observed patterns and scaling laws in the prop-erties of galaxies. One of the most powerful of these has become known as the Kennicutt–Schmidt law (KS relation), which relates the concentration (surface density) of star forma-tion in galaxies to the concentration of cold gas2,3. The relation, consisting of a nonlinear power law at high gas densities with a turn over or threshold at low densities4, applies across virtually all galaxies today, from relatively qui-escent systems such as our Milky Way to the most active ‘starburst’ galaxies. For the galaxy modeller, it provides a one-step means of pre-dicting the amount of star formation from the distribution of cold gas, and this law or close variants of it are incorporated into nearly all models of galaxy formation and evolution.

The question posed by Gnedin and Kravtsov1 is whether this star-formation law, which applies to present-day galaxies, is invariant over cosmic time. For simplicity, most models incorporate a time-invariant law, but observa-tions of distant (early-epoch) galaxies present a mixed picture. For example, quasar-absorp-tion-line galaxies — distant galaxies identi-fied through their gas absorption of light from background quasars (extremely bright galactic nuclei) — show far lower rates of star forma-tion (by more than an order of magnitude) than would be expected if they followed the present-day KS law5,6. This is consistent with a strong evolution in at least the threshold in the KS law when the Universe was 15–50% of its current age. On the other hand, luminous star-burst galaxies at similar cosmic epochs show no such suppression of star formation relative to the local KS law7. Indeed, a wide range of observations show that early-epoch galaxies overall were forming stars much more rapidly than they are today.

Gnedin and Kravtsov offer a tentative solu-tion to this paradox. They do so by using numerical simulations as a virtual laboratory to explore how changes in the properties of gal-axies and the Universe at early cosmic epochs might change the star-formation law. The simulations use a theoretically motivated star-formation prescription in which the forma-tion rate per unit volume scales as the ratio of the molecular-gas density to the gravitational free-fall time (the time taken for a gas cloud to collapse freely to its centre), multiplied by star-formation efficiency per free-fall time. Their prescription comes out of recent theoretical work8,9, and is broadly consistent with the observed properties of local, present-day star-forming galaxies and the phenomenological basis of the KS law2,8. However, when Gnedin and Kravtsov examined the properties of the high-redshift galaxies in their simulation (in this case z = 3, or 16% of the present age of the Universe), they found that star formation was suppressed at much higher gas surface den-sities than is found today, in a regime where present-day galaxies readily form stars.

Dissecting the simulations revealed the physical explanation for such suppression.

Young galaxies tend to be strongly depleted in heavy elements and interstellar dust (which build up gradually over multiple generations of stars and stellar explosions called super novae), and this depletion acts together with the more-intense ultraviolet radiation in the early Uni-verse to suppress the formation of molecular hydrogen, a necessary ingredient for forming stars. This would suppress star formation over wide swathes of a young galaxy, without inhib-iting it in unusually dense regions, where suffi-cient shielding remains to form molecular gas. In this way, the models can account for the lack of star formation in the quasar-absorption-line galaxies while preserving the efficient star formation in the bright starburst galaxies.

Apart from offering an elegant explanation for a seemingly contradictory set of obser-vations, Gnedin and Kravtsov’s study1 has broader implications. It demonstrates that local scaling relations such as the KS law may them-selves evolve over cosmic time, complicating the task of the galaxy modeller. The robust-ness of this result rests heavily on the validity

of the star-formation prescription used in the calculations, but studies by many other groups9 support it. Viewed more broadly, this renais-sance of theoretical interest in the physics of star formation in the Universe is sure to lead to new insight into the formation of galaxies and stars alike. ■

Robert C. Kennicutt Jr is at the Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK. e-mail: robk@ast.cam.ac.uk

1. Gnedin, N. Y. & Kravtsov, A. V. Astrophys. J. 714, 287–295 (2010).

2. Kennicutt, R. C. Jr Astrophys. J. 498, 541–552 (1998).3. Schmidt, M. Astrophys. J. 129, 243–258 (1959).4. Martin, C. L. & Kennicutt, R. C. Jr Astrophys. J. 555, 301–321

(2001).5. Wolfe, A. M. & Chen, H.-W. Astrophys. J. 652, 981–993

(2006).6. Wild, V., Hewett, P. C. & Pettini, M. Mon. Not. R. Astron. Soc.

374, 292–304 (2007).7. Bouché, N. et al. Astrophys. J. 671, 303–309 (2007).8. Krumholz, M. R., McKee, C. R. & Tumlinson, J. Astrophys. J.

699, 850–856 (2009).9. McKee, C. F. & Ostriker, E. C. Annu. Rev. Astron. Astrophys.

45, 565–687 (2007).

ORGANIC CHEMISTRY

Symmetrizing the unsymmetricalScott A. Snyder

You might think that the partial symmetry of the molecule complanadine A makes it easy to prepare, but the reverse is true. Two syntheses of this compound offer insight into how to make partly symmetrical molecules.

When the molecular structures of naturally occurring organic compounds are symmetri-cal, it is relatively easy for synthetic chemists to come up with a blueprint to make them in the laboratory. But if that symmetry is broken just a little — say, by the presence of a single bond in the ‘wrong’ place, or by the insertion of an atom into one of the two otherwise equivalent parts of the molecule — then all bets are off. Two recent syntheses1,2 of the non-symmetri-cal molecule (+)-complanadine A, reported in the Journal of the American Chemical Society, provide insight into how to access such non-symmetrical structures. Ironically, the key to success involves symmetry.

Nature weaves together carbon, nitrogen, oxygen, hydrogen and a few other select atoms into millions of different organic molecules of varying size and complexity. The diversity of architectures produced from just these few components is truly stunning3. Equally striking is that these structures include what is arguably a disproportionate number of symmetrical molecules — compounds that are easily bro-ken into two identical halves that are usually natural products in their own right.

The fact that such dimeric structures are formed — and that there are so many of them — is probably born of the need to make the most

efficient use of energy. The biosynthesis of any small molecule requires a lot of energy-carrying ATP molecules, not to mention highly evolved enzymatic machinery. It therefore seems logical for the organism producing it to form a second compound from that molecule if the second product can be accessed for ‘free’ or at a mod-est additional energetic cost. After all, the new molecule might confer an added evolution-ary advantage on the organism in the form of different and/or improved biological activity.

Usually, nature will make such a dimer through the easiest and most energetically economical reaction available, producing molecules that are fully symmetrical and that chemists can typically make using available synthetic tools. But every once in a while, nature abandons this low-cost approach and becomes more extravagant, joining together monomers in a manner that is not only less obvious to chemists, but that also creates dis-symmetry. Although the monomer in these compounds is still clear, finding the means to unite two of them often proves highly chal-lenging — if not impossible — to achieve in a laboratory flask, because the union requires each monomer to behave in a chemically very different way. Complanadine A (Fig. 1) is one example of this relatively rare phenomenon4.

560

NATURE|Vol 465|3 June 2010NEWS & VIEWS

© 20 Macmillan Publishers Limited. All rights reserved10

Synthetic chemists have been interested in making complanadine A, as well as its many alkaloid cousins, for two main reasons. The first is to generate sufficient supplies of the compound to explore its biological activity. Complanadine A has potential as a medicine for regenerating neurons, and other structur-ally related compounds already constitute the basis for drug-discovery programmes in the fight against Alzheimer’s disease. The second reason is that the connectivity of atoms and bonds in complanadine A makes it an ideal tar-get with which to explore the power of tools and strategies for organic synthesis. The synthetic challenge is, in fact, readily underestimated.

On paper it is easy to see that the structure of complanadine A contains two identical pieces, the structures of which correspond to that of the natural product lycodine (Fig. 1). But in practice, uniting two lycodine molecules to form the desired dimeric product is almost impossible to achieve selectively, because dif-ferent carbon atoms in each lycodine unit must be connected.

There are few attractive options available for solving such a synthetic problem. One choice would be to fashion two different precursors, each based on the monomeric unit, but with extra chemical groups attached to the carbon atoms that need to be connected. These chemi-cal groups would then allow a bond to form controllably between the key carbons. Although this approach is almost guaranteed to work, it requires two separate synthetic endeavours to prepare each fragment, thereby doubling the total effort and financial outlay involved. Another possibility would be to simply target the monomer, in the hope that some process can be developed to dimerize it into the final structure without forming large amounts of unwanted side products. This idea is conceptually attrac-tive, but history reveals that such dimerizations are rarely achieved in the absence of the enzymes used for such purposes in nature.

Fischer and Sarpong1 and Siegel and col-leagues2 have developed a third, potentially powerful approach that is effectively a hybrid of the two methods discussed above. In their strat-egy, a single intermediate is prepared and then differentiated into two different analogues of the monomer at a late stage of the synthesis. These analogues are then connected using a controlled carbon–carbon bond-forming reaction.

In Fischer and Sarpong’s work1, the late-stage intermediate was a lycodine derivative (a triflate; Fig. 1a). The authors converted this compound into another intermediate (known as a boronic ester) in a two-step process that involved an inventive use of a recently discov-ered iridium-catalysed reaction5. This process ‘desymmetrized’ the reactivity of the lyco-dine — that is, it allowed the authors to form a bond between different carbon atoms in the triflate and the boronate ester, yielding the non-symmetrical core of complanadine A.

Siegel and colleagues’ late-stage intermedi-ate2, an alkyne–nitrile, was less obviously related

structures and offer advances towards what is potentially the biggest challenge for twenty-first-century synthetic chemistry: not just pre-paring a given target molecule, but doing so with a level of efficiency and cost-effectiveness that rivals that of nature itself. ■

Scott A. Snyder is in the Department of Chemistry, Columbia University, New York, New York 10027, USA. e-mail: sas2197@columbia.edu

1. Fischer, D. F. & Sarpong, R. J. Am. Chem. Soc. 132, 5926–5927 (2010).

2. Yuan, C., Chang, C.-T., Axelrod, A. & Siegel, D. J. Am. Chem. Soc. 132, 5924–5925 (2010).

3. Nicolaou, K. C. & Snyder, S. A. Classics in Total Synthesis II (Wiley-VCH, 2003).

4. Kobayashi, J., Hirasawa, Y., Yoshida, N. & Morita, H. Tetrahedron Lett. 41, 9069–9073 (2000).

5. Ishiyama, T. et al. J. Am. Chem. Soc. 124, 390–391 (2002).6. Vollhardt, K. P. C. Angew. Chem. Int. Edn 23, 539–556 (1984).

CorrectionThe News & Views article “Nuclear physics: Doubly magic tin” by Paul Cottle (Nature 465, 430–431; 2010) gave the isotope of lead as lead-28 (28Pb), when the correct notation is of course lead-208 (208Pb).

N

HN

Me

H

H

Lycodine

NNH

MeHH

N

HN

Me

H

H

N

N

Me

H

H

N

N

Me

H

HO

BOMe

Me

MeMe

Triflate

PGPG

NPG

a

Boronic esterTriflate

TfO

b

NNPG

MeHH

SiMe3

NN

MeHH

N

Me

H

H

Two steps

Me3Si

(+)-Complanadine A

Twosteps

Twosteps

Alkyne–nitrile

Alkyne–nitrile

(+)-Complanadine A

[2+2+2]reaction

[2+2+2]reaction

then two more steps PG

N

MeHH N

PG

Figure 1 | Synthetic routes to complanadine A. The natural product complanadine A consists of two identical structural units, which correspond to another natural product, lycodine. The bond between the two lycodines in complanadine A (green) connects different carbon atoms in each unit. Me represents a methyl group. a, Fischer and Sarpong1 prepared complanadine A by making a lycodine analogue (a triflate) as an intermediate. They converted this in two steps to a boronic ester, which they then reacted with another equivalent of the triflate, forming the target molecule in two steps from the boronic ester. Tf represents SO2CF3. PG is a protecting group — a chemical group that prevents the nitrogen atom to which it is attached from taking part in unwanted reactions. b, Siegel and colleagues2 used a different key precursor, an alkyne–nitrile, in their synthesis of complanadine A. They subjected this precursor to a [2 + 2 + 2] reaction to form the pyridine ring (red) of lycodine, and modified the product in a further two steps. The authors reacted the resulting compound with another alkyne–nitrile in a second [2 + 2 + 2] reaction, generating an intermediate that contained the core structure of complanadine A. They then converted this intermediate into complanadine A in two steps.

to lycodine (Fig. 1b). In particular, it lacked a pyridine ring system (a nitrogen-containing ring of atoms, similar to benzene). The authors used a [2 + 2 + 2] reaction6 to build a pyridine ring into their intermediate, modified the prod-uct in two steps and then reacted the resulting compound with another alkyne–nitrile in a second [2 + 2 + 2] reaction to form the core structure of complanadine A. Usually, [2 + 2 + 2] reactions yield multiple products, but Siegel and colleagues developed and optimized their sys-tem to give them the control needed for success, thereby enabling them to ‘desymmetrize’ their single building block in a different way from that used by Fischer and Sarpong1, and thus to ultimately access the non-symmetrical target.

Because the two research groups evaluated the structure of complanadine A from distinct perspectives, both of these syntheses1,2 involve very different steps. Yet both teams found creative ways to use the idea of symmetry inherent in a common intermediate, cou-pled with power ful synthetic methods, to ultimately fashion the target molecule. These papers, when taken together, may well reveal how to access other non-symmetrical dimeric

561

NATURE|Vol 465|3 June 2010 NEWS & VIEWS

© 20 Macmillan Publishers Limited. All rights reserved10