Article Text

Download PDFPDF

Extended report
Granulocyte-macrophage colony-stimulating factor is a key mediator in inflammatory and arthritic pain
  1. Andrew D Cook1,
  2. Jarrad Pobjoy1,
  3. Shannon Sarros1,
  4. Stefan Steidl2,
  5. Manuela Dürr2,
  6. Derek C Lacey1,
  7. John A Hamilton1
  1. 1Department of Medicine, Arthritis and Inflammation Research Centre, University of Melbourne, Victoria, Australia
  2. 2MorphoSys AG, Martinsried/Planegg, Germany
  1. Correspondence to Dr Andrew D Cook, Department of Medicine, University of Melbourne, The Royal Melbourne Hospital, Parkville, VC 3010, Australia; adcook{at}


Objectives Better therapies are needed for inflammatory pain. In arthritis the relationship between joint pain, inflammation and damage is unclear. Granulocyte-macrophage colony-stimulating factor (GM–CSF) is important for the progression of a number of inflammatory/autoimmune conditions including arthritis; clinical trials targeting its action in rheumatoid arthritis are underway. However, its contribution to inflammatory and arthritic pain is unknown. The aims of this study were to determine whether GM–CSF controls inflammatory and/or arthritic pain.

Methods A model of inflammatory pain (complete Freund's adjuvant footpad), as well as two inflammatory arthritis models, were induced in GM–CSF−/− mice and development of pain (assessment of weight distribution) and arthritic disease (histology) was assessed. Pain was further assessed in a GM–CSF-driven arthritis (methylated bovine serum albumin/GM–CSF) model and the cyclooxygenase-dependence determined using indomethacin.

Results GM–CSF was absolutely required for pain development in both the inflammatory pain and arthritis models, including for IL-1-dependent arthritic pain. Pain in a GM–CSF-driven arthritis model, but not the disease itself, was abolished by the cyclooxygenase inhibitor, indomethacin, indicating separate pathways downstream of GM–CSF for pain and arthritis control.

Conclusions GM–CSF is key to the development of inflammatory and arthritic pain, suggesting that pain alleviation could result from trials evaluating its role in inflammatory/autoimmune conditions.

View Full Text

Statistics from


Relief from inflammatory pain, for example that associated with arthritis, represents a significant unmet medical need. Neuronal sensitisation is thought to arise in response to the actions of mediators, which during inflammatory pain can be derived from many types of immune cells.1 Prostaglandins and sympathetic amines are key mediators of this process and their release can be stimulated by cytokine cascades.2 Increasing evidence supports the hypothesis that cytokines are not only inflammatory mediators but also should be regarded as neuromodulators, particularly as cytokines can act directly via receptors on nociceptive terminals that innervate inflamed tissues.1 Blockade of proinflammatory cytokines, such as tumour necrosis factor (TNF) and interleukin (IL)-1β, reduces hyperalgesia in inflammation models and anti-TNF therapies in rheumatoid arthritis (RA) lead to pain reduction.3

The relationship between pain, inflammation and tissue damage, for example in arthritis, is complex and not well understood. Non steroidal anti-inflammatory drugs (NSAID) can alleviate pain yet there is minimal evidence of their preventing arthritis progression. Also, TNF blockade in RA results in a very rapid and profound downregulation of nociceptive activity, even before effects on joint swelling and other manifestations of inflammation.3

Granulocyte-macrophage colony-stimulating factor (GM–CSF) was originally defined as a haemopoietic growth factor. However, it can act on mature myeloid cells as a pro-survival and ‘activating’ factor,4 as a proinflammatory cytokine4–7 and in dendritic cell function.8 Its depletion can have profound effects on disease severity and progression in many inflammation/autoimmune models.6 ,9–16 Networks linking GM–CSF and other proinflammatory cytokines, such as TNF, IL-1β and IL-23, have been found,6 ,9 ,11 ,12 ,15 ,16 although evidence for a dissociation from TNF action has also been reported.13 It was recently reported that GM–CSF plays a non-redundant role in T helper (Th) type 17-mediated inflammation,15 ,16 with one report15 stating for experimental autoimmune encephalomyelitis (EAE) ‘of all known T cell cytokines, GM–CSF seems to be the only one absolutely essential for endowing T cells with pathogenic properties’.

The GM–CSF receptor is also found on neurons and it can have pro-survival and stimulatory functions for this cell type.17 It has been published that GM–CSF can act as a pain mediator in a mouse model of bone tumour-induced pain and it directly activated receptors located on primary afferent nerve fibres to produce peripheral sensitisation.17 There is no information as to whether GM–CSF controls inflammatory or arthritic pain, and this is the subject of the current study.

Materials and methods


GM–CSF gene-deficient (GM–CSF−/−) mice were backcrossed onto the C57BL/6 background for 12 generations.10 ,18 Mice of both sexes (8–12 weeks) were used. GM–CSF−/− and wild-type control mice were age and sex-matched. All experiments were approved by the University of Melbourne Animal Ethics Committee.

Complete Freund's adjuvant model

Inflammatory pain was induced by a single intraplantar injection of 20 μl of complete Freund's adjuvant (CFA; Difco, Basel, Switzerland) into the left hind footpad.19 Paw swelling was measured using spring callipers (Mitutoyo, Tokyo, Japan), accurate to 0.01 mm.

Antigen-induced arthritis

Mice were immunised (day 0) with methylated bovine serum albumin (mBSA; Sigma-Aldrich, St Louis, Missouri, USA), emulsified in CFA, intradermally in the base of the tail.20 Arthritis was induced 7 days later by an intra-articular injection of mBSA into the right knee, the left knee being injected with phosphate-buffered saline (PBS).

mBSA-induced arthritis models

Monoarticular arthritis was induced in C57BL/6 mice, as before10 ,21 but with slight modifications, by intra-articular injection of 100 µg mBSA in 10 µl PBS into the right knee on day 0, the left knee being injected with PBS, followed by a subcutaneous injection in the scruff of the neck on days 0–2, of either IL-1β (250 ng; R&D Systems, Minneapolis, Minnesota, USA), GM–CSF (500 ng; R&D Systems) or saline. Mice were killed (day 7) and knee joints were collected for histology.

Pain readings

As an indicator of pain, the differential distribution of weight over a 5-s period between the inflamed paw or limb relative to the non-inflamed paw or limb was measured using an incapacitance metre (IITC Life Science Inc, California, USA).22 Mice were acclimatised to the incapacitance metre on at least three occasions before the start of the experiment. Three measurements were taken for each time point and averaged.


At termination, the knee joints were removed, fixed, decalcified and paraffin embedded.9 ,14 ,20 Frontal sections (5 μm) were stained with H&E. For antigen-induced arthritis (AIA), infiltration of cells, cartilage damage and bone erosions were scored separately from 0 (normal) to 3 (severe).20 For the mBSA/IL-1 and mBSA/GM–CSF models, cellular infiltration, synovitis (synovial hyperplasia), pannus formation, cartilage damage and bone erosions were scored separately from 0 (normal) to 5 (severe).21

Prostaglandin E2 measurement

Prostaglandin E2 (PGE2) levels were measured in knee joint washouts9 using a mouse PGE2 ELISA kit (Cusabio Biotech Co., China). The ELISA was sensitive down to 0.32 pg/ml.


For pain readings, Student's t test or two-way analysis of variance were used, and for histological scores, the Mann–Whitney two-sample rank test or two-way analysis of variance were used; values are expressed as the mean±SEM; p≤0.05 was considered statistically significant.


Inflammatory pain is GM–CSF dependent

We first tested whether the absence of GM–CSF could modulate pain in the CFA footpad model, which is widely used for inflammatory pain, using the incapacitance metre, as previously reported.23 This method for measuring pain does not rely on the application of any additional experimental stimulus or withdrawal reflexes. We used GM–CSF−/− mice, which have essentially an intact myeloid system.18

Following intraplantar injection of CFA into the left footpad, mice developed local swelling, which was similar in magnitude in C57BL/6 (wild-type) and GM–CSF−/− mice (figure 1A). Treatment of CFA-injected wild-type mice with indomethacin had no effect on the swelling (figure 1A), noting, however, that its first injection was at 24 h when maximum swelling was already evident. There was no swelling of the contralateral (right) paw (data not shown). In contrast to the increase in swelling, assessment of weight distribution showed that, while wild-type mice developed pain between 24 and 72 h post-CFA injection, GM–CSF−/− mice did not (figure 1B). Treatment of wild-type mice with the NSAID, indomethacin, alleviated the pain (figure 1B), validating this weight distribution method as monitoring pain.22

Figure 1

Granulocyte-macrophage colony-stimulating factor (GM–CSF) is required for inflammatory pain. C57BL/6 and GM–CSF−/− mice were injected with complete Freund's adjuvant (CFA) into the left footpad. The degree of (A) paw swelling (indicative of inflammation) and (B) change in weight distribution (indicative of pain) were measured. Weight distribution is expressed as a percentage of the weight placed on the injected paw versus the control paw. A value of 100 indicates no pain and a decrease in weight applied to the injected paw (ie, a reading  <100) indicates pain.22 Indomethacin (indo) (1 mg/kg) was injected intraperitoneally 1 h before each reading from 24 h onwards. Results are means±SEM; n=12 mice/group. *p<0.05, **p<0.001, GM–CSF−/− versus C67BL/6 mice; #p=0.05, C57BL/6 mice with or without indomethacin.

Therefore, GM–CSF is absolutely required for the development of inflammatory pain in this commonly used model.

Inflammatory arthritis pain is GM–CSF dependent

We next examined whether GM–CSF could also control pain in an inflammatory disease model involving tissue remodelling. We measured pain in monoarthritis models by the incapacitance metre as the ratio of the weight bearing on an arthritic limb relative to that on a non-arthritic one, and is a common method of determining pain in such models22—again a relative decrease in weight bearing indicates pain.

Antigen-induced arthritis

We first examined the GM–CSF dependence of disease and pain in the AIA model as it is widely employed in arthritis pain studies as a model for RA.24

Following systemic antigen (mBSA) priming, at 1 week post intra-articular mBSA challenge wild-type mice showed significant swelling of their mBSA-injected knee joint compared with GM–CSF−/− mice, in whom only mild swelling was evident (2.6±0.2 vs 1.0±0.3, p=0.01; data not shown). There was significantly more inflammatory cell infiltration (p=0.02; figure 2A,B) in the arthritic joints of wild-type mice compared to those of GM–CSF−/− mice, but little joint destruction at this time point. By 6 weeks post intra-articular injection, ie, during the chronic phase, wild-type mice showed increased joint destruction, including bone erosion (figure 2A,B), compared to week 1, with GM–CSF−/− mice having significantly less. Therefore the disease severity for both the acute inflammatory and chronic destructive phases was significantly lower in GM–CSF−/− mice.

Figure 2

Granulocyte-macrophage colony-stimulating factor (GM–CSF) is required for full expression of disease and pain in antigen-induced arthritis (AIA). C57BL/6 and GM–CSF−/− mice were immunised intradermally for AIA on day −7 (methylated bovine serum albumin (mBSA) in adjuvant), then challenged intra-articularly (i.a.) with mBSA on day 0. (A) and (B) Mice were killed at 1 and 6 weeks post intra-articular injection. (A) Representative histological pictures of knee joints (H&E stain); (B) quantification of histology, showing cell infiltration (indicative of inflammation) and bone erosion (indicative of joint destruction); (C) change in weight distribution (pain) measured over time expressed as a percentage of the weight placed on the arthritic limb versus the control limb.22 Results are means±SEM; n=5–10 mice/group. *p<0.01, **p<0.005, ***p<0.001, C57BL/6 versus GM–CSF−/− mice. This figure is only reproduced in colour in the online version.

We next assessed whether GM–CSF−/− mice develop pain. Following the initial immunisation of mBSA in adjuvant, wild-type mice showed no articular pain. However, within 1 day following the intra-articular injection of mBSA, significant pain was evident (figure 2C), which remained so for approximately 2 weeks. For GM–CSF−/− mice, following the intra-articular injection of mBSA, the level of pain was significantly lower and more transient, disappearing at approximately 1 week at which time wild-type mice still had maximal pain (figure 2C). Overall, there was a significant difference in the degree of pain between wild-type and GM–CSF−/− mice (p<0.0005).

The development of both arthritis and pain in this model is heavily dependent on GM–CSF.

mBSA/IL-1 arthritis

We next used the monoarticular mBSA/IL-1 model, which involves intra-articular mBSA injection and systemic IL-1 administration.10 ,21 It also allows cytokine interactions and mechanisms to be dissected more easily. We have shown the disease in this IL-1-driven model to be GM–CSF dependent by using both GM–CSF−/− mice and wild-type mice administered neutralising monoclonal antibodies to GM–CSF.10 Arthritic pain has not been examined in this model.

Arthritis was induced in wild-type mice and knee pain assessed over a 7-day period. The left knee received mBSA while the contralateral right knee received saline. Following systemic IL-1 administration, significant pain in the left knee was evident from approximately day 4 after IL-1 administration (p<0.001, compared to t=0; figure 3A). Indomethacin given 1 h before each pain reading blocked the pain (p<0.0001), again validating the fact that pain is being monitored.22 GM–CSF−/− mice showed no detectable left knee pain throughout, with wild-type mice having significantly more pain from day 4 onwards (p<0.0001; figure 3A).

Figure 3

Requirement of granulocyte-macrophage colony-stimulating factor (GM–CSF) for pain in methylated bovine serum albumin (mBSA)/IL-1 arthritis. C57BL/6 and GM–CSF−/− mice were immunised for mBSA/IL-1 arthritis. (A) Change in weight distribution (pain) (figure 2). Indomethacin (indo) (1 mg/kg) was injected in C57BL/6 mice 1 h before each reading; (B) representative histological pictures of knee joints (H&E stain); (C) quantification of histology (see Materials and methods section). Results are means±SEM; n=12 mice/group. *p<0.05, **p=0.004, ***p<0.0001, GM–CSF−/− versus C57BL/6 mice; #p<0.01, ##p<0.001, C57BL/6 mice with or without indomethacin. This figure is only reproduced in colour in the online version.

Therefore, the pain in this IL-1-driven inflammatory arthritis model is completely GM–CSF dependent.

In the left knee GM–CSF−/− mice also showed significantly less inflammatory cell infiltration (p=0.02), synovitis (p=0.04) and cartilage damage (p=0.03; figure 3B). The overall severity of arthritis (the sum of each histological feature)10 ,21 was significantly lower in GM–CSF−/− mice compared with that in wild-type mice (figure 3B,C). Interestingly, indomethacin treatment throughout arthritis development in wild-type mice had no effect on disease severity (p=0.004, figure 3B,C) even though it completely blocked pain induction (see Discussion section).

GM–CSF induces cycloxygenase-dependent arthritic pain

While the findings above show that the pain in the above models depends on the presence of GM–CSF, they do not inform as to whether GM–CSF itself can induce inflammatory pain; they also do not tell us whether GM–CSF could be involved in the cyclooxygenase dependence of the pain in the mBSA/IL-1 model (figure 3). To address these questions we took advantage of the availability of a monoarticular arthritis model developed by us in which systemic GM–CSF administration replaced that of IL-1 in the mBSA/IL-1 model.21 GM–CSF exacerbates the mild synovitis induced by intra-articular mBSA21—in other words, the arthritis is driven by exogenous GM–CSF, thereby allowing easier analysis of its mode of action, including the involvement of any downstream mediators. Joint pain has not been monitored in this mBSA/GM–CSF arthritis model.

Pain was measured over a 7-day period. mBSA/GM–CSF-injected mice developed pain in the mBSA-injected joint relative to the contralateral joint (figure 4A), which was similar in magnitude to what was found in mBSA/IL-1-injected mice (p<0.0005, mBSA/IL-1 or mBSA/GM–CSF vs mBSA/saline) and consistent with the GM–CSF dependence of the pain in the latter model (figure 3A). As a negative control, mBSA/saline-injected mice did not show joint pain (figure 4A). mBSA/GM–CSF-injected mice were also treated with indomethacin (1 mg/kg) from day 5 onwards, once significant pain was evident. Indomethacin was able to reverse the GM–CSF-induced pain, again indicating eicosanoid involvement (p<0.01, GM–CSF with or without indomethacin, figure 4A).

Figure 4

Mice with granulocyte-macrophage colony-stimulating factor (GM–CSF)-driven arthritis develop cyclooxygenase-dependent pain. C57BL/6 mice received methylated bovine serum albumin (mBSA) intra-articularly and GM–CSF, interleukin (IL)-1 or saline subcutaneously. In mBSA/GM–CSF mice, once significant pain was evident, indomethacin (indo) (1 mg/kg) was injected 1 h before subsequent readings (from day 5 onwards). (A) Change in weight distribution (pain) (figure 2); (B) representative histological pictures of knee joints (H&E stain); (C) quantification of histology (see Materials and methods section); (D) Prostaglandin E2 (PGE2) levels in joint washouts. Results are means±SEM; n=8–12 mice/group. ND=not detected. *p<0.05, **p<0.01, mBSA/GM–CSF, mBSA/GM–CSF plus indomethacin, or mBSA/IL-1 versus mBSA/saline; #p<0.01, mBSA/GM–CSF versus mBSA/saline. This figure is only reproduced in colour in the online version.

mBSA/GM–CSF-injected mice developed arthritis to a similar magnitude as mBSA/IL-1-injected mice, with both groups of mice having significantly greater cellular infiltration, synovitis and pannus formation than mBSA/saline-injected mice, ie, there is amplification of a mBSA-induced synovitis21 (p<0.01, for each histological feature; figure 4B,C). Treatment of mBSA/GM–CSF-injected mice with indomethacin once again had no effect on arthritis development even though pain was abolished. We therefore investigated whether PGE2 levels were increased in the joints following mBSA/GM–CSF-injection. Joint PGE2 levels were significantly increased in mBSA/GM–CSF-injected mice but not in the mBSA/saline-injected controls when measured on day 5 (p<0.05, mBSA/GM–CSF vs mBSA/saline, figure 4D). As expected, indomethacin treatment reduced the GM–CSF-stimulated levels. IL-1 administration led to a similar increase in PGE2 as GM–CSF.

These findings indicate that the GM–CSF-driven pain, but not disease, in this model requires a cyclooxygenase product(s), such as PGE2, for its manifestation.


We demonstrated above for the first time that GM–CSF is absolutely required for pain development in an often used inflammatory pain (CFA footpad) model as well as in the AIA and the more acute (mBSA/IL-1) arthritis models. We also showed that GM–CSF-driven arthritic pain, but not the disease itself, is cyclooxygenase dependent, indicating that they are able to be dissociated.

For the inflammatory pain model, there was no pain in the absence of GM–CSF. Seeing that both IL-1 and a cyclooxygenase product(s) contribute to the pain in this CFA model,19 it could be that GM–CSF provides the common link between these mediators both in this model as well as in the mBSA/IL-1-induced arthritis model. It would be interesting to know how critical GM–CSF is in other inflammatory pain models. The absence of GM–CSF was also found to reduce dramatically the severity of both pain and arthritis development in the AIA model. This model can now be added to the list of GM–CSF-dependent inflammatory arthritis models.4 ,9 ,10 ,13 ,14 ,18 Full-blown expression of this arthritis requires IL-23-mediated regulation of helper T cells and γδ T cells.25 As GM–CSF can be produced by IL-23-responsive helper T cells, such as Th1 and Th17 cells,15 ,16 it could be that T cells are again the source of the pathogenic GM–CSF in the AIA model as in mBSA/IL-1 arthritis and EAE.15 ,16 ,26

In the mBSA/IL-1 arthritis model, in addition to disease severity, the IL-1-driven pain was completely dependent on the presence of GM–CSF; it could be that this dependence applies to some of the other examples of hyperalgesia reported for IL-12 ,27 and for other cytokines.2 GM–CSF may be an important downstream mediator in IL-1 biology in general because it can be induced by IL-1 in many cell types.6 ,27 Networks linking GM–CSF and cytokines, such as IL-1 and IL-23, are beginning to emerge, highlighting the key role of GM–CSF as a pathogenic cytokine in inflammation/autoimmunity,4 ,6 ,11 ,13 ,15 ,16 with GM–CSF being the only known T-cell-derived cytokine with a non-redundant function in the initiation of EAE.15 The complete suppression by indomethacin of the pain but not arthritis development in the mBSA/IL-1 and mBSA/GM–CSF models is similar to what is often reported clinically with NSAID.

Even though we have now the first definitive evidence for a non-redundant role for GM–CSF in inflammatory pain, there are several critical unknowns regarding how and when GM–CSF can have such algesic effects, for example, the relative contribution of its direct17 or indirect neuronal sensitisation via cells/mediators, the cell type(s)4 and mediators (cytokines, prostanoids) governing GM–CSF formation and its algesic action, the relative contribution of peripheral and central sensitisation, and the nature of the GM–CSF-dependent noiceptive system. Our GM–CSF/mBSA model could be conveniently exploited further to address these questions using eicosanoid-dependent pain as the readout. The requirement of an eicosanoid and of a time delay of 3–4 days for a change in weight distribution to be manifested, ie, the time period during which the mBSA-induced myeloid and lymphocyte joint infiltration and synovitis occur,21 suggest an indirect mechanism for nociceptor sensitisation via a target inflammatory cell, at least in this model.

Preliminary results from a RA trial neutralising the GM–CSF receptor support an important role for GM–CSF as an inflammatory mediator.28 From the above, it is proposed that GM–CSF is also a key mediator in inflammatory pain, such as that associated with arthritic disease.


The authors would like to thank Jennifer Davis and Lara Mizhirisky for assistance with the maintenance and care of the mice.


View Abstract


  • Contributors ADC designed and supervised the study, performed some of the experiments, analysed the data and wrote the paper. JP and SSa performed some of the experiments and MD and SSt were involved in the design of the experiments and writing of the paper. DCL supervised and perfomred some of the experiments and JAH designed and supervised the study and wrote the paper.

  • Funding This work was supported by grants from Morphosys AG and the National Health and Medical Research Council (NHMRC), and by a NHMRC Senior Principal Research Fellowship (JAH).

  • Competing interests SSt is a full-time employee of MorphoSys AG, Germany. JAH has received consulting fees from MorphoSys AG, Germany (less than US$10 000 a year). The University of Melbourne has licensed to MorphoSys AG, Germany, patented technology relating to therapeutically targeting granulocyte-macrophage colony-stimulating factor, for which licensing fees and milestone payments have been made. The other authors state no conflict of interests.

  • Ethics approval All experiments were approved by the University of Melbourne Animal Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.