n e w s a n d v i e w s

nature medicine volume 15 | number 7 | july 2009 729

for human cDNAs that permitted mouse cells to bind EV71. They isolated multiple clones encoding PSGL-1. Because PSGL-1 is expressed primarily on leukocytes, it is unlikely to serve as the receptor on nonleukocyte cells, such as neurons. Indeed, these authors show that sev-eral susceptible cell lines derived from epithe-lial, fibroblast and neuroblastoma cells do not express PSGL-1 and that PSGL-1–specific anti-bodies do not prevent infection of these cells as they do with T cells. Nishimura et al4. also show that, although multiple independently isolated strains of EV71 use PSGL-1 to infect Jurkat T cells, some EV71 strains infect these cells without binding PSGL-1.

The evidence in each paper suggests that there may be two or more receptor molecules for EV71. It remains to be seen whether PSGL-1 is the possible second receptor detected by Yamayoshi et al.3 and whether the PSGL-1–independent infection of T cells is mediated by SCARB2 or by additional receptors to be identified.

Other questions also arise. First, how do the receptors permit infection at the cellular level? Both PSGL-1, a mucin, and SCARB2, a type 3 glycoprotein with amino- and carboxy-termi-nal cytosplasmic domains, differ in structure

from other enterovirus receptors, and it will be interesting to see how they interact with the virus and what happens once attachment has occurred. In addition to facilitating attachment to the cell, virus receptors may transmit signals required for viral entry, mediate internaliza-tion of the virus, or trigger release of the viral genome into the cell11. When expressed in mouse cells, PSGL-1 and SCARB2 each seem to be sufficient for virus attachment and entry, but it remains to be determined whether each receptor mediates entry in a different way.

Of even greater interest is how the receptors contribute to the pathogenesis of EV71 and whether they contribute to neurotropism, the affinity of a virus for the brain. Neither PSGL-1 nor SCARB2 is localized exclusively to the brain, and neurotropism may depend on many fac-tors other than receptor expression. If viruses differ in their tropism for the two receptors, is there a correlation between receptor preference and neurovirulence? CVA16, which causes the benign hand, foot and mouth disease, can bind both PSGL-1 (ref. 4) and SCARB2 (ref. 3), but Nishimura et al.4 point out that a prototype strain of CVA16 does not depend on PSGL-1 for infection of Jurkat cells. SCARB2 is widely expressed and may be directly involved in

infection of the brain by EV71. PSGL-1 is not expressed on neurons, but infected leukocytes that express PSGL-1 may traffic to the central nervous system and allow the virus to spread.

The interaction of EV71 with PSGL-1 on lymphocytes may induce production of the inflammatory cytokines involved in encepha-litis or pulmonary edema; cytokines could also disrupt the blood-brain barrier, permitting direct viral invasion.

Understanding how these receptors work will take additional effort, but their identifica-tion is a major step forward.

1. Fan, Y. et al. J. Clin. Microbiol. published online, doi:10.1128/JCM.00563-09 (13 May 2009).

2. Huang, C.C. et al. N. Engl. J. Med. 341, 936–942 (1999).

3. Yamayoshi, S. et al. Nat. Med. 15, 798–801 (2009).4. Nishimura, Y. et al. Nat. Med. 15, 794–797 (2009).5. Melnick, J.L. Rev. Infect. Dis. 6 Suppl 2, S387–S390

(1984).6. Eckhardt, E.R. et al. J. Biol. Chem. 281, 4348–4353

(2006).7. Vishnyakova, T.G. et al. Proc. Natl. Acad. Sci. USA 103,

16888–16893 (2006).8. Philips, J.A., Rubin, E.J. & Perrimon, N. Science 309,

1251–1253 (2005).9. McEver, R.P. & Cummings, R.D. J. Clin. Invest. 100,

485–491 (1997).10. Lin, T.Y., Hsia, S.H., Huang, Y.C., Wu, C.T. & Chang,

L.Y. Clin. Infect. Dis. 36, 269–274 (2003).11. Marsh, M. & Helenius, A. Cell 124, 729–740 (2006).

Coupling bone degradation to formationJameel Iqbal, Li Sun & Mone Zaidi

To maintain skeletal integrity and prevent fractures, degradation and rebuilding of bone must occur in synchrony. Transforming growth factor-β1 is now found to coordinate this restructuring process: the molecule is released during bone degradation and stimulates bone rebuilding (pages 757–765).

The Mount Sinai Bone Program, Mount Sinai

School of Medicine, New York, New York, USA.

e-mail: mone.zaidi@mssm.edu

Although often depicted in the lay media and cartoons as a hunk of mineralized rock, bone in reality is a highly dynamic and specialized organ. As a connective tissue, bone can increase in mass in response to stimulation such as exer-cise and, conversely, can rapidly destroy itself if not needed, as astronauts in orbit have come to learn1.

This remodeling, which occurs in response to microcracks that accumulate during everyday life, is a constant process in adult vertebrates. Old bone is removed by osteoclasts and restored by osteoblasts, which lay down matrix pro-teins, particularly collagen, which is later min-eralized2. To maintain skeletal integrity, bone

degradation and formation must be coupled temporally, spatially and quantitatively so that packets of old bone are replaced with packets of new bone in perfect synchrony. If these pro-cesses are decoupled, crippling skeletal diseases result; if the bone remodeling cycle is desyn-chronized, pathological changes in bone mass cause osteoporosis or osteosclerotic disorders, and if bone remodeling is not coordinated spa-tially, bones become misshapen, as in Paget’s bone disease and the rare Camurati-Engelmann disease (CED)2.

In this issue of Nature Medicine, Tang et al.3 show in mice that transforming growth factor-β1 (TGF-β1) is the key chemoattractant that tightly couples bone degradation and for-mation in space and time.

The coupling of bone resorption and forma-tion has been challenging to study because it is an in situ phenomenon that cannot be ade-

quately replicated in a culture dish. The adult remodeling cycle begins at areas of stress, where bone lining cells are dissociated from the under-lying bony trabeculae. Chemoattractants such as sphingosine-1-phosphate help steer osteo-clast precursors to these sites, where receptor activator of nuclear factor-κB ligand (RANKL) produced by stromal cells and osteoblasts induces the differentiation of mature bone-resorbing osteoclasts. After bone degradation, osteoblast precursors are recruited to these sites, and then they undergo differentiation and sub-sequently lay down unmineralized matrix. The mineralization of this matrix completes the remodeling cycle.

Several experiments have hinted that TGF-β1 is released from sites of bone resorption to stimulate bone formation. TGF-β1 is a ubiq-uitous cytokine, produced by many different cells, including osteoblasts, and most cell types

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

n e w s a n d v i e w s

730 volume 15 | number 7 | july 2009 nature medicine

CED, in which a mutation within the latent pro-tein of TGF-β1 causes bilateral, symmetric and progressive dysplasia of long bones, ultimately leading to sclerosis and fragility10. The defect in the latent protein causes its premature dis-sociation, leading to increased production of activated TGF-β1 at sites nearly everywhere within the marrow cavity of CED mice (Fig. 1) instead of only at sites of osteoclastic resorp-tion. Similar to TGF-β1–overexpressing mice7, CED mutant mice had increased osteoclast and osteoblast activity, indicating that the overpro-duction of TGF-β1 caused high turnover of bone. However, in contrast to mice overex-pressing TGF-β1 (ref. 7), CED mice showed disorganized osteoclast and osteoblast activ-ity, recapitulating the human CED phenotype. This observation indicated that prematurely activated TGF-β1 interferes with the proper migration of BMSCs to sites of resorption and confirmed unequivocally that the spatial cou-pling of resorption to formation depends on the gradient of active TGF-β1.

Finally, Tang et al.3 showed that treatment with an inhibitor of the TGF-β receptor-1 normalizes the grossly aberrant remodel-ing in CED mice3. It is possible that a similar approach could be used to treat other diseases where there is uncoordinated osteoblastic bone formation and osteoclastic bone resorption, such as Paget’s disease11. A recent study showed that inhibitors of the type 1 TGF-β1 receptor increased bone mineral density and mineral concentration in a dose-dependent manner, improved the architecture of trabeculae and enhanced matrix elasticity in wild-type mice12. In those experiments, the increased bone min-eral density may be mediated through reduced osteoclast numbers and bone degradation, as mildly blocking TGF-β1 signaling seemed to allow bone remodeling to occur at the appro-priate locations.

Regardless of the mechanism, those findings, together with the elegant studies of Tang et al.3, present the possibility of increasing bone mass in a nonhaphazard manner. Despite this excit-ing news, such a therapeutic approach may have an Achilles heel: the near ubiquitous expres-sion of TGF-β receptors. Without the ability to regulate where TGF-β inhibitors localize after absorption, it remains an open question whether the TGF-β axis could be targeted for the treatment of osteoporosis, bone metastases and other metabolic bone diseases.

1. Iqbal, J. & Zaidi, M. Biochem. Biophys. Res. Commun. 328, 751–755 (2005).

2. Zaidi, M. Nat. Med. 13, 791–801 (2007).3. Tang, Y. et al. Nat. Med. 15, 757–765 (2009).4. Bonewald, L.F. & Mundy, G.R. Connect. Tissue Res. 23,

201–208 (1989).5. Oreffo, R.O., Mundy, G.R., Seyedin, S.M. & Bonewald,

L.F. Biochem. Biophys. Res. Commun. 158, 817–823 (1989).

they found that osteoclasts release activated TGF-β1 during bone resorption3. The cytokine was the major stimulus for migration of bone marrow stromal cells (BMSCs) toward sites of resorption. Knowing that TGF-β1 is a chemoat-tractant for osteoblasts9, Tang et al.3 predicted that the major action of TGF-β1 is not regulat-ing osteoblast proliferation but rather guiding BMSCs correctly to sites of resorption.

To test this hypothesis, Tang et al.3 generated TGF-β1–deficient mice; these mice develop autoimmune disease, so the researchers pre-vented immune activation by also making the mice deficient in recombination-activating gene-2 (Rag2)3. Tang et al.3 injected these mice with wild-type GFP-labeled BMSCs. In con-trol Rag2-deficient mice, the labeled cells lined bony bone trabeculae at one week and became embedded within the bone matrix by four weeks, as expected. However, when TGF-β1 was absent, the donor BMSCs failed to local-ize to the trabeculae and remained in bone marrow. This experiment established beyond doubt that TGF-β1 is necessary for the local-ization of BMSCs to sites of osteoclastic bone resorption. Thus, the signal for coupling sites of bone resorption to subsequent formation is osteoclast-released active TGF-β1.

To further explore these major observations, Tang et al.3 developed a mouse model of human

express receptors for it4. Importantly, TGF-β1 is released from cells in an inactive form, inhibited by an associated latent peptide that prevents receptor binding4. Because of the generalized expression of TGF-β1 receptors, the crucial step for TGF-β1 regulation is the production of the active cytokine.

Bone houses an abundant supply of inac-tive TGF-β1, but the cytokine is only activated after osteoclastic resorption, when enzymes and acidic pH cause the dissociation of the latency-associated peptide5. Although the role of TGF-β1 in bone remodeling has remained unclear, the prevailing hypothesis has been that TGF-β1 induces the proliferation of nearby osteoblasts, which would be predicted to promote bone formation. Altering TGF-β1 signaling, however, does not cause defects in osteoblast activity or bone formation6. Overexpression of TGF-β1 stimulates bone turnover, resulting in bone loss, but does not decouple bone resorption and for-mation7. Conversely, blocking TGF-β1 signaling causes excessive bone accretion despite normal bone formation rates, indicating less bone deg-radation8. These confusing phenotypes suggest TGF-β1 has a broader role in the matrix than the regulation of osteoblast formation.

Tang et al.3 bring clarity to these mouse mod-els by a complementary set of in vitro and in vivo experiments. Using a two-well culture system,

a Normal remodeling

Bone

Bone

BMSCsBMSCs

Osteoblasts

Osteoclast Osteoclast

TGF-β1Bound

LP

LP

FreeLP

b Remodeling in CED

TGF-β1

Osteoblasts

Figure 1 Premature TGF-β1 activation in Camurati-Engelmann disease (CED) leads to haphazard bone formation and poor bone quality. (a) Under normal circumstances, osteoblasts secrete TGF-β1 with its latent protein as they lay down matrix. In the resorption pit, the TGF-β1 is freed from latent protein (LP) and diffuses from the resorption site, acting as a chemoattractant for BMSCs. Osteoblasts follow the activated TGF-β1 gradient and subsequently fill in the resorbed cavities. (b) In CED, mutations in the latent protein cause premature dissociation, resulting in the premature release of activated TGF-β1. The resulting effect is a distortion of the resorption-induced TGF-β1 gradients. Because of inadequate BMSC recruitment to sites of resorption, poor-quality bone is produced with unfilled resorbed areas and haphazard sclerotic areas.

Kat

ie V

icar

i

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

n e w s a n d v i e w s

nature medicine volume 15 | number 7 | july 2009 731

(2005).11. Kurihara, N. et al. J. Bone Miner. Res. 21 Suppl 2,

55–57 (2006).12. Mohammad, K.S. et al. PLoS One 4, e5275 (2009).

8. Filvaroff, E. et al. Development 126, 4267–4279 (1999).

9. Lucas, P.A. Bone 10, 459–463 (1989).10. Gupta, S. & Cheikh, I.E. Endocr. Pract. 11, 399–407

6. Kakonen, S.M. et al. J. Biol. Chem. 277, 24571–24578 (2002).

7. Erlebacher, A., Filvaroff, E.H., Ye, J.Q. & Derynck, R. Mol. Biol. Cell 9, 1903–1918 (1998).

Tumor immunotherapy: making an immortal armyBrent H Koehn & Stephen P Schoenberger

Manipulation of cell renewal pathways creates T memory stem cells that can generate a sustained and targeted immune response. These findings have broad implications for vaccine development and immunotherapy.

Brent H. Koehn and Stephen P. Schoenberger

are in the Laboratory of Cellular Immunology,

La Jolla Institute for Allergy and Immunology,

La Jolla, California, USA.

e-mail: sps@liai.org

The Greek historian Herodotus tells of an elite corps of Persian warriors who fought the vastly outnumbered Spartans at the battle of Thermopylae in 480 BCE1. Herodotus called these fabled Persians the “immortals,” because as each man fell to disease or injury, he was immediately replaced by another well-trained soldier, maintaining the numbers of the force at a constant strength1.

In this issue of Nature Medicine, echoes of this epic story appear in a new immunother-apy tactic. Gattinoni et al.2 have found a way to generate optimal CD8+ T cell responses2—through activation of the Wnt–β-catenin pathway. Activation of this pathway can convert CD8+ T cells into a self-renewing ‘army’ of stem cells that generates specialized effector T cells capable of eradicating tumors many thousands of times larger on a per-cell basis2. If reproducible in the clinical setting, this strategy could markedly improve the effi-cacy of both vaccines and adoptive immuno-therapies through the continuous generation of antigen-specific CD8+ T cells from a self-renewing reservoir.

CD8+ T cells are the ‘killer’ lymphocyte subset, charged with executing virtually any cell in the body that may become infected with an intracellular pathogen. They do this by inducing apoptosis through a vari-ety of mechanisms and have been shown to be remarkably effective against tumors that express mutated self-proteins or tissue-specific differentiation antigens3. As such, they represent a powerful weapon within the immunotherapist’s arsenal.

Antigen-specific CD8+ T cells can be iso-lated from the blood of patients with cancer and expanded in vitro to large numbers that can be transferred back to the patient and

attack the tumor. Alternatively, these cells can be readily induced within an individual through vaccination.

The rapid disappearance of CD8+ T cells after vaccination or adoptive transfer, how-ever, limits their potential efficacy. After primary activation by antigen, naive T cell precursors undergo clonal expansion to generate a range of functionally distinct subsets of T cells that differ in their longev-ity, location and cytotoxic potential. Most of the CD8+ T cells that arise from vaccination and during in vitro expansion are short-lived effector cells that are terminally differenti-ated. Such cells can effectively kill the first wave of targets they encounter, but, in the absence of memory cells able to persist and generate new effectors, the immune response will be temporary at best.

The capacity for self-renewal and contin-ued differentiation is found within two other subsets of T cells: the effector memory cells (TEM cells) found mostly in peripheral tis-sues and the central memory cells (TCM cells) that reside in lymphoid organs such as spleen

and lymph nodes. How these memory sub-sets are induced and maintained, however, is unclear4.

The Wnt pathway involves a number of evo-lutionarily conserved proteins that regulate many cellular events, ranging from embryo-genesis to differentiation5. Binding of Wnt pro-teins to cell surface receptors leads to a change in the amount of β-catenin (an intracellular signaling molecule) that reaches the nucleus, where it interacts with members of the TCF/LEF family of transcription factors to promote new gene expression6. In hematopoietic stem cells, Wnt controls self-renewal by limiting proliferation and differentiation so that divi-sion can regenerate both multipotent daughter cells and additional pluripotent stem cells7.

In their search for a sustained immune response against particular tumor antigens, Gattinoni et al.2 investigated whether mature T cells could become memory cells after activation of the Wnt–β-catenin pathway. The researchers treated a line of T cells that specifically recognize a melanoma antigen with several small molecules that activate the

Kat

ie V

icar

i

Bone marrow

HSC

MSC

LPC

MPC

T, B and NK cells

Monocytes and macrophages, RBCs, neutrophils, eosinophils, basophils, megakaryocytes

Proliferation anddifferentiation

GSK-3βinhibition

Proliferation anddifferentiation

Wnt–β-catenin pathway

In vitro

Naive CD8+

T cell

Ag + APC

In vivo

Adoptivetransfer Effector

cells

Memorycells?

Proliferation anddifferentiation

Wnt–β-catenin pathwayAg + APC

TMSC

Figure 1 Same pathway, different outcomes? During hematopoiesis in the bone marrow, the Wnt–β-catenin pathway limits the proliferation and differentiation of hematopoietic stem cells (HSCs) so that division can regenerate HSCs in addition to the multipotent stem cells (MSCs) that give rise to lymphoid progenitor cells (LPCs) and myeloid progenitor stem cells (MPCs). Gattinoni et al.2 induced this pathway in mature CD8+ T cells through pharmacological inhibition of GSK-3β during priming, resulting in the generation of memory cells that possess stem cell–like qualities of self-renewal and multipotency. NK, natural killer; RBCs, red blood cells; Ag, antigen; APC, antigen-presenting cell.

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.