news and views annually degasses about 1 gigatonne (109 not nitrate exhaustion. on the microbial algae. The growth perfor- tonnes) of CO2 (ref. 3), equivalent to 20 per Yet the paradox persists. Virtually all phy- mance of the diatoms is all the more surpris- cent of current anthropogenic output. The toplankton species, including the picoplank- ing as nitrate reduction requires energy regression lines and intercepts of dissolved ton, are able to use nitrate; indeed, phyto- and the mediating enzyme contains iron. inorganic carbon, silicate and nitrate con- plankton other than diatoms routinely Indeed, why diatoms can be so much more centrations, measured in the upper 200 m of exhaust nitrate in the surface waters of the efficient than the other algae despite the the EUZ, indicate that silicate availability non-HNLC ocean. So why does this not hap- nitrate handicap needs to be explained. regulates both carbon and nitrate uptake pen in the EUZ and other low-silicate HNLC Balancing pelagic ecosystem budgets is and fate. The slope of the highly significant regions? Differences in grazing pressure still an art because we know so little about the 5 regression between nitrate and silicate is 1, have been proposed as an explanation . One abilities and predilections of the organisms8 which coincides with the known require- widely held view is that the small algae of the and their interactions with one another5. ment of these two elements by diatoms. The microbial loop are kept in check by heavy Whatever the outcome of studies on the lim- intercept represents the excess nitrate (about grazing pressure, whereas diatoms, because iting factors in the various HNLC regions 4 mmol m–3) that confers HNLC status on of their larger size (and possibly also the pro- and their subsystems, the status of diatoms as the EUZ. tection offered by the silica shell), are less key players will not be challenged. The work- Silicate is bound in mineral shells (bio- prone to being grazed by the smaller proto- horses running pelagic systems are recruited genic silica) of no food value. In the EUZ zoa6. So relaxation of a limiting factor (such from this algal group: their new production model, the diatom shells are packaged in as iron) results in accumulation of diatom not only fuels the food chains leading to fish zooplankton faeces and exported wholesale but not picoplankton cells. Whether contin- but also provides the raw material driving from the surface layer. In contrast, nitrogen ued iron fertilization will eventually lead to the biological pump and ultimately the great is selectively retained by the diatom grazers nitrate exhaustion by non-diatom phyto- biogeochemical cycles of the ocean. It is time which, in the course of their metabolism, plankton in low-silicate HNLC regions we gained a better understanding of the excrete ammonia; in turn the ammonia is remains to be tested. properties that make diatoms so special. assimilated by the photosynthesizing cyano- In Dugdale and Wilkerson’s steady-state Victor Smetacek is at the Alfred-Wegener Institute bacteria (<2 Ȗm) and small flagellates (2–10 EUZ model, the biomass-to-production for Polar and Marine Research, Am Handelshafen Ȗm) of the ‘microbial loop’. This ubiquitous ratio of the diatoms indicates that they were 12, 27570 Bremerhaven, Germany. community (pico- and nanoplankton) is growing at least as fast as, if not faster than, e-mail: [email protected] additionally composed of small protozoa the microbial algae. To maintain steady state, 1. Dugdale, R. C. & Wilkerson, F. P. Nature 391, 270–273 (1998). that feed on the minute algae and bacteria, as grazing pressure on the diatoms, presumably 2. Doney, S. C. Nature 389, 905–906 (1997). well as on one another, which also release by copepods (zooplanktonic crustacea, 3. Tans, P. P., Fung, I. Y. & Takahashi, T. Science 247, 1431–1438 ammonia (see Fig. 2 of Dugdale and equipped with mandibles edged with silica (1990). 4. Coale, K. H. et al. Nature 383, 495–501 (1996). Wilkerson’s paper, page 272). The ammonia- the better to crush diatom shells with), must 5. Banse, K. Oceanography 7, 13–20 (1994). based production within this tightly geared have been similar to or even higher than that 6. Banse, K. ICES J. Mar. Sci. 52, 265–277 (1995). community — termed regenerated produc- tion — contributed four-fifths of the total daily production (new plus regenerated) in the EUZ. Because of the low variability Hox gene variation and in nutrient and biomass concentrations Axel Meyer observed over different cruises and seasons, the pelagic system of the EUZ seems to have settled into a steady state which can be oth developmental and evolutionary questioning of this commonly held hypoth- likened to a silicate-limited chemostat. biologists try to explain patterns in the esis — that genomic and morphological The silicate-to-nitrate ratios in the other Bdiversity among organisms, and the complexity are causally linked2–7. HNLC regions — the Southern Ocean and Hox genes encode a class of transcription Using an experimental approach based the subarctic Pacific — are much higher than factors that have provided ample material on the polymerase chain reaction, Prince et in the EUZ, yet fairly similar community for such discussions. Because they may be al. unambiguously identified 34 Hox genes structures, total production rates and ratios pivotal in specifying regional identity in and determined their linkage with somatic- of new-to-regenerated production are rou- body plans, differences in their expression cell hybrids. Surprisingly, they found that tinely measured there. Diatom growth is evi- could (at least partly) explain the evolution the has three Hox genes (HoxC3, dently not silicon-limited, so other factors, of phyla. The Hox genes are arranged HoxA8 and HoxB10), with no direct mouse such as deep mixing, iron deficiency and in genomic clusters and, importantly, they equivalents (Fig. 1). Moreover, the expres- heavy grazing pressure, have been invoked to are expressed in a spatially colinear fashion sion domains of the anterior Hox genes are explain the HNLC condition. An in situ iron — anterior genes are expressed early in partly overlapping, and restricted to a short- fertilization experiment (IronEx II), carried development and towards the front of the er anterior region. Possibly the most impor- out in the South Equatorial Current adjacent body, posterior genes later in development tant finding is that the zebrafish has at least to the EUZ, yielded a spectacular diatom and in more distal portions of the body. two additional Hox gene clusters for a total of bloom4 and showed that iron availability In invertebrates, only a single Hox gene six and not, as previously thought8, the typi- indeed limited new production in this region. cluster has been found (although it is split in cal vertebrate number of four. All of the Dugdale and Wilkerson argue that, as the ). The common ancestor of all genes on these additional clusters have prob- EUZ diatoms are already silicate-limited, chordates is surmised to have had a single ably not yet been discovered, and three are adding iron should have no effect on them. cluster as well. This cluster is thought to reported so far1. They suggest that iron upwelling with the have duplicated to four clusters (A–D) on These two additional clusters lead us to other nutrients in the EUZ is sufficient to different , accompanying the question a simple, ‘more clusters, more com- meet the demands of the diatoms. In the increasing complexity of body plans during plexity’ model of evolutionary diversifica- iron-limited waters of the South Equatorial the evolution of vertebrates (Fig. 1). But a tion — in terms of phenotypic complexity, Current, iron-fertilized diatom growth report by Prince et al.1, shortly to appear however measured, a zebrafish is probably would eventually be halted by silicate but in Development, is likely to cause some not 50 per cent more complex than a mouse

NATURE | VOL 391 | 15 JANUARY 1998 225 Nature © Macmillan Publishers Ltd 1998 news and views

Zebrafish (42) Pufferfish (31) Mouse (39)

13 12 11 10 9 8 7 6 5 4 3 2 1 13 12 11 10 9 8 7 6 5 4 3 2 1 13 12 11 10 9 8 7 6 5 4 3 2 1

A A A B B B C C C D D D X Y 100 YEARS AGO Loss of 1 Hox gene Loss of 12 Hox genes Loss of 5 Hox genes Messrs. Swan Sonnenschein and Co. Cluster duplications announce that they will shortly publish a 13 12 11 10 9 8 7 6 5 4 3 2 1 work, entitled “The Wonderful Century: its8 Hypothetical A Successes and its Failures,” by Dr. Alfred fish ancestral B (43) condition C R. Wallace, F.R.S. The object of the D Lost Hox gene volume is to give a short descriptive Loss of 1 Hox gene Hox pseudogene sketch of all the more important mechanical inventions and scientific 13 12 11 10 9 8 7 6 5 4 3 2 1 discoveries which are distinctive of the Hypothetical A B (44) nineteenth century. ... The author vertebrate ancestral C condition D maintains that our century is altogether unique; that it differs from the eighteenth Loss of 8 Hox genes or seventeenth centuries, not merely as those differed from the centuries which 13 12 11 10 9 8 7 6 5 4 3 2 1

A immediately preceded them, but that it B (52) C has initiated a new era, and that it may D be more properly compared with the whole preceding historical period. Cluster duplications

13 12 11 10 9 8 7 6 5 4 3 2 1 The January number of the National Hypothetical chordate Review has an admirable article by Mr. ancestral condition Amphioxus (13) Gerald Arbuthnot, entitled “In Defence of Figure 1 Evolutionary conservation of the Hox gene clusters in zebrafish, pufferfish, mouse and the Muzzle.” The temperate spirit in Amphioxus. The zebrafish Hox gene clusters are shown with the 34 genes confirmed by Prince et al.1, which it is written, and the conscientious and an additional eight, for the total of 42 that the zebrafish is likely to have. Ten gene losses manner in which the statistics referred to differentiate the mouse from the pufferfish, but there are only three losses between the mouse and have been collected, ought to materially zebrafish, even though fish and tetrapods shared their last common ancestor about 400 million years strengthen the hands of those who are ago. The common ancestor of human and mouse probably lived around 60 million years ago, but the upholding the muzzling order for dogs, in architectures of their Hox gene clusters are identical. The losses in the AbdA (abdominal A)-related the face of the selfish and short-sighted genes (light green) in the D-cluster probably occurred early in the history of vertebrates. opposition which it is receiving from a (Figure drawn by Heike Haunstetter.) certain section of the public. From Nature 13 January 1898. or a human. The extra clusters cannot be precisely known), they would be more likely explained by entire- duplications to contain a D-like Hox than a B- 50 YEARS AGO because, although is known from or C-cluster. The suggestion that the D- and A symposium arranged by the New York other carp-like fish9 and is common in A-clusters are the oldest also seems to fit the Academy of Sciences and held in salmonids, zebrafish are diploid. It also observation that the D-cluster is the most December 1946 on “Nutrition in Relation seems unlikely that the additional clusters ‘deteriorated’ of all (Fig. 1). to Cancer” covered a wide field and are remnants of a polyploid ancestral condi- Following the principle of Dollo parsi- included a number of interesting articles tion. Instead, the Hox-cluster duplications mony — which assumes that losses of genes which have now been published. ... in zebrafish might be a unique evolutionary are much more common and likely than Although it may seem disappointing that event. But such events may turn out to be independent evolutionary origins12 — we after so much study of cancer the common, at least in fish. can speculate which Hox genes might have fundamental cause or nature of it is Prince and colleagues’ work on zebra- been present in the common ancestors of unknown, the papers [in the symposium fish1, combined with studies on the puffer- vertebrates, tetrapods and fish (Fig. 1). Based volume] show advances. The carcinogenic fish10, now enables us to reconstruct the on the genomic organizations available so process can often be influenced by diet, evolutionary history of the Hox gene clusters far, the rates of evolution of Hox clusters which means that the process can be in vertebrates (Fig. 1). The initial chordate do not seem to be constant. For example, resolved into separate parts, and this ancestral cluster of 13 Hox genes (the archi- whereas the zebrafish is likely to have lost must help in the understanding of the tecture that is still present in the cephalo- only one Hox gene since it shared a common process. Many of the speakers at the chordate Amphioxus7) probably duplicated ancestor with the pufferfish (probably more symposium compared carcinogenesis to in a three-step process, to form four com- than 200 million years ago), the losses along mutations. Both cancer and mutations can plete clusters with a total of 52 genes. One the pufferfish lineage were possibly several be induced in living organisms by similar phylogenetic study11 indicates that the D- times faster (12 Hox genes were lost, if the agents. The hypothesis that cancer is a cluster is the most ancestral, and that the B- zebrafish really has 42 Hox genes) (Fig. 1). somatic mutation relates carcinogenesis and C-clusters are the youngest. Hagfish Ignoring the additional clusters of the to other biological changes and the and lamprey are phylogenetic intermediates zebrafish, and estimating that it has 42 Hox stability of the nuclear and cytoplasmic between Amphioxus and more derived ver- genes, we find that 13 differences separate genes. tebrates such as zebrafish. So, if these fishes the zebrafish from the pufferfish. In the From Nature 17 January 1948. have only two or three clusters (which is not pufferfish, the HoxC1 and C3 genes are still

NATURE | VOL 391 | 15 JANUARY 1998 227 Nature © Macmillan Publishers Ltd 1998 news and views recognizable, but they are only pseudogenes the evolution of genes or networks of inter- stable isobars, in which more than one stable (genes that are not transcribed). They might actions through regulatory elements drives nucleus exists at a given mass number. have lost their function at different points in most morphological diversification might, There are three basic nuclear processes the evolution of fish, because HoxC3, at least, then, have different answers in different evo- that produce heavy elements: the s-, r- and p- is still present in the zebrafish. lutionary lineages. In the most species-rich processes, standing for slow, rapid and pro- The apparent acceleration of genomic group of vertebrates — fish — organization ton-rich. The abundances produced by each evolution along the pufferfish lineage might of the Hox genes might not be completely of these5 are shown in Fig. 1. be correlated with accelerated morphologi- constrained by interwoven regulatory net- The s-process probably takes place in red cal evolution. Pufferfish belong to one of the works, and differentiation might be driven giant stars that are burning helium, with a most morphologically derived groups of fish, by gene evolution. However, in the lineage given nucleus capturing a neutron perhaps8 and they lack ribs, pelvic fins and the pelvic that leads to reptiles and mammals, the dri- every few thousand years. But only fractions girdle. Are the missing genes those that are no ving force behind morphological diversifica- of a second separate successive photodisinte- longer necessary because these structures tion might have been newly evolving interac- grations in the p-process or neutron captures have been lost during evolution? If so, the tions in networks of regulatory elements16. in the r-process, implying that these process- missing Hox genes in the pufferfish might Axel Meyer is in the Department of Biology, es must occur in supernova explosions. So also be absent in the other groups of fish University of Konstanz, D-78457 Konstanz, some details of supernova explosions can be which have secondary loss of pelvic fins (such Germany. deduced from isotopic analysis of meteoritic as eels) or even tail fins (for example, the e-mail: [email protected] material — in particular, interstellar grains ocean sunfish Mola mola). Moreover, the loss 1. Prince, V. E., Jolly, J., Ekker, M. & Ho, R. Development 125, that have undergone relatively little thermal of Hox genes might also be accompanied by 407–420 (1998). processing during and since the formation of the secondary loss (or simplification) of 2. Akam, M. Phil. Trans. R. Soc. 349, 313–319 (1995). the Solar System. appendages in land vertebrates, such as in 3. Holland, P. W. H. & Garcia-Fernández, J. Dev. Biol. 173, One important recent discovery was that 382–395 (1996). 6 limbless amphibians, reptiles and whales. 4. Lewis, E. B. Nature 276, 565–570 (1978). there are two different r-processes : one What selective forces maintain or modify 5. Krumlauf, R. Cell 78, 1991–202 (1994). responsible for the r-process nuclides up to genomic organizations? The observation 6. Duboule, D. & Morata, G. Trends Genet. 10, 358–364 (1994). about A = 140 (where A is the mass number), that Hox genes are clustered, and that the 7. Garcia-Fernández, J. & Holland, P. W. H. Nature 370, 563–566 the other for nuclides aboveA = 140. The dis- (1994). architecture of these clusters is highly con- 8. Misof, B. Y. & Wagner, G. P. Mol. Phylogen. Evol. 5, 309–322 covery stemmed from extinct radionuclides, served in evolution, has led to the suggestion (1995). which lived long enough to survive with that the regulatory elements that control 9. Buth, D. G., Dowling, T. E. & Gold, J. R. in Cyprinid Fish, measurable abundances from the time of Systematics, Biology and Exploitation (eds Winfield, I. J. & expression of the Hox genes cannot be sepa- Nelson, J. S.) 83–126 (Chapman & Hall, London, 1991). their production to their injection into the rated from these genes without jeopardizing 10.Aparicio, S. et al. Nature Genet. 16, 79–83 (1997). forming Solar System, but not long enough their proper functioning and, possibly, deter- 11.Bailey, W. J., Kim, J., Wagner, G. P. & Ruddle, F. H. Mol. Biol. to be measurably present now. They are mination of morphology along the antero- Evol. 14, 843–853 (1997). detected through excesses of the daughter 13,14 12.Meyer, A. in New Uses for New Phylogenies (eds Harvey, P. H., posterior axis. For vertebrates these ideas Leigh Brown, A. J., Maynard Smith, J. & Nee, S.) 322–340 elements that result from their decay. 182 have been partially confirmed experimen- (Oxford Univ. Press, 1996). The abundance of Hf in the early Solar tally15, and this tight functional linkage might 13.Gérard, M. et al. Genes Dev. 10, 2326–2334 (1996). System is consistent6 with the continuous be particularly strong along the lineage that 14.Gould, A. et al. Genes Dev. 11, 900–913 (1997). production of the actinides in the Galaxy, 15.Mann, R. S. BioEssays 19, 661–664 (1997). leads to reptiles and mammals (Fig. 1). 16.Davidson, E. H., Peterson, K. J. & Cameron, R. A. Science 270, with mixing to maintain a roughly constant The long-standing question of whether 1319–1325 (1995). abundance level on a timescale consistent with the mean life of 182Hf — about 107 years. Isotope astrophysics But trouble arises from two lighter radio- nuclides, 107Pd and 129I. If these were made Two cradles for the heavy elements by the same r-process that produced the A. G. W. Cameron 10–5

here do the Solar System’s heavy ments within stellar interiors in which many 10–5 )

elements come from? We know of those processes take place. The process 6 Wthat many of the elements heavier whose environment has proved most elusive 0 10–5 than iron must have been formed in super- has been the r-process, in which neutron novae; but exactly how? Taken together, capture takes place on a very rapid timescale. 10–5 three new papers (one on page 261 of this In the evolution of a star considerably issue1, and two to appear in the Astrophysical more massive than the Sun, nuclear fusion 10–5 Journal 2,3) imply that two different types of reactions build toward products in which the supernova are responsible for elements in binding energy per nucleon becomes maxi- 10–5 Abundance (silicon = 1 different mass regimes. mized. This produces an abundance peak at s–process It is now some four decades since Suess 56Fe. Making nuclides much heavier than this 10–5 r–process and Urey4, largely working from analyses of involves neutron capture, which generally p–process 10–5 meteorites, assembled a table of abundances produces nuclides on the neutron-rich side 80 100 120 140 160 180 200 220 of the elements that enabled nuclear physi- of the region of stable elements, known as the Mass number cists to identify the principal processes in valley of beta stability. Only in the violence of stellar interiors that had produced those a supernova explosion can the nuclides on Figure 1 The solar abundances of the nuclides, as elements. The actual nuclear physics of those the neutron-deficient side be produced, a function of mass number, showing the p-, s- processes became reasonably well under- primarily by losing nucleons in photodis- and r-process contributions. For the s- and r- stood within the following few years. How- integration. It is possible to distinguish these processes, average smooth lines have been drawn ever, it has taken considerably longer for us processes by examining the abundances of through the characteristic zig-zag patterns of the to understand the astrophysical environ- nuclei in the Solar System, particularly of odd and even mass numbers. (From ref. 5.)

228 NATURE | VOL 391 | 15 JANUARY 1998 Nature © Macmillan Publishers Ltd 1998