This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
Author's personal copy
Macrophage colony stimulating factor: Not just for
macrophages anymore! A gateway into
Thomas G. Douglassa, Lara Driggersb, Jian Gang Zhangb,c, Neil Hoab,
Christina Delgadob, Christopher C. Williamsb, Qinhong Danb,
Ramon Sanchezb, Edward W.B. Jeffesd, H. Terry Wepsicb, Michael P. Myerse,
Kirston Kothsf, Martin R. Jadusb,c,g,⁎
aBiology Department, California State University Long Beach, 1250 Bellflower Blvd, Long Beach CA 90840, United States
bDepartment of Diagnostic and Molecular Medicine, Box 151 Veterans Affairs Medical Center, 5901 E. 7th Street, Long Beach,
CA 90822, United States
cPathology Department, University of California, Irvine, CA 92697, United States
dDepartment of Dermatology, Veterans Affairs Medical Center, Long Beach, CA 90822, United States
eChemistry and Biochemistry Department, California State University Long Beach, 1250 Bellflower Blvd, Long Beach CA
90840, United States
fIndependent Biotechnology Consultant, 2646 Mira Vista Drive, El Cerrito, CA 94530, United States
gNeuro-Oncology Program, Chao Comprehensive Cancer Center, University of California, Irvine. Orange, CA 92868, United States
Received 19 April 2008; accepted 21 April 2008
Foundation via the University of California at Irvine Cancer Research Program (MRJ). Faculty–Student Collaborative Research Seed Grant (MM).
⁎ Corresponding author. Box 113 Diagnostic and Molecular Medicine Healthcare Group. Veterans Affairs Medical Center, 5901 East 7Th Street,
Long Beach, CA 90822, United States. Tel.: +1 562 826 8000x4079.
E-mail address: email@example.com (M.R. Jadus).
Macrophage colony stimulating factor (M-CSF, also called colony stimulating factor-1) has
traditionally been viewed as a growth/differentiation factor for monocytes, macrophages, and
some female-specific tumors. As a result of alternative mRNA splicing and post-translational
processing, several forms of M-CSF protein are produced: a secreted glycoprotein, a longer secreted
form containing proteoglycan, and a short membrane-bound isoform. These different forms of M-CSF
all initiate cell signaling in cells bearing the M-CSF receptor, called c-fms. Here we review the biology
of M-CSF, which has important roles in bone physiology, the intestinal tract, cancer metastases to the
bone, macrophage-mediated tumor cell killing and tumor immunity. Although this review
concentrates mostly on the membrane form of human M-CSF (mM-CSF), the biology of the soluble
forms and the M-CSF receptor will also be discussed for comparative purposes. The mechanisms of
1567-5769/$ - see front matter © 2008 Elsevier B.V. All rights reserved.
International Immunopharmacology (2008) 8, 1354–1376
Author's personal copy
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Molecular biology and biochemistry of human M-CSF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
M-CSF receptors and its homologues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
M-CSF signal transduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
M-CSF involvement in bone physiology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Intestinal tract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A paradox: why doesn't an immunosuppressive hybridoma form a tumor in vivo? . . . . . . . . . . . . . . . . . . . .
Mechanisms of macrophage-mediated killing of mM-CSF transduced tumor cells . . . . . . . . . . . . . . . . . .
The molecular mechanisms by which monocytes/macrophages kill mM-CSF-expressing glioma cells . . . . . . . . .
Paraptotic tumor cells produce “danger signals” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Immunized T cells are produced once mM-CSF-expressing tumor cells are rejected . . . . . . . . . . . . . . . .
A novel T cell epitope is identified as alt-M-CSF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Therapeutic vaccination with mM-CSF-expressing T9 glioma cells . . . . . . . . . . . . . . . . . . . . . . . . . .
Combination therapies using an mM-CSF vaccine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Summary/conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The colony stimulating factors (CSF) were initially described
in the late 1960s and early 1970s as glycoproteins that
induced the clonal growth of various hematopoietic lineages
from the bone marrow . A variety of these factors were
named for their utilitarian functions: M-CSF stimulated
macrophages, G-CSF stimulated granulocytes, GM-CSF sti-
mulated both granulocytes and macrophages, while multi-
CSF induced the growth of all hematopoietic different cell
types. Richard Stanley and his colleagues first identified the
undefined M-CSF, purified to homogeneity, and named it
colony stimulating factor-1 (CSF-1) [2–5] to reflect a better
characterization of this protein. Initially, the human form of
M-CSF was isolated from urine. Later it was discovered that a
characterization. For this review we will use the functional
name M-CSF. There have been several excellent general
reviews previously written about this cytokine [6–10].
Recombinant M-CSF (called Lanimostim/MacroTac) is used
clinically in bone marrow transplantation patients [11,12],
whose innate immune system has not been fully restored,
and consequently suffer from recurrent fungal and bacterial
infections due to the lack of myeloid cells. M-CSF stimulates
the growth of mononuclear phagocytes from the hemato-
poietic stem cells, producing more precursor monocytes and
macrophages. Infused M-CSF also activates these monocytoid
cells to become better phagocytic cells, thereby clearing the
microbes by directly engulfing the pathogens. One major
toxicity that limits the therapeutic use of this cytokine is
thrombocytopenia . This should not be surprising that the
M-CSF-activated monocytes/macrophages phagocytosized
the platelets, which have the same approximate size as the
microbes. The M-CSF-induced thrombocytopenia is reversi-
ble, following cessation of the treatment.
M-CSF primarily stimulates the growth of macrophages
and resident macrophages of local tissue (Kupffer cells in
liver, microglial cells in bone, mesangial cells in the kidney,
dendritic cells. In the skin, the Langerhans cells are
stimulated [14,15], while in the blood and lymph nodes,
plasmacytoid dendritic cells  are produced. M-CSF is
normally found in detectable levels (2.4 ng/ml or 120 units/
ml)in theserumofhealthy individuals .M-CSF expression
is elevated during pregnancy, up to 400 units/ml, from 9 to
33 weeks of gestation . Low M-CSF levels were associated
levels were linked with a pathology of the mother called pre-
eclampsia . This linkage with pregnancy ultimately
proved to be a particularly useful in the fertile area of M-
CSF research over the last 22 years. M-CSF plays important
roles in many gynecological aspects such as egg follicle
fetal and placental growth. Since these topics have been
covered elsewhere [21–30], we will not review it here.
M-CSF is speculated to be a useful biomarker for a number
of cancers (reviewed by Khatma, ), including pancreatic
, colorectal [33,34], breast [35,36], and ovarian cancers
[37,38]. In acute rejection of kidney transplant patients ,
circulating M-CSF levels correlated with poor prognosis. It is
not yet clear whether the high levels of M-CSF are the cause
or the effect of these diseases.
the biological effects of the membrane-bound M-CSF reveal that this cytokine is unexpectedly
involved in many complex molecular events. Recent experiments suggest that a tumor vaccine
based on membrane-bound M-CSF-transduced tumor cells, combined with anti-angiogenic
therapy, should be evaluated further for use in clinical trials.
© 2008 Elsevier B.V. All rights reserved.
1355Macrophage colony-stimulating factor: Not just for macrophages anymore!
Author's personal copy
M-CSF also has a dark side. Some tumor cells such as
MiaPaca and PANC-1 pancreatic cancer cells make M-CSF
. Other cancer types, including: prostate [41,42],
Hodgkins lymphoma , J6-1 AML leukemia , cervical
, endometrial [46–48], breast [36,49–52] and ovarian
[37,38,52] make both this cytokine and its receptor. So the
possibility does exist that many of these cancer cells use M-
CSF as an autocrine growth factor. Here the M-CSF enhances
growth and aggressiveness, while stimulating the tumor-
infiltrating macrophages into a pro-tumor phenotype by
stimulating macrophages . These activated macrophages
in turn make angiogenic growth factors, proteases that
facilitate tumor metastases and immunosuppressive mole-
cules. Hence, M-CSF is often viewed as a pro-tumor cytokine.
We have previously published a review of our earlier work
 dealing with the anti-tumor properties of a membrane-
bound form of human M-CSF, abbreviated as mM-CSF. There
we speculated on the differences of macrophages activated
by either soluble M-CSF or the mM-CSF. In the present review,
we detail the basic biology of M-CSF, while highlighting some
of the complex biological pathways that we have investi-
gated during our explorations of the distinct biology of mM-
2. Molecular biology and biochemistry of human
The M-CSF gene resides at 1p21–p13 in man and on
chromosome 3, (51 cM) in mice. Presently the NCBI describes
74 single nucleotide polymorphisms (SNPs) in the human M-
CSF gene. Three SNPs exist in the general population, the
other SNPs arise from different tumor types (http://www.
cDNAs encoding M-CSF have been cloned [55–59]. Histori-
Although up to 19 different transcripts are predicted by
confidence=B&uniid=no). Four distinct forms of M-CSF proteins
can be derived from these five major transcripts. Fig. 1 shows
the different mRNAs and the different isoforms of M-CSF
proteins that have been characterized. Also included in the
figure are the various names/nomenclatures that have been
used over the years to describe each of these proteins. The
various cloned M-CSF forms were named in order of their
discovery: M-CSF-α, β, and γ.
After the primary mRNA is made, alternative splicing
pathways produce different lengths of M-CSF transcripts
[40,60,61]. The best known form of M-CSF is the secreted or
soluble form (sM-CSF). Three mRNAs (4, 3.7 and 2.6 kb forms)
lead to this secreted glycosylated cytokine, but there are 2
2.6 kb transcripts produce a precursor protein of 554-amino
amino acid protein. The 3.7 kb transcript produces a 438-
amino-acid protein. This form also is processed into the same
The 3.1 and 1.6 kb transcripts produce a shorter form of M-
CSF that remains on the cell-surface or membrane of the
producing cells for extended times. Hence the name
membrane-bound or cell-surface M-CSF was given.
The general composition of the M-CSF protein consists of a
32-amino-acid signal peptide, a 149-amino-acid receptor
binding domain, a variable spacer region, a 24-amino-acid
transmembrane region with a 35-amino-acid cytoplasmic
tail. Fig. 2 shows the processing of the various forms of M-CSF
following synthesis [62–64]]. As the M-CSF monomers are
translated and channeled into the lumen of the endoplasmic
reticulum (ER), they form a number of intrastrand disulfide
bonds, depending upon the splice form. One interstrand
disulfide linkage is present in all of the stable mature dimeric
forms. As the nascent protein enters the lumen of the ER, the
32-amino-acid leader sequence is quickly removed by a
signal peptidase. During the post-translational modification
process O-linked and N-linked carbohydrate moieties are
added. As the M-CSF traverses the Golgi complex, a protease
(M-CSF-β convertase) cuts at or near Arg223, releasing the
soluble form of this M-CSF . The secretory vesicle
eventually fuses with the membrane, permitting the freshly
cut M-CSF to be released as the soluble cytokine.
An additional form of soluble of M-CSF exists as a
proteoglycan form, pM-CSF [65,66]. Production of this form
proceeds in the identical manner as the 554-amino-acid
glycosylated form encoded by the 4 and 2.6 kb mRNA.
Chondroiten sulfate glycosaminoglycans are added to the
Ser277residue. As the M-CSF transits through the Golgi, it is
cleaved by a second protease (M-CSF-α convertase). This
second protease cleaves closer to the membrane. As a result
of the actions of M-CSF-α convertase, this version of M-CSF
possesses a longer portion of the M-CSF mature N-terminal
sequence, generating a molecule with a molecular weight
between 100 and 250 kD. The pM-CSF displays some
additional versatility in that it can bind to either basic
fibroblast growth factor [67,68] or low density lipoproteins
(LDL) . Thus, pM-CSF has important ramifications for
angiogenesis, wound healing and atherosclerosis.
The 3.1 and 1.6 kb transcripts form when exon 6 is excised
from the primary transcript. This spliced mRNA encodes a 256-
convertase sensitive region, this shorter M-CSF form is not
vesicle fuses with the membrane, whereas the soluble M-CSF
diffuses away from the synthesizing cell. This cell-surface or
membrane M-CSF (mM-CSF) lacks the Ser277residue, where the
is added to this shorter 256-amino-acid isoform. The mM-CSF
form does acquire O-linked and N-linked glycosylation.
Membrane M-CSF is eventually shed from the membrane.
This active shedding process allows this membrane form to
become a chemoattractant for cells possessing M-CSF
receptors. A 15-amino-acid juxtamembrane region is where
the M-CSF-α convertase cleaves the mM-CSF . Deng et al.
[71–74] showed that by site-directed mutagenesis, the
following substitutions: asp-asp-asp, for ser158-ser159-ser160
or ile for leu163, pro for gln164all significantly reduced the
cleavage of mM-CSF. The elimination of pro161-gln162-leu163-
gln164-glu165within the juxtamembrane region of mM-CSF
completely prevented the M-CSF-α convertase from success-
fully cleaving the membrane M-CSF. The probable mechanism
for this inhibition of the cleavage is physical prevention of
the active protease from reaching the cleavage region. The
proteolytic processing of mM-CSF can be accelerated by
1356T.G. Douglass et al.
Author's personal copy
adding phorbol esters, presumably increasing synthesis of
the M-CSF-α convertase .
Tumor necrosis factor (TNF) is another cytokine that is
processed as a membrane intermediate. TNF forms trimers and
via a long N-terminal leader. There is evidence that the same
protease that cleaves TNF (known as TNF converting-enzyme
as an M-CSF-α convertase . TNF convertase cleaves pro-TNF
after the residues LAQA, generating the terminal VRSS N-
terminus of mature TNF. The release of mM-CSF occurs after
receptor . Matrix metalloproteinase-9 has also been
reported to cut mM-CSF . The possibility does exist that
other proteases can also cleave mM-CSF as well.
As a result of the different pathways of M-CSF synthesis,
cleavage, and post-translational modification, it is not
surprising that the molecular weights previously reported
for the M-CSF proteins have spanned a wide range from 24,
30, 45 and 60–70 kD , 100, 250 and 375 kD forms .
3. M-CSF receptors and its homologues
single high affinity, type III tyrosine kinase receptor whose
molecular weight ranges from 130–170 kD. On a molar basis
when all soluble M-CSF forms are tested for their ability to
stimulate cell growth, all these species display equal stimula-
tory activities . The M-CSF receptor (designated as CD115)
was identified as the cellular homologue (c-fms) of the active
viral oncogene, v-fms, found in the feline Susan McDonough
sarcoma (SM-FeSV) retrovirus [79,80]. The gene for c-fms is
amino-acid glycosylated protein (the homologous murine M-
M-CSF). An extracellular 512-amino-acid region possesses five
immunoglobulin-like domains that are responsible for physical
binding to the cytokine. This extracellular region is joined to a
25-amino-acid transmembrane region. An intracellular 435-
amino-acid sequence in the C-terminal region contains the
autocatalytic kinase domain. There are 10 known SNPs of this
variantsof c-fms, duetodifferent processing pathways. The c-
fms found on macrophages/monocytes is a shorter version of
the receptor, while in trophoblasts, c-fms is 15 amino acids
longer on its N-terminus . There are apparently no
significant differences in the ligand affinity between these
two receptor isoforms.
The oncogenic version of this receptor, v-fms, possesses
three apparently significant mutations when compared to c-
splicing. The 1.6 and 3.1 kb transcripts encode 256-amino-acid proteins. This version of M-CSF has been given the following names over
the years: short M-CSF, M-CSFα, CSF-1256, membrane M-CSF, cell-surface M-CSF, and slow released M-CSF. The 3.7 kb transcript encodes
a 438-amino-acid form and has been given the name M-CSFγ or secreted M-CSF. The 4 and 2.6 kb mRNA produce a 554-amino-acid long
peptide. The names given to these proteins are: M-CSF-β, long M-CSF, CSF-1554(human) or CSF-1552(mouse), secreted M-CSF and Fast
released M-CSF. At amino acid 277, a proteoglycan moiety is added to create pM-CSF (ovals). Depending on where specific proteolytic
enzymes subsequently cut, either a 223- or 456-amino acid M-CSF species is formed. The 456-amino-acid form is the pM-CSF species.
The various sections of M-CSF protein are schematically represented in color. From left to right, purple: leader sequence, green:
common 149-amino-acid M-CSF core, red: a spacer region, dark pink: a 25-amino-acid transmembrane region, and orange: a 35-
amino-acid intracellular region. The first arrow indicates where the leader sequence is cleaved quickly during processing by a signal
peptidase. The larger arrow represents where other protease activity subsequently cleaves M-CSF to yield various soluble species. The
majority of M-CSF produced in vivo (95–99%) is rapidly processed to a soluble form in the Golgi by M-CSF β-convertase. A minor species
of M-CSF (1–5%) remains in a membrane-bound form that appears on the extracellular region, membrane M-CSF. This form is slowly
released by M-CSF α-convertase.
The various forms of M-CSF. The M-CSF gene is transcribed into 4.0, 3.7, 3.1, 2.6 or 1.6 kb mRNA forms via alternative
1357 Macrophage colony-stimulating factor: Not just for macrophages anymore!
Author's personal copy
fms. There are five point mutations in the extracellular
domain. Two of these point mutations (301 and 374) turn v-
fms into a “constitutively activated” transforming agent
[81–84]. In the C-terminus of v-fms there is a 50-amino-acid
truncation that is replaced by 14 unrelated amino acids.
A novel use of c-fms has been reported by Lo et al.  T
cells (murine CTLL-2 or human peripheral blood T cells)
genetically engineered to express c-fms showed enhanced
responses to either sM-CSF or mM-CSF positive tumor cells.
CTLL-2 cells proliferated better with sM-CSF with sub-
optimal amounts of interleukin-2. The c-fms transduced
CTLL-2 cells made more interferon-γ (IFN-γ) when co-
stimulated with both phorbol myristate acetate and M-CSF.
Soluble M-CSF acted as a chemokine to c-fms transduced
human T cells. Dual-transduced human CD8+ cells (c-fms
along with a receptor for prostate membrane surface
antigen) killed human LNCaP prostate cancer cells. These T
cells responded better when the tumor cells also co-
expressed mM-CSF. The therapeutic rationale for this
approach is that ex vivo engineered human T lymphocytes,
when re-infused into patients, will be specifically attracted
to M-CSF-producing tumor cells. These redirected T cells
should then attack those tumor cells making M-CSF. Many
human tumors make M-CSF (see Introduction), so this
therapy could have broad clinical applications.
The Epstein–Barr virus encodes a receptor-like protein,
designated as Barf-1. This molecule  binds to M-CSF. Barf-
1 has no homology to c-fms, but it has 18% homology to the
co-stimulatory molecule, CD80. Barf-1 does activate Bcl-2
, so current speculation is that this molecule provides a
selective advantage by protecting EBV-infected cells from
dying of apoptosis. This molecule could down-regulate
immune function during infection in the host in a couple of
ways. First, by inhibiting macrophage function by interfering
with the normal binding of free M-CSF. This EBV M-CSF
receptor mimic is not a transmembrane protein like c-fms or
v-fms, but it is secreted from EBV-infected cells as a 23 kD
protein. By preventing macrophage activation via M-CSF,
macrophages may be less effective phagocytic cells, thereby
allowing free virions a better chance to infect their
appropriate target/host cell before being cleared by the
immune system. Second, Barf-1 could be disrupting T cell-
antigen-presenting cell signaling. Barf-1 is homologous to
CD80, which is a co-stimulatory molecule. So this molecule
could act as a molecular decoy and prevent proper co-
stimulation of Tcells by dendritic cells. Transduction of Barf-
protein species shown here. The 4 and 2.6 kb transcripts (green) are processed into the ER as 554-amino-acid polypeptides. The 3.7 kb
transcript (orange) encodes a 438-amino-acid protein. The 3.1 and 1.6 kb mRNA (purple) produce a 256-amino-acid species. As these
proteins are translated and translocated into the ER, the 23-amino-acid leader sequence is quickly cleaved by a signal peptidase. The
pro-M-CSF species form disulfide bonds, generating homodimers that are linked by a interstrand single interstrand disulfide bond. The
longer forms-, 554-amino acids, acquire proteoglycan moieties at position ser277. As M-CSF enters the Golgi, the sM-CSF is subject to
proteolytic attack by M-CSF β-convertase, which releases sM-CSF of 223-amino acids in length. A second protease, M-CSF α-
convertase, cleaves near the membrane, so that pM-CSF is released. As the mM-CSF transits the Golgi, it is resistant to M-CSF β-
convertase, so when the secretory vesicle fuses with the cell membrane, the mM-CSF now resides on the surface. Membrane M-CSF is
susceptible to slow proteolytic cleavage by M-CSF α-convertase which produces a 158-amino-acid soluble form of M-CSF. The exact
physical location where M-CSF α-convertase becomes active is not known, but apparently occurs on the surface of the cell.
The M-CSF processing pathways. The M-CSF gene transcripts undergoes variable mRNA splicing, ultimately generating the
1358T.G. Douglass et al.
Author's personal copy
1 into either mouse 3T3 cells or human B cells causes a
temporary oncogenic transformation of these cells [88–91].
As the Barf-1 gene is gradually lost, these infected cells lose
oncogenic potential. EBV is thought to be the cause of
nasopharyngeal carcinoma, Burkitt's lymphoma, Hodgkin's
lymphoma, some gastric cancers and some B cell lymphomas
after transplantation. Barf-1 may help avoid T cell-depen-
dent clearance of EBV-infected cells; thereby allowing the
carcinogenesis of these infected cells to emerge. This unique
EBV M-CSF receptor may have important ramifications for
both virology and oncology.
4. M-CSF signal transduction
Stein et al.  and Uemura et al.  first showed that the
less abundant, membrane-bound form, mM-CSF, was func-
tional and stimulated the growth of bone marrow cells via a
juxtacrine signaling pathway utilizing the M-CSF receptor. In
juxtacrine stimulation, a cell producing a membrane-bound
cytokine is in direct contact with a cell containing its
receptor. In the case of mM-CSF, binding to the receptor
involves residues in or near alpha helices A and C in the first
149 amino acids of the N-terminus of mature M-CSF . All
forms of M-CSF, secreted and membrane forms, all bind and
stimulate cells presenting c-fms. Mature M-CSF is a disulfide-
bonded dimer, so when one M-CSF monomer binds to one
receptor, the second M-CSF subunit binds another c-fms
molecule resulting in receptor dimerization. Following
dimerization of c-fms, autophosphorylation of at least eight
internal tyrosines (Fig. 3) is accomplished within minutes by
the internal kinase domain of c-fms . After the phosphor-
ylation of c-fms occurs, various docking/linker proteins are
recruited to the phosphorylated tyrosines. Numerous signal-
ing transduction pathways are then initiated, including those
involving: phosphoinositol-3 (PI-3) kinase [95–97], Stat 1,
Stat 3, tyk2 and JAK1 proteins , Ras , Raf and MAP
kinase , phospholipase A2 , phosphatase 1C ,
protein kinase C-δ , c-myc , and Mona . As a
to either proliferate or differentiate. For proliferation to be
induced within murine bone marrow-derived macrophages, 8
to 12 h of continuous M-CSF stimulation is required to get
these cells to proliferate . So constant M-CSF signal
transduction is required for c-fms+ macrophages to begin cell
division. Mona  is a monocyte/macrophage restricted
protein, so some cell-lineage proteins may account for some
different downstream biologies, when comparing the non-
monocytic to the monocytic c-fms-positive cells responding
One of the C-terminal tyrosines of c-fms (tyr969in human)
interacts with c-cbl . Upon interacting with c-cbl,
ubiquitin ligation of the c-fms complex occurs. This
ubiquitination then targets the receptor complex for
degradation via a lysosomal pathway that does not involve
the proteasome . After the association with clathrin-
dependent coated pits, the receptors are pinocytosed and
degraded [94,108]. Phosphorylated c-fms receptors are
internalized within 12–15 min, and are gradually replaced
with newly synthesized unphosphorylated receptors .
The c-cbl protein does not bind to v-fms, since v-fms lacks
the 50 terminal amino acid residues that include the
phosphotyrosines that dock to c-cbl . Hence, this viral
mutant M-CSF receptor provides a prolonged signaling
stimulus because the receptor is removed more slowly.
intracellular tyrosines in human c-fms that become phosphorylated after binding to the M-CSF are shown. The M-CSF dimer binds
to two c-fms molecules, facilitating receptor dimerization and autophosphorylation of the tyrosines by the intrinsic kinase
region. These phosphorylated tyrosines allow various docking proteins to bind and activate different signal transduction
pathways. Eventually, these docking proteins become ubiquitinated (green shapes) and are removed via clathrin-dependent
endocytosis. M-CSF binding, dimerization, autophosphorylation, ubiquitation and clearance of the M-CSF/c-fms complex occurs
within 12–15 min.
M-CSF signal transduction pathways. The M-CSF receptor is called c-fms. The numbered positions of the various
1359Macrophage colony-stimulating factor: Not just for macrophages anymore!
Author's personal copy
5. M-CSF involvement in bone physiology
Bone is a dynamic structure that constantly remodels itself in
response to various physical and hormonal stresses. Osteo-
blasts, derived from mesenchymal/stromal cells, are respon-
sible for bone formation, especially in response to calcitonin.
Hematopoietic-derived monocytes eventually differentiate
into multinucleated cells called osteoclasts. Osteoclasts are
responsible for bone resorption. In the normal physiological
state, there is a homeostasis between osteoblasts and
osteoclasts (Fig. 4A). Osteoblasts and osteoclasts cross-
regulate each other by releasing a variety of growth factors
and hormones that maintain the proper cellular balance for
bone mineralization (formation) and demineralization
The first indication that M-CSF plays an active role in bone
physiology was the discovery that the bone defect in the
osteopetrotic (op/op) mouse was caused by a nucleotide
insertion in the M-CSF gene [110,111]. This insertion causes a
frameshift mutation resulting in an inactive protein. Osteope-
lack of recruited osteoclasts results in a decrease in bone
resorption. Thus, the bones acquire a dense, short, and thick
pathology. The M-CSF deficiency also causes lack of teeth,
facial deformity, along with retarded skeletal growth. The
osteopetrotic phenotype can also be reproduced by knocking
deficient in systemic and local macrophages. These mice have
liver, sex organs and intestinal tract.
mM-CSF is biologically active in bone and does stimulate the
growth of osteoclasts [113,114]. Transgenic reconstitution of
the op/op mice with the sM-CSF or mM-CSF partially restores a
normal phenotype . The sM-CSF gene replacement best
restored the wild type phenotype and was superior to that
achieved by daily bolus injections of recombinant M-CSF. The
constant infusion of sM-CSF apparently allows more hemato-
poietic osteoclastic precursors to be recruited into the growing
the defective protein with membrane M-CSF corrected several
aspects of the pathology, but not all the defects of op/op mice
were completely reversed [115,116]. This was possibly due to
the inability of the cytokine to act as a chemoattractant for the
Bone physiology is very complex and requires other
molecules in addition to M-CSF [117,118]. Other membrane-
bound cytokines and ligands, such as Receptor Activator for
Nuclear Factor κ B-Ligand (RANK-L), intercellular adhesion
molecule (αvβ3) , and tumor necrosis factor (TNF) allow
multiple differentiation/activation signals to be delivered to
the osteoclasts via juxtracrine interactions with their appro-
priate receptors on the osteoblasts. When membrane-bound
cytokines bind to the receptors, they are not cleared within
12–15 min by the receptor-bearing cells, as is the case with
their soluble counterparts.
To illustrate this effect, we used rat bone marrow-derived
macrophages that reacted to rat T9 glioma cells making mM-
CSF (the T9-C2 clone). We observed an exaggerated and
prolonged PI-3 kinase signal transduction response. Fig. 5A
and B shows a time-lapse progression of a macrophage
contacting several mM-CSF-expressing T9-C2 glioma cells.
The macrophage (black arrow) was the attached cell, while
the T9-C2 cells (white arrows) were added as free-floating
cells at Time 0. After 15 min (Fig. 5A), the T9-C2 cells start to
settle down and some inter-cell contacts are seen. By 1 h,
the T9-C2 cells have attached to the macrophages and more
membrane interactions are seen (Fig. 5B). Fig. 5C shows the
time-dependent kinetics of the functional PI-3 kinase
responses of macrophages reacting to the mM-CSF-positive
glioma cells. At 10 min there is a brief burst of PI-3 kinase
activity within the macrophages. We attributed this early
response to the shed mM-CSF from the T9-C2 cells that were
added to the macrophages. This finding is consistent with
previous studies following the ligand/phosphorylation of
soluble M-CSF with c-fms [94,108]. After 1 h the macro-
phages have responded to the mM-CSF, after making
prolonged contacts with the T9-C2 cells (Fig. 5B). This
stronger PI-3 kinase activity was still seen after 2 h of
incubation, when the experiment was terminated.
This prolonged/exaggerated response may be analogous
to a process called frustrated phagocytosis, where phago-
cytes try to engulf larger particles than they can successfully
envelop . As a result of this increased cell-surface
binding, more intracellular docking proteins are recruited
into the membrane area than would normally be participat-
ing in the phagocytosis/pinocytosis of smaller particles/
molecules. These additional recruited proteins create a
prolonged signal. The exaggerated membrane cytokine
stimulation by mM-CSF via c-fms either alone or in combina-
tion with other membrane cytokines/ligands therefore could
induce the monocytes to assume the more differentiated
morphology that is the hallmark of osteoclasts.
Studies have shown that transcription of M-CSF can be
induced by a number of agents (lipopolysaccharide, tumor
necrosis factor, interleukins 1 and 2, interferon-γ, 1,25-
dihydroxy vitamin D, dexamethasone, parathyroid hormone
and parathyroid hormone related peptide [PTH and PTH-rP ,
respectively]) [121–124]. Rubin, et al. [122,123] concluded
that dihydroxy-1,25 vitamin D and dexamethasone stimulated
osteoblasts to make mM-CSF and that mM-CSF expression was
due to increased transcription and differential mRNA splicing.
Endocrinologists have known for decades that PTH/PTH-rP
causes bone resorption by stimulating the osteoclasts to
release various mediators such as hydrogen ions and/or
proteases such as cathepsin K, which causes the demineraliza-
tion of bone. Both PTH and PTH-rP bind to the same PTH
receptor ; by autoradiography, these receptors are only
found on osteoblasts, not osteoclasts . So the mechanism
of action of PTH and PTH-rP must be proceeding through the
osteoblasts via an indirect pathway.
M-CSF production by osteoblasts may explain why certain
cancers, such as breast and prostate, have a propensity to
metastasize into the bone. Some breast and prostate cancers
possess c-fms [41,49–51]]. We speculate that M-CSF acts as a
chemoattractant for those circulating c-fms+ metastatic
clones. Osteoclast activation may be influenced by these
recruited cancer cells by two different scenarios (Fig. 4C).
Osteoblasts making mM-CSF may directly bind to the c-fms+
cancer cells as they arrive into the bone. The mM-CSF on the
osteoblasts then stimulates the cancer cells by producing
enhanced juxtracrine signaling (see above, Fig. 5). If these
tumor cells make M-CSF, then they can directly activate the
osteoclasts. Alternatively, the cancer cells can release
metabolites that cause demineralization.
1360T.G. Douglass et al.
Author's personal copy
For other tumors that do not make M-CSF or c-fms, these
cancer cells (whether they metastasize into the bone or via
endocrine stimulation) can make PTH-rP that causes
increased levels of circulating Ca+2ions . Here the
circulating PTH-rP can activate osteoblasts.
Thus, both soluble and membrane isoforms of M-CSF and
6. Intestinal tract
An intriguing observation concerning the effect of M-CSF on
the intestines was made by Ramsay et al. in 2004 . They
described that the murine intestinal epithelial cells that
form villi possess c-fms on the basolateral side of the cells. As
these epithelial cells leave the crypt area, where the
intestinal stem cells reside, they begin to express c-fms.
The M-CSF made by the macrophages, endothelial or other
stromal cells allows the villi cells to differentiate and mature
as these epithelial cells migrate. Isolated single cell prepara-
with recombinant M-CSF, produced epitheloid-type colonies,
analogous to hematopoietic colonies grown under similar
conditions. Ryan et al.  showed that the macrophage
content within the intestinal tract within op/op mice is
severely depleted of F4/80+ macrophages. So it is not
surprising that the intestines of op/op mice display some
abnormal physiology. The op/op mouse intestinal tract lacks
normal-sized villi and the surface area of the intestines is
greatly diminished, when compared to their wild type
littermates. Because the op/op macrophages and endothelial
cells lack M-CSF production, then the intestinal cells fail to
cells fail to mature and differentiate into normal villi.
Zapata-Velandia and her coworkers  confirmed some
of the above studies. Immunohistochemistry showed that c-
fms was found in the epithelial cells of normal human ileum
and colon. This group reported that there was a single
nucleotide polymorphism in c-fms (CSF1R-A2033T) near the
RUNX1 binding site in the eleventh intron. This variant was
found in27% oftheir Crohn's disease cohort fromLouisiana. In
contrast, a group from New Zealand could not reproduce the
their study group of 182 individuals . This discrepancy
could be due to the fact that the latter group examined a
different genetic population of individuals. Crohn's disease
has other genetic susceptibility genes such as NOD2, MDR1,
mechanism of this disease in certain populations.
Other groups have linked inflammatory bowel disease
(IBD) with M-CSF. Makiyuma et al.  demonstrated
elevated serum levels of M-CSF in patients with active IBD
(both ulcerative colitis and Crohn's disease). Klebl et al.
, showed that M-CSF mRNA and protein were found
within normal human intestines and bowels. They also
reported that in ulcerative colitis and Crohn's disease the
percentage of M-CSF-positive cells significantly increased.
Marshall and colleagues,  using a dextran sodium sulfate
(DSS)-induced colitis model in mice, confirmed that
increased M-CSF levels were found in mice with active
lesions. When neutralizing anti-M-CSF antibody was added to
the drinking water, the severity of this DSS-induced
pathology was reduced on multiple parameters. These latter
studies provide some evidence that M-CSF is directly involved
with the pathology of this disease. Whether this pathology is
due to the actions of the activated macrophages or problems
associated with the epithelial cells differentiation due to
excess M-CSF remains to be determined.
hybridoma form a tumor in vivo?
Our work with mM-CSF began with a hybridoma cell line,
called NBXFO. This cell line was created by Beverly Barton
 who took splenocytes from newborn mice (NB) and
fused them (X) with a FO fusion partner (FO). This cloned
NBXFO hybridoma induced immune suppression via a soluble
mediator having a non-reduced molecular weight of 90 kD
and a pI of 4.5. These characteristics roughly match those of
one of the large soluble forms of M-CSF . This hybridoma
exhibited a fibroblastic/stromal cell phenotype . The
NBXFO cells are adherent cells and make extracellular matrix
proteins such as collagen, laminin and fibronectin. The M-
CSF gene was most likely contributed from the neonatal
fibroblast found either in the spleen capsule or in the spleen
stromal components. Northern blot analysis of neonatal
splenocytes and NBXFO cells showed M-CSF transcripts (4 and
1.6 kb M-CSF mRNAs) were made, while FO cells did not make
any M-CSF mRNA. The supernatants derived from the NBXFO
hybridoma, but not those from FO cells, supported the
growth of macrophages from the bone marrow. Flow
cytometry confirmed that the NBXFO cells expressed mM-
CSF on their cell surface.
When syngeneic mice were injected with 16 million NBXFO
cells, theyfailedtoinitiate growtheitherinintraperitoneal or
subcutaneous sites. If the mice were sub-lethally irradiated or
treated with immunosuppressive agents such as low-dose
cyclophophosphamide, the hybridoma still failed to grow. The
lack of tumor formation lead to an interesting paradox: how
could this hybridoma, which was making an immunosuppres-
sive factor, fail to form into a tumor when the host's immune
directly make immunosuppressive agents or indirectly stimu-
late suppressor cells that shut down anti-tumor immune
responses[136,137]. Low-dose cyclophosphamide kills mature
and activated lymphocytes, while simultaneously stimulating
the growth of immature monocyte/myeloid suppressor cells
[138–140]. These myeloid suppressor cells are immature
monocytes that express c-fms and most likely eliminated the
NBXFO cells after injectionintothe mice. When bone marrow-
derived macrophages were incubated with the radiolabeled
NBXFO cells, the macrophages killed these hybridoma cells
within 18 h . We speculated that the mM-CSF on the
NBXFO cells was responsible for eliciting the macrophage-
To prove that mM-CSF was responsible for allowing this
type of macrophage-based cytotoxicity to occur, we geneti-
cally engineered non-M-CSF producing tumor cells with LXSN-
based retroviruses that would stably transduce either human
sM-CSF or human mM-CSF. Only human mM-CSF was cloned at
that time. Human M-CSF does stimulate rodent cells
possessing M-CSF receptors. The first tumor cells we
1361Macrophage colony-stimulating factor: Not just for macrophages anymore!
Author's personal copy
transduced were rat T9 gliomas , human U251 glioma
cells , and murine Hepa1-6 hepatomas . When
these mM-CSF transduced cells were cloned, the T9-C2,
U251-2F11 and Hepa1-6 D1 clones were the best mM-CSF
expressing cell clones, in that order. These cloned trans-
duced cells were killed by monocytes and macrophages,
confirming our previous conclusions with the NBXFO cells.
Unmodified tumor cells or viral vector control cells not
expressing mM-CSF were not killed by either rodent or human
monocytes/macrophages in vitro. In vitro, sM-CSF trans-
duced T9 cells (T9-H1 clone) were not killed by the
macrophages. In vivo, these sM-CSF-expressing tumor cells
were not eliminated, and they quickly formed subcutaneous
tumors . Thus, mM-CSF apparently stimulated
1362 T.G. Douglass et al.
Author's personal copy
macrophages into a tumoricidal state, in contrast to the
result obtained with the soluble form of M-CSF.
8. Mechanisms of macrophage-mediated killing
of mM-CSF transduced tumor cells
Early in vitro work used bone marrow-derived macrophages
to perform cytotoxicity studies. These bone marrow cells
were quite easy to culture and provided a ready source of
macrophages on a daily basis. These macrophages specifi-
cally killed mM-CSF transduced T9 glioma [141,145] or
Hepa1-6 hepatoma  cells through a phagocytosis-
dependent pathway. This cytotoxicity was directly depen-
dent upon the surface concentration of mM-CSF. The more
mM-CSF the clones expressed, the better the macrophage
killing was. Importantly, cells expressing the rapidly secreted
form of M-CSF failed to elicit macrophage cytotoxicity.
Macrophage-mediated phagocytosis of the mM-CSF cells
could be prevented by either saturating the c-fms receptors
on the effector macrophages by 100× molar excess of
recombinant M-CSF, or by chemical inhibitors of phagocytosis
(D-deoxyglucose, iodoacetate, gadolinium chloride or cyto-
chalasin B) [141,143]. Tumor cells transduced with sM-CSF
failed to promote macrophage conjugation and failed to
allow cytotoxicity to occur.
Oneof thedangers of using a pharmacological approach to
inhibit a biological response in a two-cell reaction, is that
one can't exclude the possibility that the drug, however
specific as it is supposed to be, will not affect another
different biochemical pathway, especially in the target cell.
As a result, we also used a genetic approach to confirm the
pharmacology data. Bone marrow-derived macrophages
derived from Hck−/−, Fgr−/−, Lyn −/− triple knock-out
mice, that are genetically incapable of phagocytosis, also
showed that these knock-out macrophages were incapable of
killing the rat T9 or mouse Hepa1-6 hepatomas 
expressing mM-CSF via a phagocytic mechanism.
removal. Osteoclasts (pink) make M-CSF and mM-CSF, along with other membrane ligands (RANK ligand and tumor necrosis factor),
which may stimulate the osteoclasts (pink) via a prolonged signaling mechanism for the membrane receptors for RANK, TNFR or αvβ3)
to induce bone resorption. Osteoclasts also make growth factors and hormones that maintain osteoblasts, which can down-regulate
the actions of pro-bone resorption. As a result, bone maintains itself in a normal physiology. Panel B shows what occurs when M-CSF is
knocked out in the osteopetrosis (op/op) mice. Because the mice fail to make M-CSF, there is no M-CSF available to recruit the
osteoclast precursors. Hence, no mature osteoclasts are present (shown by dotted lines). As a result, the osteoblasts continue to lay
down bone matrix. The bone becomes dense and short, resulting in the osteopetrosis phenotype. Panel C shows how osteolytic disease
can occur when metastatic tumor cells (blue) arrive in the bone. Since the osteoblasts release M-CSF, c-fms+ tumor cells, such as those
in some breast and prostate cancers, can be recruited into the bone. The c-fms tumor can either release M-CSF/mM-CSF or PTH/PTH-
rP that induce osteoclastic activity. The mM-CSF (if below the threshold value to induce macrophage-mediated cytotoxicity)
stimulates the osteoclast into bone resorption. Non-M-CSF producing tumor cells can also release PTH/PTH-rP which directly
stimulates the osteoblast into making M-CSF/mM-CSF and initiating other osteoclastic activity that causes osteolysis.
The role of M-CSF in bone metabolism. Panel A shows the normal homeostatic balance between bone formation and bone
1363 Macrophage colony-stimulating factor: Not just for macrophages anymore!
Author's personal copy
As described earlier (See M-CSF signal transduction), M-
CSF stimulates monocytes and macrophages by binding to
the c-fms receptor (Fig. 2). A number of pathways are
activated upon c-fms stimulation including the PI-3 kinase
pathway. The use of PI-3 kinase chemical inhibitors (wortman-
nin [Fig. 6A] and Ly294002 [Fig. 6B]) inhibited in a dose-
dependent manner the cytotoxic actions of bone marrow-
derived macrophages towards either T9-C2 or Hepa1-6 D1
cloned cells in 24 h assays (C.C. Williams, Masters' Degree
Thesis, California State University, Long Beach). With the help
that the macrophages derived from PI-3 kinase knock-out mice
had reduced cytotoxicity (Fig. 6C). Macrophages derived from
at killing the mM-CSF expressing tumor cells. The macrophages
from the heterozygotic (PI-3K +/−) mice displayed an unex-
pected inhibition ofthis macrophage-mediated cytotoxicity, an
cytotoxicity of Hepa1-6 D1 cells. Mouse bone marrow-derived
macrophages were incubated with mM-CSF-expressing Hepa1-6
D1 cells (pre-labeled with
absence of PI-3 kinase inhibitors (wortmannin [top panel] or
Ly294002 [middle panel]) for 24 h. The inhibitor concentration is
shown within the legends inside each panel. The amount of lysis
was then calculated and shown as % Specific Kill with the
standard deviation (SD) of quadruplicate cultures. Panel C shows
the killing of the mM-CSF-expressing Hepa1-6-D1 cells by bone
marrow macrophages derived from various PI-3 kinase knock-out
(KO) littermates. The macrophages derived from 3 wild type
mice(PI-3K +/+), 8 heterozygotic (PI-3K +/−) and 3 homozygous
(PI-3 K −/−) KO mice were tested against the Hepa1-6 D1 (pre-
labeled with3H-Thymidine) in a 24-h cytotoxicity assay. Each
macrophage cell line was tested in triplicate cultures with the
Hepa1-6 D1 hepatoma cells at 3 different effector to tumor cells
ratios (10:1, 5:1 and 2.5:1). Data from all experiments were
pooled at each macrophage:tumor ratio. The data is expressed
as % Kill±Standard Error of the Means (SEM). The asterisks
indicate significant differences (Pb0.05).
PI-3 kinase plays a role in macrophage-mediated
3H-thymidine) in the presence or
rat macrophages responding to mM-CSF-expressing T9-C2 cells.
The top panels show the time-lapse photography of adherent rat
macrophage contacting the mM-CSF transduced T9-C2 cells.
Panel A is at 30 min and Panel B is at 60 min. As time progresses,
the mM-CSF-positive glioma cell (white arrows) and the
macrophage (black arrow) have greater cell-surface contact.
We speculate that this prolonged contact allows exaggerated
signal transduction responses to occur/accumulate. Panel C
shows the PI-3 kinase enzymatic activity of the bone marrow
macrophages after responding to the T9-C2 for similar times. At
the indicated times, the cultures were collected and the cells
lysed. The protein concentrations were measured and 100 µg of
total protein was immunoprecipitated with anti-PI-3 Kinase
antibody. The immunoprecipitate was washed and then incu-
bated in the presence of 20 µCi γ32P-ATP along with the
substrate, phosphatidylinositol (PI) for 20 min. The reaction
was stopped by 1 N HCl and the lipids were extracted by a 1:1
CHCl3/MeOH solution. The organic phase was then spotted onto
a potassium oxalate treated TLC plate and allowed to develop in
a CHCl3/acetone/methanol/acetic acid/water solvent. After
the lipids separated, the amount of32P transferred onto the PI
was quantitated by a phosphorimager and is presented as a
Time-lapse photography and PI-3 kinase activity of
1364T.G. Douglass et al.
Author's personal copy
inhibition that was almost as complete as that in the
homozygous (PI-3K −/−) knock-out macrophages.
We concluded from this in vitro work using bone marrow-
derived macrophages that the mechanism of killing was
through direct phagocytosis of mM-CSF expressing cells.
Our initial in vivo studies began with the rat T9-C2 glioma
cells. These transduced cancer cells, failed to grow, whether
implanted surgically in the brains of rats  or subcuta-
neously [144,146]. In over 400 rats used to date, T9-C2 cells
never formed subcutaneous tumors, even if 10×106cells were
injected. Similar results were seen with: 1) human U-251
(U251-2F11 clone) injected into either immunodeficient nude
mice or SCID/bg mice ; 2) with the aggressive weakly
immunogenic rat MADB106 breast cancer cells, ; 3) rat
NUTU-19 ovarian cancer cells; 4) rat D74 glioma cells (data
from these lasttwo syngeneicF-344 based-tumor models were
not published, Martin Graf, Scott Rose, Sanjay Rao, and Martin
Jadus); and 5) non-immunogenic murine Hepa1-6 hepatocel-
lular carcinoma cells . Most interestingly, when Bcl2 was
co-transducedinto mM-CSFexpressingcells, the subcutaneous
injected tumor cells still failed to form tumors . In
Tumor cells expressing mM-CSF could grow in animals that
had been depleted of their myeloid cells using polyclonal
antibodies . When monocytes/macrophages in rats were
eliminated, the T9-C2 cells could grow in some nude rats.
Because the T9 glioma cells were spontaneously making
interleukin-8 , (also known as cytokine-induced neutro-
phil chemoattractant, CINC), neutrophils also were recruited
into the injection site. Some groups reported that immature
neutrophils do express c-fms . When both macrophages
In general, our in vivo data correlated very well with that
predicted from in vitro cytotoxicity assays.
Upon electron microscopic analysis, we recognized that
the in vivo macrophage-killing process of mM-CSF-positive
cells was not proceeding through direct phagocytosis of the
tumor cells, as predicted by earlier in vitro studies done with
bone-marrow-derived macrophages. The in vivo morphology
of these dying mM-CSF tumor cells revealed a swollen and
vacuolated phenotype, best described as paraptosis
[142,146]. Paraptosis is thought to be a programmed form
of death that leads to necrosis [152,153]. Paraptosis begins
with the swelling of the mitochondria and the endoplasmic
reticulum, but then the cells themselves continue to swell
until they physically rupture. The triggers for this lysis are
produced by the recruited myeloid cells: HOCl, H2O2or OH−.
Using both rat macrophages and human monocytes, we
showed that ROS is released within 15–30 min of being
exposed to the mM-CSF-expressing tumor cells, even though
rat monocytes are slightly slower in generating ROS [149–
151]. The U251 and T9 glioma cells all proved to be
susceptible to HOCl, H2O2 and OH−. T9-C2 cells resisted
nitric oxide or peroxynitrite mediated killing, while U251
cells were very susceptible to killing by these two reactive
nitrogen intermediates; NO  and peroxynitrite (unpub-
lished data). Apparently, glioma cells have different suscept-
ibility to killing by various reactive intermediates.
9. The molecular mechanisms by which
monocytes/macrophages kill mM-CSF-
expressing glioma cells
When human monocytes were exposed to U251-2F11 glioma
cells for 4 h, distinct morphological changes occurred within
the mM-CSF transduced U251 cells. The glioma cells failed to
attach themselves to the plastic surface while simulta-
neously contacting the monocytes. The affected mM-CSF
transduced glioma cells appeared bloated. This was not seen
when either the unmodified or the viral vector control cells
interacted with the monocytes. During this time frame, ROS
production was detected by fluorescent microscopy, using
H2DCFDA. By microarray analysis, we found several genes
were modestly up-regulated: hemoxgenase-2 (1.6±0.3 fold)
and NADPH 450 reductase (1.7±0.8 fold). A report by
Williams et al.  linked these two molecules functionally.
Since these two up-regulated genes are enzymes, the
targeted paraptotic cells did not need de novo transcription
or protein synthesis to accomplish these induced morpholo-
gical changes. Hemoxygenase-2 and the NADHPH 450
reductase catalyze a reaction to produce carbon monoxide,
which acts as a secondary messenger to open potassium ion
channels called BK (big potassium, to reflect large electrical
conductances). This ion channel also known as Maxi-K, hSlo,
of mM-CSF expressing tumor cells via reactive oxygen species
(ROS). Following prolonged contact with mM-CSF via the c-fms on
the macrophages, ROS are made. These ROS activate hemox-
ygenase-2 (HO-2) and NADPH450 reductase (450 reductase) to
catalyze the formation of carbon monoxide (CO). CO then
stimulates the opening of the BK channels found in the
mitochondria, endoplasmic reticulum, and the cell membrane.
As a result, potassium ions are released from their normal stores.
Eventually, potassium leaves the cell, and is replaced by sodium
ions. When sodium ions enter, so does water, eventually causing
prolonged BK channel activation, they eventually lose the ability
to produce ATP . Once ATP levels are depleted, the cell lyses
osmotically due to excess intracellular Na+ions and water.
Proposed mechanism of macrophage-mediated killing
1365 Macrophage colony-stimulating factor: Not just for macrophages anymore!
Author's personal copy
calcium dependent (BKCa), large conductance- or voltage
activated-channels now allows the release of K+ions from
various K+ion stores. BK channels were found on the glioma
cell membrane and patch-clamping studies showed that
these BK channels were functional . Laser confocal
microscopy indicated that these potassium channels co-
localized to the mitochondria and the ER. This data provided
a possible explanation for how the tumor cells are attacked
by the ROS made by the myeloid cells.
Our proposed mechanism by which monocytes/macro-
phages kill mM-CSF-transduced cells via BK channel activa-
tion is shown in Fig. 7. The proposed process involves six
steps. Before macrophages come in contact with the mM-CSF
tumor cells, normal homeostasis is present in the targeted
cells. Intracellular ATP, K+and Na+are at baseline physiolo-
gical levels; i.e., intracellular concentrations are high in ATP
as H2O2and HOCl, are produced by the myeloid cells (Step 1).
As the c-fms-positive myeloid cells bind the mM-CSF, they
generate a respiratory burst, which includes the production
of ROS. The presence of ROS within the mM-CSF expressing
target cells, allows hemoxygenase and P450 reductase to
produce CO (Step 2). Acting as a secondary signal, CO
mediates the opening of BK channels (Step 3) on the cell
membrane, in the ER, and in the mitochondria, thereby
allowing pooled stores of K+to be released. Using CO-
saturated tissue culture media, T9 glioma cells rapidly swell
As the cell releases K+(Step 4), Na+cations enter, preserving
When Na+enters the cell (Step 5) water follows, producing
and mitochondrial swelling (Step 5) due to the presence of
extra Na+and water. Cellular homeostatic mechanisms are
activated to expel the excess intracellular Na+through the
ATP dependent Na+/H+anti-porter  or other Na+/K+
ATPase exchangers (Step 6). The cell expends more ATP to
expel the excess Na+ions. Physical disruption of the
mitochondria reduces its ability to generate sufficient
studies (submitted), we show that as the T9 cells die in
response to various BK channel activators or H2O2, intracel-
lular levels of ATP are depleted, just prior to cell death. In
support of this model, Na+/H+anti-porter inhibition has been
shown to produce paraptosis in cerebellar neurons .
Panel A shows untreated U251 glioma cells stained foranti-HMGB1 antibody (red) and DAPI (blue) stained forthe nucleus. The HBMG1 is
retained at nuclear/perinuclear locations. Very little HMGB1is found either in the cytoplasm or along the membrane of the cells.
Panels B, C and D show U251 cells treated for 1 h at 37 °C with 1 mM phloretin (Panel B), 0.01 mM pimaric acid (Panel C) or 1 mM H2O2
(Panel D). Inall cases theHMGB1 has translocated from the nucleus/perinuclear regions through the cytoplasm towards theedge of the
membranes of the effected cells.
HMGB1 is translocated from the nucleus to the membrane as a result of BK channel activation or by hydrogen peroxide.
1366T.G. Douglass et al.
Author's personal copy Download full-text
10. Paraptotic tumor cells produce “danger
Actively dying cells possess distinct morphologies, reflecting
autophagy, necrosis/paraptosis and apoptosis . Autop-
hagy consists of cells self-digesting internal organelles due to
a lack of vital nutrients. Apoptosis is identified by nuclear
condensation, DNA cleavage, cell shrinkage, membrane
blebbing and high-mobility group box-1 (HMGB1) transloca-
tion into the nucleus from its normal perinuclear location
activated, facilitating many of these changes. This process
culminates in apoptotic body formation, which restricts the
release of various intracellular contents thereby limiting
healthy tissue damage and inflammation. Paraptotic cells, as
described above, are characterized by a swelling and
vacuolization of the endoplasmic reticulum (ER) and the
mitochondria [142,146]. Osmotic lysis releases substances
that have been labeled as “danger signals” [159–161]. These
include HMGB1 , heat shock proteins (HSP) such as
HSP60, HSP70, HSP90 and grp94/gp96 [159–161]. HMGB1 has
been reported to bindto interleukin-1β and enhances its pro-
inflammatory nature . These HSPs can act as cytokines
and bind to various receptors on immature dendritic cells via
TLR2, 4, CD14 and CD91 [163–167]. As a result these
immature dendritic are stimulated into becoming mature
antigen-presenting cells and stimulating the Tcells.
Our paraptotic model postulates the loss of intracellular
ATP level is a key element in killing cells via this pathway. Low
“danger signals” promotes massive inflammation, ultimately
stimulating cell-mediated immunity [159–161]. U251 cells
that were treated/killed by the BK channel activators or by
H2O2 showed that HMGB1 is translocated away from the
nucleus towards the membrane and eventually released (Fig.
8). Thus, several released proteins probably account for the
“naturally” immunogenic properties of paraptotic cells.
Chiang et al.  reported that HOCl caused human SK-OV-
3 ovarian cancer cells to die of necrosis; these necrotic cells
demonstrated increased immunogenicity for dendritic cells
stimulating Tcells. Besides HOCl, another ROS member, H2O2,
has also been reported to stimulate human dendritic cells in
vitro . Thus, as a result of all these released molecules:
ROS, HSP , HMGB1, tumor antigens, an “immunostimulatory
storm” is initiated after paraptotic cells osmotically lyse.
Our “grand unifying theory” (Fig. 9) brings together
several divergent lines of research. Some groups demon-
strated that the loss of intracellular ATP leads to necrosis
[172–174]. Low intracellular levels induce HSP 70 transcrip-
tion [168,169]. HSPs and HMGB1 release are danger signals
that lead to enhanced immune responses, apparently
explaining the improved immunogenicity of paraptotic
cells. Other groups have concluded that necrotic tumor
cells produce stronger immune responses than apoptotic
cells do [175–181]. The model in Fig. 9 may provide the
molecular mechanisms by which paraptotic cells produce
such effective immune responses.
11. Immunized T cells are produced once mM-
CSF-expressing tumor cells are rejected
Others showed that HSPs and HMGB1 enhance immune
responses to antigens and tumors [159,161–163]. We
contact mM-CSF-expressing cells, reactive oxygen species are made. A proposed mechanism for the resultant cell swelling is shown in
Fig. 7. As the intracellular levels of ATP decline, several genes are activated including those for heat shock proteins (HSP), and HMGB1
translocates to the membranes. As the cell lyses from unopposed osmotic pressure, the HSPs and tumor antigens are released. HSPs
can bind to surface markers such as CD91, CD14 and toll-like receptors-2 and 4 and activate immature dendritic cells (DC) to become
mature DC. The tumor antigens can be collected by the immature DC. After the DC matures, it acts as an antigen-presenting cell that
can then stimulate both CD4 and CD8 T cells. These T cells then mediate anti-tumor immunity. As a result, tumor immunity can be
generated against tumor cells that express the same tumor antigens.
Mechanism by which mM-CSF induces paraptosis and facilitates induction of anti-tumor immunity. When macrophages
1367 Macrophage colony-stimulating factor: Not just for macrophages anymore!