NEWS & VIEWS

NATURE|Vol 457|1 January 2009

orchestrate inflammation remotely to overt inflammatory conditions An inflammatory (such as breast cancer), the activation by producing antibodies that become microenvironment of oncogenes can orchestrate the prolocalized in the extracellular matrix11. Tissue invasion duction of inflammatory molecules What’s more, a macrophage-derived & metastasis extracellular-matrix protein called SPARC and the recruitment of inflammatory facilitates tumour-cell motility and metacells. In the tumour microenvironstasis12. So it seems that extracellularment, inflammatory cells and molecules influence almost matrix components are much more than Evasion of Insensitivity to every aspect of cancer progress, a scaffold, or a substrate to be consumed apoptosis growth inhibitors including the tumour cells’ abilduring tumour-cell invasion, but instead ity to metastasize3. Thus, whereas represent a central component of cancerthere were previously six recognized related inflammation. hallmarks of cancer — unlimited The present study2 offers unexpected replicative potential, self-suffivistas on the molecular pathways that link Sustained ined Self-sufficiency in Self-suf ciency in growth signals, insensiinflammation to acquisition of the capacity angiogenesis enesis growth growt signals tivity to growth inhibitors, evasion to metastasize during tumour progression. Limitless replicative of programmed cell death, ability It will be essential to assess the significance potential to develop blood vessels, and tissue of versican and other extracellular-matrix invasion and metastasis5 — cancerproteins in models that reflect the diversity of related inflammation now emerges human cancer, for from such work innovative Figure 1 | The hallmarks of cancer. In 2000, Hanahan therapeutic strategies may follow. as number seven (Fig. 1). ■ 5 and Weinberg proposed a model to define the six A group of cytokine proteins, Alberto Mantovani is at the Istituto Clinico properties that a tumour acquires. These are unlimited including IL-1, IL-6, TNF and replicative potential, ability to develop blood vessels Humanitas IRCCS, and the University of Milan, RANKL, activate inflammation and (angiogenesis), evasion of programmed cell death Rozzano, Milan 20089, Italy. are known to augment tumour cells’ (apoptosis), self-sufficiency in growth signals, insensitivity e-mail: [email protected] ability to metastasize by affecting to inhibitors of growth, and tissue invasion and metastasis. 1. Fidler, I. J. Nature Rev. Cancer 3, 453–458 (2003). several steps in the cells’ dissemina- Kim and colleagues’ findings2, together with those of 2. Kim, S. et al. Nature 457, 102–106 (2009). 3,4 tion and implantation at secondary other studies , indicate that this model should be revised 3. Mantovani, A., Allavena, P., Sica, A. & Balkwill, F. Nature 3,6,7 to include cancer-related inflammation as an additional 454, 436–444 (2008). sites . Inflammatory cytokines hallmark. (Adapted from ref. 5.) 4. Coussens, L. M. & Werb, Z. Nature 420, 860–867 lie downstream of the ‘master’ gene (2002). transcription factor for promot5. Hanahan, D. & Weinberg, R. A. Cell 100, 57–70 (2000). ing inflammation — NF-κB — which is itself particular vascular endothelial growth fac- 6. Giavazzi, R. et al. Cancer Res. 50, 4771–4775 (1990). activated by them3. A major source of inflam- tor, which is mobilized by enzymes originat- 7. Luo, J. L. et al. Nature 446, 690–694 (2007). 8. Wyckoff, J. B. et al. Cancer Res. 67, 2649–2656 (2007). matory cytokines in the tumour microen- ing from inflammatory white blood cells and 9. Mantovani, A., Schioppa, T., Porta, C., Allavena, P. & Sica, A. Cancer Metastasis Rev. 25, 315–322 (2006). vironment are specialized white blood cells which promotes blood-vessel formation durcalled macrophages. Tumour-associated ing tumour progression4,5 . Moreover, during 10. Yang, L. et al. Cancer Cell 13, 23–35 (2008). 11. de Visser, K. E., Korets, L. V. & Coussens, L. M. Cancer Cell 7, macrophages assist the malignant behav- the development of cancers caused by human 411–423 (2005). iour of tumour cells, not just by producing papillomavirus, immune cells called B cells 12. Sangaletti, S. et al. Cancer Res. 68, 9050–9059 (2008). cytokines, but also by secreting growth factors and matrix-degrading enzymes8–10. Kim et al.2 explored the molecular pathways linking tumour cells, macrophages and ASTROPHYSICS metastasis. By purifying the components of the medium in which the tumour cells (the Lewis lung carcinoma cell line) were grown, they isolated a factor that induced cytokine proRalph E. Pudritz duction by macrophages. They identified this tumour-derived macrophage activator as ver- Deciphering how stars form within turbulent, dense clouds of sican, a protein of the extracellular matrix that molecular gas has been a challenge. An innovative technique that is frequently upregulated in human tumours. The authors found that versican is recognized uses a tree diagram provides insight into the process. by TLR2 and TLR6, two receptor proteins that belong to a family of cellular sensors of An understanding of how stars form has been away from the observer, the wavelength of the microbially derived molecules and tissue dam- hampered by the complexity of the clouds millimetre emission from a molecule such as age. They then went on to silence versican in of cold molecular gas within which their carbon monoxide is shifted towards shorter tumour cells by an RNA interference technique, formation occurs. To elucidate this process, (‘blueshifted’) or longer (‘redshifted’) waveand to use mice in which TNF and TLR were the effects of gravity must be traced across a lengths. By measuring these Doppler shifts at absent. On the basis of the evidence obtained, wide range of scales, particularly at the large each point on the sky, one can determine the the authors propose that, in the Lewis lung car- distances over which it operates in these gas relative velocities of parcels of gas in the cloud cinoma model, tumour-derived versican acts clouds. On page 63 of this issue, Goodman along each line of sight. It turns out that the on macrophages through TLR2/TLR6, leading et al.1 show how a hierarchical tree diagram — gas motions in such clouds are mainly highly to the production of inflammatory cytokines, a ‘dendrogram’ — can be used to disentangle supersonic. Indeed, computer simulations which enhance metastasis. the gravitational connections that tie the gas show2 that the network of dense filamentary Kim and colleagues’ observations highlight together on many different scales. structures seen in clouds is probably a direct the importance of the extracellular matrix in The movements of gas in molecular clouds consequence of such supersonic gas flows cancer-related inflammation. The matrix acts are measured by spectral shifts. Depending on (Fig. 1). as a depot of cytokines and growth factors, in whether a parcel of gas is moving towards or As with many astronomical observations,

Star formation branches out

© 2009 Macmillan Publishers Limited. All rights reserved

37

NEWS & VIEWS

BLACKWELL PUBLISHING

NATURE|Vol 457|1 January 2009

Figure 1 | Stellar nursery. The image shows a computer simulation2 of the formation of stars within a turbulent, self-gravitating cloud of gas. The initial mass of this star-cluster-forming cloud, which is modelled as a sphere of uniform density, is 500 solar masses. The sphere’s radius is 83,300 astronomical units (1 au is the mean distance between Earth and the Sun) and its temperature is 10 kelvin. Supersonic turbulence compresses the gas into many filaments and smaller, dense regions. The simulation is viewed after one free-fall time — the time taken for a gas parcel to collapse freely to the cloud centre — has elapsed, which is 1.9×105 years for these simulation parameters. The white dots correspond to small, dense gas ‘cores’ that collapse to form individual stars. These would correspond physically to regions such as those denoted by the billiard balls in Figure 1 of the paper by Goodman et al.1 (page 63).

however, we do not know the distance of any object (gas in this case) from Earth without using further painstaking methods. In observing molecular clouds, one is limited to measuring the two-dimensional position of the total gas emission on the sky. This measurement involves two coordinates that are akin to latitude and longitude on Earth’s surface, as well as the relative velocity of gas at that position. Thus, a map of such a cloud is a sequence of position–position–velocity (p–p–v) measurements of the gas across the whole cloud. But without the ability to observe the full, threedimensional gas cloud, how can its true structure be deduced, let alone the strength of the various forces that control where and when stars will form? The approach usually taken to tackle this problem consists of segmenting the clumpy cloud into a collection of structures (clumps) using a computer program called CLUMPFIND. The end result is analogous to a topographical relief map of a mountain range. Such a map typically shows peaks that stand 38

out from ridges or are isolated, and provide a series of contours that demarcate different elevations. If one now decided to break the range up into discrete ‘mountains’, one would identify the peaks and, using the various contour levels, try to decide whether smaller outcrops ‘belonged’ to a given mountain or were independent structures. In a p–p–v map of a molecular cloud, it is the contours and peaks of gas-column density — the sum of the emission from all gas parcels moving at a given velocity along a given line of sight — that play the role of topological relief in this mountainrange analogy. The problem with this approach arises as soon as the results are used to try to provide insight into how stars form. For example, the column densities allow one to measure the masses of the clumps. One can then count the number of clumps with a given mass. The resulting distribution of clump masses is used to work out how star formation might occur3,4. With programs similar to CLUMPFIND, the data suggest that the © 2009 Macmillan Publishers Limited. All rights reserved

clump-mass distribution closely resembles the distribution of star masses. A plausible but debated inference is that the origin of stellar masses may derive directly from the turbulent process that produced the clumps. The fly in this ointment, however, as Goodman et al.1 show (see Fig. 1 of their Supplementary Information), is that if one adopts different threshold levels for contouring the maps, the column-density distribution changes. This is similar to the situation in a topographical relief map of a mountain range: the list of discrete mountains and their properties depends on the choice of threshold level picked to define mountains and smaller outcrops. This is unsettling — such an approach does not provide a completely objective way of measuring the actual distribution of column densities. Enter the dendrogram technique advocated by the authors1. Rather than dividing the p–p–v data into a priori distinct structures in a subjective way, they use a method that is sensitive to the structures’ intrinsic hierarchy (structure within structure). The data are broken up into ‘leaves’, ‘branches’ and ‘trunks’. Leaves are identified as sufficiently strong maxima in the column-density maps, and connections between them are made by branches (their environs). The collection of physically related branches defines a trunk. Every point on the dendrogram corresponds to a closed ‘isosurface’ on the map that encloses one or more column-density maxima (see Fig. 2 on page 64). The authors then show how physical properties can be ascribed to regions within these isosurfaces. A critical property is the ‘virial’ parameter — the ratio of the energy of the gas motions (which depends on the gas velocity in the region) to the gravitational energy (which depends on the mass and size of the region). If this ratio is sufficiently small, the gas in that region will be self-gravitating and prone to form stars. The dendrogram thus traces the relative strength of the gravitational force across the gas cloud. Application of the dendrogram technique to observations of the gas cloud L1448, and to computer simulations in which only turbulence — and not self-gravity — is taken into account, shows inconsistencies. In contrast to the simulations, in which most of the gas is found to be self-gravitating on all spatial scales, the observations show that, although a large fraction of the gas is self-gravitating at large scales, at smaller scales that fraction is much lower. Interestingly, strong local column-density maxima — which correspond to the dense gas ‘cores’ in which stars are observed to form — are sparser in Goodman and colleagues’ dendrogram than those found with the CLUMPFIND algorithm, and turn out to be in larger regions of self-gravitating gas. And here we arrive at what may be the most tantalizing point of all. A debate that has enlivened star-formation theory for nearly a decade is how stars acquire their mass5. Do they accrete their gas from relatively isolated cores, or do

NEWS & VIEWS

NATURE|Vol 457|1 January 2009

they accrete material as they move about in a broader gravitational potential, gathering mass through competition with other dense regions in the same gravitationally bound region6 (the ‘competitive accretion’ picture)? Although neither hypothesis is amenable to definitive observational tests, the dendrogram method developed by Goodman et al. has the potential to answer this question and to identify the real conditions in which stars form. ■ Ralph E. Pudritz is in the Department of Physics and Astronomy, McMaster University, 1280 Main

Street West, Hamilton, Ontario L8S 4M1, Canada. e-mail: [email protected] 1. Goodman, A. A. et al. Nature 457, 63–66 (2009). 2. Bate, M. R. Mon. Not. R. Astron. Soc. (in the press); preprint at http://arxiv.org/abs/0811.0163 (2008). 3. Motte, F., André, P. & Neri, R. Astron. Astrophys. 336, 150–172 (1998). 4. Johnstone, D. et al. Astrophys. J. 545, 327–339 (2000). 5. McKee, C. F. & Ostriker, E. C. Annu. Rev. Astron. Astrophys. 45, 565–687 (2007). 6. Bonnell, I. A. & Bate, M. R. Mon. Not. R. Astron. Soc. 370, 488–494 (2006).

GAME THEORY

How to treat those of ill repute Bettina Rockenbach and Manfred Milinski A much-needed theoretical analysis deals with whether the principle known as ‘costly punishment’ helps to maintain cooperation in human society. It will prompt a fresh wave of experiments and theory. Human societies are built on cooperation, especially on reciprocation1 — I help you and you help me, or I help you and someone else helps me. In the first case, help is directly reciprocated by help. In the second, called indirect reciprocity, I gain a good reputation and so I can expect help when in need. But how shall I treat someone with a bad reputation? Shall I just refuse help or shall I punish this person at a cost to myself? Costly punishment can enhance cooperation 2,3 in experiments with human subjects, but potentially with no net benefit4: the costs of punishment usually, although not always5, neutralize gains from enhanced cooperation. On page 79 of this issue, Ohtsuki et al.6 describe a theoretical test of whether either refusing help to or punishing someone with a bad reputation might lead to a cooperative society. They conclude that, except under certain rare conditions, punishment does not produce that outcome. When you meet someone needing help, you can help (cooperate), refuse to help (defect) or not only refuse to help but, in addition, decrease the needy person’s wealth (punish). Both cooperation and punishment are costly for you, but respectively create a larger benefit or larger loss for the person needing help. Defection is cost neutral. How you behave depends on the reputation — good or bad — of the needy person, and depends upon your ‘action rule’. An example is ‘cooperate with someone with a good reputation and defect with someone with a bad reputation’ (CD). The reputation you yourself gain by applying your action rule depends on the social norm of your society. Under the norm ‘stern-judging’, for example, you gain a good reputation when cooperating with good or when defecting with bad, and a bad reputation

in all other cases. Thus CD always leads to a good reputation under stern-judging. Another action rule, CP, prescribes ‘cooperate with good and punish bad’. Under sternjudging, with CP you will achieve a good reputation when you interact with someone with a good reputation and a bad reputation when you interact with someone with a bad reputation. But under a different social norm, ‘shunning’ (cooperation with good or punishment of bad leads to a good reputation), CP will always provide you with a good reputation (Fig. 1). In their simplest model, Ohtsuki et al.6 assume that everybody has the same opinion of the reputation of another person or has the same level of fallibility in assigning an

incorrect reputation. For such a society, they test for each of the 64 different social norms (Fig. 1) whether an action rule exists that both generates a cooperative society and is evolutionarily stable — meaning one that resists replacement (invasion) by any of the other eight possible action rules (Fig. 1). Ohtsuki et al. find that the two action rules that induce cooperation and resist invasion are those described above — CD under sternjudging and CP under shunning. Nonetheless, the average pay-off is lower if the action rule uses costly punishment, while the stability conditions are less restrictive. However, which parameters determine which rule is most efficient in the sense of leading to the highest average pay-off at equilibrium? It turns out that a crucial one is the accuracy of assigning the correct reputation to everybody. If this accuracy is too low then only a DD action rule is efficient, under which nobody cooperates. If the accuracy is high enough, then CD can be efficient. For intermediate values of the accuracy, there is a small window in which CP can be efficient, as reflected in the title of the paper6: “Indirect reciprocity provides only a narrow margin of efficiency for costly punishment.” In a further step in their modelling, Ohtsuki et al.6 dropped the assumption that all good or bad reputations are publicly known, and allowed individual knowledge of reputations. They found that the stability of both CD and CP is lost when there is the smallest error in distinguishing between good and bad. When individuals start to communicate with each other and adjust their assessments of everybody’s reputation, the CP action rule can be stably maintained. Then, when reputations become even more publicly agreed upon through more efficient gossip, both CD and CP are stably maintained under their

Alice’s reputation

Alice’s reputation

Alice’s reputation

Good

Good

Good

Bad

Bad

Bad

Cooperate Cooperate Cooperate Bob’s Cooperate action rule

Defect

Social norm

Punish

Bob’s new reputation

Defect

Cooperate

Defect

Defect

Defect

Punish

Punish

Cooperate

Punish

Defect

Punish

Punish

Stern-judging Good

Bad

Shunning Good

Good

Figure 1 | Action rules, social norms and the story of Alice and Bob. Bob meets Alice and learns whether Alice has a good or a bad reputation. On the basis of that, Bob may either help (cooperate), refuse help (defect) or refuse help and, in addition, decrease Alice’s wealth (punish). How Bob reacts is specified in his action rule. Bob’s own reputation depends on how society’s social norm evaluates Bob’s reaction to Alice’s reputation. The social norm ‘stern-judging’ evaluates cooperation with good as good and punishment of bad as bad; the social norm ‘shunning’ evaluates cooperation with good as good and punishment of bad as good. Each social norm specifies which reputation to assign for each of the 6 possible scenarios (3 actions of Bob for 2 reputations of Alice). This leads to 26 = 64 social norms. © 2009 Macmillan Publishers Limited. All rights reserved

39