagen may have differing relationships to disease has stimulated recent discussions on the importance of distinguishing between increases in these two types of fibres in bone marrow biopsies (Bain et al, 2001; Thiele et al, 2005). Stains that identify reticulin and collagen are routinely available and the use of both stains to evaluate a single biopsy specimen can provide a more complete picture of the amount and nature of bone marrow stromal fibres than performing either one of these stains alone (Bauermeister, 1971; Beckman & Brown, 1990; Bain et al, 2001).
The use of very clear terminology is also very important. Different investigators may use the same term very differently and the same disease may be known by various names. For example, the term ‘myelofibrosis’ has been used in some contexts to describe any increase in bone marrow stromal fibres, regardless of the type of fibre or the associated disease (Groopman, 1980; Bain et al, 2001), and in other contexts to describe the specific constellation of bone marrow, haematological, and clinical findings associated with chronic idiopathic myelofibrosis (CIMF), a myeloproliferative disease (Barosi, 1999; Dingli et al, 2004; Thiele et al, 2005). Moreover, increase in bone marrow fibrotic tissue may be called ‘collagen fibrosis’ even if the dominant fibre type seen was reticulin (Tefferi, 2000). The growing awareness of the importance of distinguishing between increased bone marrow reticulin and increased bone marrow collagen is motivating a drive for greater precision and consistency in terminology (Bain et al, 2001; Thiele et al, 2005). Recent efforts in this regard include proposals to revise the World Health Organization (WHO) diagnostic criteria for myeloproliferative disorders (Tefferi et al, 2007) and CIMF (Mesa et al, 2007) as well as proposals for grading treatment responses of CIMF by the International Working Group for Myelofibrosis Research and Treatment (Tefferi et al, 2006) and the European Myelofibrosis Network (Barosi et al, 2005).
This review describes the current state of our understanding of the pathophysiology and clinical significance of increased bone marrow reticulin fibres (reticulin fibrosis) and to distinguish reticulin fibrosis from the less common finding of increased bone marrow collagen fibres (collagen fibrosis). Although reticulin is composed partly of collagen, the terms ‘reticulin fibrosis’ and ‘collagen fibrosis’ will be retained to distinguish between the histochemical, biochemical and clinical characteristics associated with these two different types of bone marrow stromal fibres.
Bone marrow microenvironment
The bone marrow stromal microenvironment is composed of cells, structural fibrils and extracellular matrix (‘ground substance’). The cellular components include macrophages, fibroblastic cells (including several types of adventitial and reticular cells), adipocytes and endothelial cells, as well as other less well characterized cells (Weiss & Chen, 1975; Watanabe, 1985; Cattoretti et al, 1993). The most common structural fibrils in the bone marrow are collagen, reticulin, laminin and fibronectin (Bentley et al, 1981; Reilly et al, 1985; Cattoretti et al, 1993). The ground substance includes water, salts, glycosaminoglycans and glycoproteins, most of which have not been well characterized (Clark & Keating, 1995). These elements provide a connective tissue structure for the bone marrow and physical support for the haematopoietic progenitor cells. In addition, the bone marrow microenvironment plays an important role in haematopoiesis by providing humoral factors and cognate interactions that support haematopoietic cells (Harigaya et al, 1981; Ohkawa & Harigaya, 1987; Sudo et al, 1989; Zhang et al, 2004).
There are at least 16 types of collagen in the body, but the fibrillar collagens – types I, II and III – account for 80–90% of all collagen in the body. Bone marrow collagen is primarily composed of type I and type III collagen. All collagens share a similar triple-helical structure made possible by an abundance of glycine, proline and hydroxyproline residues. Collagen triple helix chains are synthesized in the endoplasmic reticulum as long procollagen precursors: type I procollagen is formed from two α1[I] and one α2[I] molecules, while type III collagen is formed from three α1[III] molecules. Post-translational modifications, such as the addition of glucosyl and galactosyl residues to lysine hydroxyl groups, are performed in the Golgi apparatus. The collagen molecules do not form into fibrils and fibres until they are secreted into the extracellular space and extracellular procollagen peptidases remove the N-terminal and C-terminal propeptides (Lodish et al, 2000).
Siegfried first used the term ‘reticulin’ in 1892 to describe the powdery residue which remained after extraction of gelatin from collagenous ‘reticulum’ fibres (as quoted by Puchtler (1964). His intent was to distinguish chemically the interfibrillar material (‘reticulin’) from the intact collagen fibres. Subsequently, the term reticulin has been defined histochemically as the argylophilic fibres identified by various silver staining methods. Although not studied in detail in the bone marrow, electron microscopic examinations of reticulin in the kidney seem to link the initial chemical definition with the more recent histochemical definition (Fleischmajer et al, 1980). These examinations revealed that reticulin is composed mainly of individual fibrils or small bunches of fibrils of type III collagen surrounding a core of type I collagen fibrils, all embedded in a matrix of glycoproteins and glycosaminoglycans (Montes et al, 1980; Lisse et al, 1991; Fleischmajer et al, 1992; Fakoya, 2002; Ushiki, 2002). Reticulin contains approximately 10% carbohydrate, in contrast to type I collagen that contains only approximately 1% carbohydrate (Kiernan, 1999; Stevens & Lowe, 2000). Whereas type I collagen forms thick fibrils, type III collagen usually forms single fibrils or small bunches of thin fibrils; presumably because the bulky glycosylated hydroxylysine groups on type III collagen create steric hindrance that reduces the lateral assembly of the triple-helical domains (Kiernan, 1999). The smaller fibre diameter and increased content of interfibrillar material thus distinguish reticulin from collagen.
Collagen and/or reticulin production are associated with fibroblastic bone marrow cells that include adventitial reticular cells (ARCs), perisinusoidal adventitial cells, periarterial adventitial cells, adipocytes and endosteal cells (Weiss & Chen, 1975; Watanabe, 1985; Cattoretti et al, 1993). A direct relationship between ARCs [detected by an antibody against low affinity nerve growth factor receptor (LNGFR)] and reticulin deposition has been observed (Cattoretti et al, 1993).
Identifying and grading changes in bone marrow fibrotic tissue
The connective tissue structure of the bone marrow is not well appreciated on a routine haematoxylin and eosin stained bone marrow biopsy and usually requires special stains. Histological stains, such as Mallory's trichrome stain, van Gieson stain, or Masson's trichrome stain, can be used to identify collagen (Bain et al, 2001; Bancroft & Gamble, 2002). Trichrome staining methods refer to the use of two different anion dyes (one of small molecular weight and the other of large molecular weight) along with a heteropolyacid, such as phosphomolybdic acid (PMA) or phosphotungstic acid (PTA). Their differential staining is based upon differences in rates of diffusion of dyes of differing molecular weight into fixed tissues. When tissue is fixed, the insoluble proteins create a matrix with different sizes of pores associated with different cellular and extracellular structures. For example, erythrocytes with their high protein content have small pores, whereas collagen has much larger pores (Kiernan, 1999; Bancroft & Gamble, 2002). Upon addition of the small molecular weight anionic dye, fuchsin, cationic binding sites in all tissues will bind the dye. When the larger (uncoloured) anionic PMA or PTA dye is added, it penetrates into more porous tissues such as collagen and displaces the fuchsin, but cannot displace the fuchsin from the less porous tissues such as erythrocytes and muscle. When an even larger molecular weight anionic dye, methyl blue, is added to the tissues, it can only penetrate into the most porous tissue, collagen, thereby displacing the PMA/PTA and ‘specifically’ staining the collagen. The final result of this Masson's trichrome stain is blue-black nuclei, red cytoplasm of muscle and erythrocytes, and blue collagen (Fig 1).
An additional finding, osteosclerosis (new bone formation in the marrow space), may also be found in some patients with collagen fibrosis. When associated with myeloproliferative diseases, osteosclerosis appears to be mediated by activation of the osteoprotegerin pathway (Chagraoui et al, 2003a,b; Tefferi, 2005).
Bone marrow biopsy sections can be stained for reticulin using a silver impregnation technique, such as Gomori's stain (Gomori, 1937, 1939; Bain et al, 2001; Puchtler & Waldrop, 1978), that probably identifies the glycoprotein matrix in which the collagen fibrils lie, rather than the collagen fibrils themselves (Puchtler, 1964; Puchtler & Waldrop, 1978). The reticulin stain is based on the high content of hexose sugars in reticulin (Kiernan, 1999). Reticulin fibres have little natural affinity for silver solutions and must be pretreated to produce sensitized sites where silver will deposit. In the first step, the adjacent hydroxyl groups of the hexose sugars of glycoproteins are oxidized to aldehydes by potassium permanganate. Subsequent exposure to an iron salt, such as ferric ammonium nitrate then ‘sensitizes’ these reactive sites by an unknown mechanism (Gomori, 1937, 1939). Silver diamine [Ag(NH3)2] solution is then added and is reduced to metallic silver [Ag] by the aldehyde groups in the glycoproteins and deposits four silver atoms for each glucosyl, galactosyl, or other sugar residues. When formaldehyde is then added to the silver diamine-impregnated tissue, the invisible local silver reaction is amplified (as in a photographic reaction) to produce visible silver deposits (Kiernan, 1999). The section is typically counterstained with neutral red or eosin so the defined reticulin fibres appear black against a background of red tissue (Fig 1).
Ideally, both a trichrome collagen stain and a reticulin stain should be performed and both the type and amount of fibrosis should be described using a clearly defined grading scale; however, in practice, collagen is almost never revealed by trichrome stain unless there is a marked increase in reticulin. Two slightly different, but widely supported grading scales for these stains have been created (Bauermeister, 1971; Thiele et al, 2005). The original Bauermeister scheme (Bauermeister, 1971) used six different grades, but has subsequently been simplified into a five grade system (Table I and Fig 1) (Manoharan et al, 1979; Bain et al, 2001). The more recent Thiele scale includes only four categories (Thiele et al, 2005). Although published decades apart (in 1971 and 2005), both of these grading scales take into account both the type of fibres seen (reticulin or collagen) and the overall amount of fibrosis (Table I). In narrative descriptions of bone marrow fibrosis, the term ‘fibrosis’ should always be specified as either ‘reticulin fibrosis’ or ‘collagen fibrosis’, with the term collagen fibrosis being reserved for collagen identifiable by trichrome stain.
Table I. Grading scales for the quantification of bone marrow reticulin and collagen.
*Measured per hematopoietic area (to take into account age- and disease-related cellularity.
Scattered linear reticulin with no intersection (cross-overs) corresponding to normal bone marrow
Loose network of reticulin with many intersections, especially in perivascular areas
Diffuse and dense increase in reticulin with extensive intersections, occasionally with only focal bundles of collagen and/or focal osteosclerosis
Diffuse and dense increase in reticulin with extensive intersections with coarse bundles of collagen, often associated with significant osteosclerosis
A biochemical correlate for reticulin fibrosis has been proposed: measurement of the serum procollagen III peptide (PIIINP) (Hasselbalch et al, 1985, 1990; Hasselbalch et al, 1986). During the synthesis of type III collagen, the procollagen peptide is cleaved and released into the circulation. In 35 healthy subjects, levels of PIIINP ranged from 4·9–11·7 ng/ml compared with levels from 9·3–25·7 ng/ml in 35 myeloproliferative disease patients. The PIIINP levels correlated with the degree of reticulin fibrosis but were not elevated in those patients with collagen fibrosis (Hasselbalch et al, 1985).
Disorders associated with increased bone marrow fibrosis
Only two studies have assessed the extent of collagen and reticulin staining in the bone marrow of healthy individuals; both used the original Bauermeister grading scheme (Bauermeister, 1971; Beckman & Brown, 1990). One study analysed primarily sternal marrow samples (Bauermeister, 1971), while the other examined iliac bone marrow biopsies (Beckman & Brown, 1990). Neither study identified any collagen fibrosis in specimens from these healthy individuals, but reticulin staining was seen in over 70% of specimens (Table II).
Table II. Bone marrow reticulin in healthy subjects.
0, no reticulin; N, a few fine fibres; 1+, occasional fine individual fibres plus foci of fine fibre network; 2+, fine fibre network throughout most of the section, no coarse fibres demonstrated; 3+, diffuse fibre network with scattered thick, coarse fibres but no true collagen – negative trichrome stain; 4+, diffuse, often coarse, fibre network with areas of collagenization – positive trichrome stain.
Examples of diseases that cause bone marrow reticulin fibrosis without collagen fibrosis, are listed in Table III. However, many diseases can cause both reticulin and collagen fibrosis in the bone marrow (Table IV). These diseases include benign or readily treatable conditions such as Vitamin D deficiency and autoimmune myelofibrosis as well as neoplastic conditions, such as CIMF (Tefferi, 2005). The peripheral blood findings of anaemia, leucoerythroblastic changes, tear-drop red blood cells and the clinical finding of splenomegaly are usually only associated with collagen, not reticulin fibrosis (Thiele & Kvasnicka, 2006).
Table III. Conditions associated with increased reticulin fibrosis, but not collagen fibrosis.
Nutritional and renal rickets (vitamin D deficiency)
Primary hypertrophic osteoarthropathy
Other granulomatous diseases
Grey platelet syndrome
Systemic lupus erythematosus
Primary autoimmune myelofibrosis
Other autoimmune myelofibrosis
Prior thorium dioxide administration
Focal or localized
Following bone marrow necrosis
Following irradiation of the bone marrow
Adult T-cell leukaemia/lymphoma
Site of previous trephine biopsy
The relationship between increases in bone marrow stromal fibres and disease pathology is not well understood but has been the subject of recent clinical investigations. These studies have attempted to correlate the amount and type of fibrosis with other clinical findings and disease prognosis. A major limitation of most studies is the failure to demonstrate reproducibility and inter-observer agreement as well as the afore-mentioned lack of uniform validated fibrosis response criteria (Barosi et al, 2005; Tefferi et al, 2006). The findings range from a partial correlation between reticulin fibrosis and disease severity in CIMF (Ivanyi et al, 1994; Tefferi, 2000; Thiele & Kvasnicka, 2006), to no correlation in disorders such as human immunodeficiency virus (HIV) infection (O'Malley et al, 2005) and paediatric lymphoblastic leukaemia (Hann et al, 1978), to an inverse relationship in a study of acute myeloid leukaemia (AML) (Thiele et al, 1991). In those conditions that are responsive to treatment, such as autoimmune myelofibrosis (Pullarkat et al, 2003) and systemic lupus erythematosus (SLE)-associated myelofibrosis (Pereira et al, 1998), reticulin fibrosis and even collagen fibrosis can decrease with successful corticosteroid treatment. Increased reticulin fibrosis due to chronic myeloid leukaemia is reversible with imatinib therapy, but is usually unaffected or may even increase following treatment with interferon (Lorand-Metze et al, 1987; Beham-Schmid et al, 2002; Hasserjian et al, 2002). Both collagen and reticulin fibrosis resolve after successful stem cell transplantation for CIMF (Udomsakdi-Auewarakul et al, 2003; Sale et al, 2006; Kerbauy et al, 2007).
The relationship between bone marrow fibrosis and disease progression has been most extensively studied in CIMF, a myeloproliferative disorder characterized by splenomegaly, leucoerythroblastic anaemia, bone marrow fibrosis and extramedullary haematopoiesis. In the early, so-called ‘cellular phase’ of CIMF, there may be little or no increase in bone marrow reticulin fibres. When an increase in bone marrow reticulin staining does occur, it is usually accompanied by an increase in bone marrow megakaryocytes, often morphologically atypical, as well as alterations in cellular and extracellular levels of cytokines with fibrogenic potential. The fibrosis in this disease is thought to represent a ‘reactive process mediated by cytokines that are produced by the cellular components of the clonal proliferation’ (Tefferi, 2000). In a retrospective study of 865 bone marrow biopsies from patients diagnosed with CIMF, a survival analysis showed that there was a significantly more favorable prognosis in patients with the cellular (pre-fibrotic) phase of the disease and that patients with higher grades of bone marrow fibrosis had a significantly higher clinical risk profile (Thiele & Kvasnicka, 2006). A higher grade of bone marrow reticulin content was also found to be an important prognostic parameter in a study of 50 CIMF patients (Ivanyi et al, 1994). In contrast, another study of 35 CIMF patients failed to find any correlation between the amount of bone marrow reticulin and the duration of disease, splenic weight, or degree of splenic myeloid metaplasia (Wolf & Neiman, 1985).
An increase in bone marrow stromal fibres has also been reported in children and adults with acute lymphoblastic leukaemia (ALL) (Hann et al, 1978) and in adults with AML (Manoharan et al, 1979). In adults with ALL and AML, Bauermeister grade 4 fibrosis was present in 33% and 9%, respectively; most fibrosis improved or resolved with therapy, returned with relapse, but showed no relationship to survival or remission rate (Manoharan et al, 1979).
Recently, a rare autoimmune syndrome, primary autoimmune myelofibrosis, has been identified that is distinct from CIMF and has a much better clinical prognosis (Bass et al, 2001; Pullarkat et al, 2003). Primary autoimmune myelofibrosis is defined as cytopenias with bone marrow lymphocyte infiltration and grade 3–4 reticulin fibrosis of the bone marrow; lack of atypical bone marrow cells or osteosclerosis; absent or mild splenomegaly; and the presence of autoantibodies. Among seven patients in a retrospective study who met these criteria, six completely normalized their cytopenias with a course of corticosteroids. Successful treatment was accompanied by partial resolution of bone marrow reticulin fibrosis (Pullarkat et al, 2003).
There have been few reports of increases in bone marrow reticulin in patients with SLE, but this may be under-recognized because it has not been systematically evaluated. In 21 patients with SLE and peripheral cytopenias, increased reticulin was seen in 76% of bone marrow biopsies (Pereira et al, 1998) while one patient also had collagen deposition. In a separate case study of a patient with SLE, reticulin fibrosis, and low platelets, steroid treatment produced an increase in platelets and a concomitant decrease in bone marrow reticulin (Konstantopoulos et al, 1998).
Interestingly, increases in bone marrow reticulin staining have been observed in a variety of diseases that one might not expect to exhibit any type of bone marrow stromal changes. These include pulmonary hypertension (or its treatment with epoprostenol) (Popat et al, 2005, 2006); HIV infection (O'Malley et al, 2005); and the human parasitic disease, visceral leishmaniasis (Rocha Filho et al, 2000). In the case of HIV infection, one study of 35 patients found no correlation between the degree of bone marrow reticulin fibrosis and CD4 cell counts, suggesting that the mechanism of fibrosis in this condition is independent of the severity of HIV infection (O'Malley et al, 2005). Similarly, in 12 cases of visceral leishmaniasis, increases in bone marrow reticulin appeared to have no significant influence on the progression of the disease, nor in the response to treatment (Rocha Filho et al, 2000).
In addition to the disorders described above, increases in bone marrow reticulin have been reported following treatment with haematopoietic growth factors. Although reticulin fibrosis has rarely been reported when recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF) was given to chemotherapy-treated cancer patients (Orazi et al, 1992; Cattoretti et al, 1993), administration of GM-CSF to patients with myelodysplastic syndrome with fibrosis (grades 1–3) produced a variable response: half of the patients had increased fibrosis (by 1–2 grades) while the other half had decreased fibrosis (by 1 grade) (Antin et al, 1990).
Interleukin (IL)-11 and IL-3 have been studied as possible treatments for chemotherapy-induced thrombocytopenia. In a study of 12 women with advanced breast cancer and no evidence of bone morrow involvement, varying doses of IL-11 were given after completion of the chemotherapy regimen. Treatment with the highest doses (50 or 75 μg/kg/d) for 14 d produced increases in megakaryocyte number and ploidy as well as modest (1–2 grades) increases in bone marrow reticulin in seven of 12 patients (Orazi et al, 1996). In a separate study, nine patients with malignancy and no evidence of bone marrow involvement were given recombinant human IL-3 or GM-CSF after high dosage chemotherapy. In both treatment groups, there was an increase in LNGFR-positive cells (putative adventitial reticular cells) but only patients treated with IL-3 exhibited increased bone marrow reticulin (Orazi et al, 1992).
It appears that treatment of thrombocytopenic patients with thrombopoietin (TPO) and TPO mimetics can also produce reversible increases in bone marrow reticulin (Bussel et al, 2006a; Kuter, 2007). In one study of 13 AML patients undergoing induction chemotherapy, eight of the nine patients treated with recombinant human TPO (rhTPO) and GM-CSF exhibited a dramatic increase in the number of bone marrow megakaryocytes as well as an increase in bone marrow reticulin, while only two of the six patients treated with GM-CSF alone had increased reticulin (Douglas et al, 2002). In the rhTPO-treated individuals, the increased reticulin fully resolved a mean of 30 d (range: 13–42 d) after discontinuation of rhTPO treatment.
A new TPO agonist, AMG531, has recently entered clinical trials and has been shown to increase the platelet count in healthy individuals (Wang et al, 2004) and in patients with immune thrombocytopenic purpura (ITP) (Bussel et al, 2006a,b; Kuter et al, 2006; Kuter, 2007). Diffuse reticulin formation in the bone marrow has been detected in two patients enrolled in clinical trials of AMG 531 in ITP patients: 1/57 patients enroled across 3 phase 2 studies (Stepan et al, 2005) and 1/36 patients from an ongoing, open-label extension study (Kuter et al, 2006). Neither had collagen fibrosis or abnormal cytogenetics (Stepan et al, 2005; Kuter et al, 2006). Both patients were asplenic and were receiving relatively high doses of AMG 531 (>10 μg/kg) with minimal or no response. A bone marrow biopsy performed in one of the patients 14 weeks after discontinuation of AMG 531 showed improvement in reticulin deposition (Stepan et al, 2005). A follow-up bone marrow biopsy has not yet been performed in the other patient. The effects of AMG 531 on bone marrow reticulin formation are currently being assessed in a prospective bone marrow biopsy study in ITP patients.
Pathophysiology of increased bone marrow stromal fibres
Most disorders with increased bone marrow stromal fibres are associated with abnormalities of the number and/or function of megakaryocytes and platelets. Cytokines from megakaryocytes and platelets appear to be necessary (but may not be sufficient) for fibrosis to occur. Platelet-derived growth factor (PDGF), a potent stimulator of fibroblast growth found in megakaryocytes and platelets, was one of the first cytokines proposed as a potential cause of bone marrow fibrosis (Groopman, 1980; McCarthy, 1985). This was supported by the observations that megakaryocytes and their precursors are increased in CIMF and that homogenates of megakaryocytes isolated from normal human bone marrow stimulated fibroblast growth (Castro-Malaspina et al, 1981). Further evidence for this came from the observation that increased bone marrow fibrosis occurs in the rare ‘grey platelet syndrome’ which is characterized by excessive release of PDGF (Drouet et al, 1981; Coller et al, 1983). However, more recent studies have shown that, although PDGF stimulates fibroblast growth, it has little effect on the production of reticulin or collagen (Kimura et al, 1989; Terui et al, 1990).
A separate study showed that latent and active TGF-β were increased in bone marrow plasma as well as in peripheral blood and serum from patients with hairy cell leukaemia who had bone marrow fibrosis (Shehata et al, 2004). TGF-β was overexpressed in the hairy cell leukaemia cells in these patients. Active TGF-β concentration was related to the degree of reticulin fibrosis and to levels of PIIINP in the serum. Interestingly, TGF-β from these patients stimulated bone marrow fibroblasts in vitro and they, in turn, produced more type III procollagen and reticulin than type I procollagen and collagen.
Further insights into the role that TGF-β plays in increasing bone marrow collagen comes from studies that demonstrated that TGF-β is required for TPO-induced bone marrow fibrosis in animal models (Yanagida et al, 1997; Chagraoui et al, 2002). In mice injected for 5 d with suprapharmacological doses (100 μg/kg) of a recombinant form of TPO (pegylated megakayrocyte growth and development factor), megakaryocytes and platelets increased as early as day 6 (Yanagida et al, 1997). Elevated plasma levels of TGF-β were detected at days 6 to 8 and increased bone marrow reticulin was noted at day 10. TGF-β was increased 23-fold in extracellular fluid from bone marrow, sevenfold in platelet-poor plasma, and twofold in platelet extract. Although most of the TGF-β measured was latent, the authors suggested that latent TGF-β was leaking from the megakaryocytes and was activated locally.
The role of TGF-β in TPO-induced reticulin deposition has also been studied in mice deficient in TGF-β (Chagraoui et al, 2002). Bone marrow from TGF-β null (−/−) or wild-type (+/+) mice was isolated, transfected with a retrovirus containing the TPO gene (TPO), and then engrafted into lethally irradiated mice. Animals engrafted with either TGF-β null or wild-type transfected marrow cells had the same elevated platelet level and both developed a myeloproliferative syndrome. Mice that received the wild-type transfected marrow, developed severe marrow fibrosis (reticulin and collagen fibrosis) and osteosclerosis and had a sixfold increase in latent TGF-β levels in the plasma and a fourfold increase in TGF-β levels in the extracellular fluid of the spleen. But animals receiving TGF-β null transfected marrow had no detectable circulating TGF-β and no fibrosis or osteosclerosis.
However, not all of the marrow fibrosis seen in mice transplanted with thrombopoietin-transfected bone marrow can be directly attributed to TGF-β. When murine marrow was transfected with TPO and transplanted into normal mice, extensive marrow fibrosis (reticulin and collagen fibrosis) ensued (Frey et al, 1998). But when transplanted into non-obese diabetic severe combined immunodeficient (NOD/SCID) mice that have reduced monocyte function, no fibrosis was seen. However, this effect seems dependent on the level of TPO. When TPO was expressed at higher levels, fibrosis did occur in NOD/SCID mice; under these conditions of high TPO stimulation, platelets are able to synthesize inflammatory cytokines (Wagner-Ballon et al, 2006). These experiments suggest the importance of some monocyte function in increasing bone marrow connective tissue. Whether monocytes activate latent TGF-β or release some inflammatory cytokine is not known (Rameshwar et al, 1994; Wagner-Ballon et al, 2006).
Taken together, these studies suggest that TGF-β and other growth factors and cytokines associated with megakaryocytes and platelets may be important mediators of increased bone marrow reticulin. Moreover, serum TGF-β may allow for non-invasive assessment of bone marrow fibrosis in general, while serum PIIINP may allow a non-invasive assessment of reticulin deposition (Yanagida et al, 1997; Chagraoui et al, 2002; Shehata et al, 2004).
The JAK2V617F mutation is present in most patients with polycythemia vera, and up to half of patients with essential thrombocythaemia and CIMF (Tefferi, 2006); indeed, the presence of this mutation plays a major role in the revised WHO diagnostic criteria for myeloproliferative diseases (Tefferi et al, 2007). When the mutated JAK2 was expressed in mice via retroviral transduction and transplantation, animals developed increased haemoglobin, splenomegaly, osteosclerosis and bone marrow reticulin/collagen fibrosis, but not thrombocytosis (Lacout et al, 2006; Wernig et al, 2006; Zaleskas et al, 2006). Reticulin fibrosis was strikingly strain-specific with BALB/c mice demonstrating markedly increased reticulin fibrosis compared with C57Bl/6 mice (Wernig et al, 2006), In humans with CIMF, the expression of collagenase genes was not influenced by the JAK2 mutation status, but rather primarily reflected the stage of disease (Bock et al, 2006). In eight CIMF patients with serial bone marrow biopsies, the progression to homozygosity for the JAK2 mutation and the onset of overt fibrosis appeared to be independent events (Hussein et al, 2007). Thus the relationship of the JAK2V617F mutation to bone marrow fibrosis in myeloproliferative diseases remains unclear.
Reticulin is a normal component of the bone marrow but may be increased in bone marrow biopsies in a wide variety of neoplastic and non-neoplastic conditions. When increased above normal, it is a general sign of a bone marrow abnormality and indicates that further investigation is warranted, but it may or may not be a sign of serious, neoplastic disease. In contrast, increased type I collagen in the bone marrow, as detected by the trichrome stain, is less common and is seen primarily with tumours metastatic to the bone marrow or in the late stages of myeloproliferative diseases. Unlike increased reticulin, it may not be reversible and represents a more diagnostically significant finding.
Although it was originally thought that increases in bone marrow fibrotic tissue contributed to the haematopoietic abnormalities seen in certain diseases, recent studies suggest that bone marrow fibrosis is more likely to be a result of the underlying cellular abnormalities rather than a cause. The pathophysiology of increased bone marrow reticulin and collagen deposition is just beginning to be elucidated, but it is becoming clear that TGF-β and other factors associated with megakaryocytes and platelets play important roles.
Going forward, it is likely that further insights will be gained into the relationship between various disease processes and bone marrow reticulin or collagen deposition. Further clarification of the role of factors, such as TGF-β, may lead to the development of non-invasive tests for bone marrow fibrosis. This may help to improve our understanding of disease processes as well as increase the opportunity for earlier clinical intervention. Potential pharmacological inhibition of bone marrow fibrosis may be clinically useful.
The authors thank Amy Lindsay, PhD for writing assistance. This work was supported in part by grants from the NIH [HL82889 (DK) and HL72299 (DK)] and from Amgen, Inc., Thousand Oaks, CA.