Atypical Combinations and Scientific Impact

Brian Uzzi et al.

Science 342, 468 (2013);
DOI: 10.1126/science.1240474

This copy is for your personal, non-commercial use only.

If you wish to distribute this article to others, you can order high-quality copies for your
colleagues, clients, or customers by clicking here.

Downloaded from www.sciencemag.org on October 27, 2013

Permission to republish or repurpose articles or portions of articles can be obtained by
following the guidelines here.

The following resources related to this article are available online at

www.sciencemag.org (this information is current as of October 27, 2013 ):

Updated information and services, including high-resolution figures, can be found in the online
version of this article at:
http://www.sciencemag.org/content/342/6157/468.full.html
Supporting Online Material can be found at:
http://www.sciencemag.org/content/suppl/2013/10/24/342.6157.468.DC1.html
This article cites 20 articles, 9 of which can be accessed free:
http://www.sciencemag.org/content/342/6157/468.full.html#ref-list-1

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
2013 by the American Association for the Advancement of Science; all rights reserved. The title Science is a
registered trademark of AAAS.
REPORTS
range of likely core temperatures. However, fur- 10. A. M. Dziewonski, D. L. Anderson, Phys. Earth Planet. 28. F. Delogu, J. Phys. Chem. B 110, 3281–3287 (2006).
ther investigations into multicomponent systems Inter. 25, 297–356 (1981). 29. F. Delogu, J. Phys. Condens. Matter 18, 5639–5653
11. D. Antonangeli et al., Earth Planet. Sci. Lett. 295, (2006).
are essential to fully understand their effect on the 292–296 (2010). 30. G. Manai, F. Delogu, Physica B 392, 288–297 (2007).
elastic properties of the core. Overall, our results 12. Z. Mao et al., Proc. Natl. Acad. Sci. U.S.A. 109, 31. G. Manai, F. Delogu, J. Mater. Sci. 42, 6672–6683
demonstrate that the inner core is likely to be in the 10239–10244 (2012). (2007).
13. M.-H. Nadal, P. Le Poac, J. Appl. Phys. 93, 2472–2480 32. J. P. Poirier, Introduction to the Physics of the Earth’s
strongly nonlinear regime; hence, there is no need
(2003). Interior (Cambridge Univ. Press, Cambridge, 2000),
to invoke special circumstances such as strong 14. M.-H. Nadal, C. Hubert, G. Ravel-Chapuis, J. Alloys Compds. pp. 230–244.
anelasticity, partial melts, or combinations of crys- 444–445, 265–267 (2007).
talline phases in order to match the observed 15. D. Alfè, G. D. Price, M. J. Gillan, Phys. Rev. B 65, 165118 Acknowledgments: Supported by Natural Environment
(2002). Research Council grant NE/H003975/1 (L.V.). Calculations were
seismic velocities and densities of the inner core. 16. D. Alfè, Phys. Rev. B 79, 060101–060104 (2009). performed in the HECTOR supercomputer facility. Computer
17. E. Sola, D. Alfè, Phys. Rev. Lett. 103, 078501–078504 code VASP is available at www.vasp.at. The data presented in
References and Notes (2009). this paper are given in the main text and in the supplementary
1. F. Birch, J. Geophys. Res. 69, 4377–4388 (1964). 18. G. Morard, J. Bouchet, D. Valencia, S. Mazevet, F. Guyot, materials. B.M. performed research, analyzed data, and wrote
2. J.-P. Poirier, Phys. Earth Planet. Inter. 85, 319–337 High Energy Density Phys. 7, 141–144 (2011). the paper. L.V. designed research, analyzed data, and wrote
(1994). 19. See supplementary materials on Science Online. the paper. J.B. analyzed data and wrote the paper. I.G.W.
3. A. Cao, B. Romanowicz, N. Takeuchi, Science 308, 20. L. Vočadlo, D. Dobson, I. G. Wood, Earth Planet. Sci. Lett. designed the research, analyzed data, and wrote the paper.
1453–1455 (2005). 288, 534–538 (2009).
4. L. Vočadlo, Earth Planet. Sci. Lett. 254, 227–232 (2007). 21. Y. P. Varshni, Phys. Rev. B 2, 3952–3958 (1970). Supplementary Materials
5. X. Sha, R. E. Cohen, Phys. Rev. B 81, 094105–094110 22. M. W. Guinan, D. J. Steinberg, J. Phys. Chem. Solids 35, www.sciencemag.org/content/342/6157/466/suppl/DC1
(2010). 1501–1512 (1974). Materials and Methods
6. X. Sha, R. E. Cohen, Geophys. Res. Lett. 37, 23. D. R. Nelson, Defects and Geometry in Condensed Matter Supplementary Text
L10302–L10305 (2010). Physics (Cambridge Univ. Press, Cambridge, 2002). Figs. S1 to S3
7. D. Antonangeli et al., Earth Planet. Sci. Lett. 225, 24. V. Sorkin, E. Polturak, J. Adler, Phys. Rev. B 68, Table S1
243–251 (2004). 174102–174107 (2003). References (33–43)
8. A. P. Kantor et al., Phys. Earth Planet. Inter. 164, 83–89 25. V. Sorkin, E. Polturak, J. Adler, Phys. Rev. B 68,
(2007). 174103–174109 (2003). 23 July 2013; accepted 26 September 2013
9. B. Martorell, J. Brodholt, I. G. Wood, L. Vočadlo, 26. K. Lu, Y. Li, Phys. Rev. Lett. 80, 4474–4477 (1998). Published online 10 October 2013;
Earth Planet. Sci. Lett. 365, 143–151 (2013). 27. F. Delogu, J. Phys. Chem. B 110, 12645–12652 (2006). 10.1126/science.1243651

between extending science with atypical com-

Atypical Combinations and binations of knowledge while maintaining the
advantages of conventional domain-level think-
Scientific Impact ing is critical to the link between innovativeness
and impact. However, little is known about the
composition of this balance or how scientists
Brian Uzzi,1,2 Satyam Mukherjee,1,2 Michael Stringer,2,3 Ben Jones1,4*
can achieve it.
In this study, we examined 17.9 million re-
Novelty is an essential feature of creative ideas, yet the building blocks of new ideas are often
search articles in the Web of Science (WOS) to
embodied in existing knowledge. From this perspective, balancing atypical knowledge with
see how prior work is combined. We present facts
conventional knowledge may be critical to the link between innovativeness and impact. Our
that indicate (i) the extent to which scientific pa-
analysis of 17.9 million papers spanning all scientific fields suggests that science follows a
pers reference novel versus conventional combi-
nearly universal pattern: The highest-impact science is primarily grounded in exceptionally
nations of prior work, (ii) the relative impact of
conventional combinations of prior work yet simultaneously features an intrusion of unusual
papers based on the combinations they draw
combinations. Papers of this type were twice as likely to be highly cited works. Novel combinations
upon, and (iii) how (i) and (ii) are associated with
of prior work are rare, yet teams are 37.7% more likely than solo authors to insert novel
collaboration.
combinations into familiar knowledge domains.
We considered pairwise combinations of refer-
ences in the bibliography of each paper (23, 24).
cientific enterprises are increasingly con- Yet the production and consumption of We counted the frequency of each co-citation

S cerned that research within narrow bound-

aries is unlikely to be the source of the most
fruitful ideas (1). Models of creativity empha-
boundary-spanning ideas can also raise well-
known challenges (16–21). If, as Einstein be-
lieved (21), individual scientists inevitably become
pair across all papers published that year in the
WOS and compared these observed frequencies
to those expected by chance, using randomized
size that innovation is spurred through original narrower in their expertise as the body of sci- citation networks. In the randomized citation
combinations that spark new insights (2–10). Cur- entific knowledge expands, then reaching ef- networks, all citation links between all papers
rent interest in team science and how scientists fectively across boundaries may be increasingly in the WOS were switched by means of a Monte
search for ideas is premised in part on the idea challenging (4), especially given the difficulty Carlo algorithm. The switching algorithm pre-
that teams can span scientific specialties, effec- of searching unfamiliar domains (17, 18). More- serves the total citation counts to and from each
tively combining knowledge that prompts scien- over, novel ideas can be difficult to absorb (19) paper and the distribution of these citation
tific breakthroughs (11–15). and communicate, leading scientists to inten- counts forward and backward in time to ensure
tionally display conventionality. In his Principia, that a paper (or journal) with n citations in the
1
Kellogg School of Management, Northwestern University, Newton presented his laws of gravitation using observed network will have n citations in the
2001 Sheridan Road, Evanston, IL 60208, USA. 2Northwestern accepted geometry rather than his newly de- randomized network. For both the observed and
Institute on Complex Systems, Northwestern University, 600 veloped calculus, despite the latter’s impor- the randomized paper-to-paper citation networks,
Foster, Evanston, IL 60208, USA. 3Datascope Analytics, 180 tance in developing his insights (22). Similarly, we aggregated counts of paper pairs into their
West Adams Street, Chicago, IL 60603, USA. 4National Bureau of
Economic Research, 1050 Massachusetts Avenue, Cambridge, MA Darwin devoted the first part of the Origin of respective journal pairs to focus on domain-level
02138, USA. Species to conventional, well-accepted knowl- combinations (24–26). In the data, there were
*Corresponding author. E-mail: [email protected]. edge about the selective breeding of dogs, cat- over 122 million potential journal pairs created
edu tle, and birds. From this viewpoint, the balance by the 15,613 journals indexed in the WOS.

468 25 OCTOBER 2013 VOL 342 SCIENCE www.sciencemag.org

 REPORTS
Comparing the observed frequency with the in the 1990s. This pattern indicates that observed 69.0 in the 1980s and 99.5 in the 1990s. More-
frequency distribution created with the random- journal pairings from the same WOS disciplines over, papers with a median z score below zero
ized citation networks, we generated a z score tend to be conventional, and interdisciplinary WOS are rare. In the 1980s, only 3.54% of papers had
for each journal pair. This normalized measure journal pairings are less substantially conventional this feature, whereas in the 1990s the percentage
describes whether any given pair appeared novel but still not consistently novel. fell to 2.67%, indicating a persistent and promi-
or conventional. Z scores above zero indicate The above method assigns each paper a dis- nent tendency for high conventionality.
pairs that appeared more often in the observed tribution of journal pair z scores based on the Focusing on each paper’s left tail combina-
data than expected by chance, indicating rela- paper’s reference list (Fig. 1A). To characterize a tions, we found that even among the paper’s
tively common or “conventional” pairings. Z scores paper’s tendency to draw together conventional relatively unusual journal combinations, the ma-
below zero indicate pairs that appear less often and novel combinations of prior work, we exam- jority of papers did not feature atypical journal
in the observed WOS than expected by chance, ined two summary statistics. First, to characterize pairs. Figure 1C shows that 40.8% of the pa-
indicating relatively atypical or “novel” pair- the central tendency of a paper’s combinations, we pers in 1980s and 40.7% in the 1990s have a
ings. For example, in the year 1980, the pairing considered the paper’s median z score. The me- 10th-percentile z score below zero. Overall, by
Tetrahedron and Experientia had a high z score dian allows us to characterize conventionality in these measures, science typically relies on highly
(21.55) indicating a conventional pairing, where- the paper’s main mass of combinations. Second, conventional combinations and rarely incorpo-
as Tetrahedron paired with Life Sciences had we considered the paper’s 10th-percentile z score. rates journal pairs that are uncommon compared
a negative z score (–17.67), indicating a pair- The left tail allows us to characterize the paper’s to chance. Analyses in the supplementary mate-
ing more unusual than chance. The supplemen- more unusual combinations, where novelty may rials (fig. S6) show that these empirical regular-
tary materials detail these computations, the null reside. ities for the WOS taken as a whole are largely
model, and an illustrative example (table S1 and We found that papers typically relied on replicated on a field-by-field basis and across time.
figs. S1 to S3). very high degrees of conventionality. Figure 1B Our next finding indicates a powerful rela-
As a simple validation of the z score mea- presents the distribution of papers’ median z tionship between combinations of prior work and
sure, we found that journal pairs from the same scores for the WOS in the indicated decades. ensuing impact. Figure 2 presents the probability
WOS disciplinary designation had significantly Considering that a z score below zero represents of a “hit” paper, conditional on the combination
higher z scores than did interdisciplinary journal a journal pair that occurs less often than expected of its referenced journal pairs. Hit papers are
pairs (table S3 and fig. S11). At the same time, by chance, the analysis of median z scores sug- operationalized as those in the upper 5th per-
only a minority (40.1%) of interdisciplinary jour- gests very high degrees of conventionality. Half centile of citations received across the whole data
nal pairs were novel, having z scores below zero of the papers have median z scores exceeding set, as measured by total citations through 8 years

Fig. 1. Novelty and conventionality in science. For a sample paper,

(A) shows the distribution of z scores for that paper’s journal pairings. The
z score shows how common a journal pairing is as compared to chance. For
each paper, we take two summary measures: its median z score, capturing the
paper’s central tendency in combining prior work, and the 10th-percentile
z score, capturing the paper’s journal pairings that are relatively unusual.
For the population of papers, we then consider these values across all papers
in the WOS published in the 1980s or 1990s. (B) considers the median z
scores and shows that the vast majority of papers displays a high propensity
for conventionality; in the 1980s and 1990s, fewer than 4% of papers have
median z scores below 0 and more than 50% of papers have median z scores
above 64. (C) considers the 10th-percentile z scores, which further suggest a
propensity for conventionality; only 41% of papers in the 1980s and 1990s
have a 10th-percentile z score below 0. Overall, by these measures, science
rarely draws on atypical pairings of prior work.

www.sciencemag.org SCIENCE VOL 342 25 OCTOBER 2013 469

REPORTS
after publication (the supplementary materials and (ii) “tail novelty” (an indicator for whether show significantly lower hit rates. Papers featur-
consider alternative definitions of hit papers). The the paper’s 10th-percentile z score is above or ing high median conventionality but low tail
vertical axis shows the probability of a hit paper below zero). novelty displayed hit rates of 5.82 out of 100 pa-
conditional on a 2 × 2 categorization indicating Papers with “high median conventionality” pers, whereas those featuring low median con-
the paper’s (i) “median conventionality” (an indi- and “high tail novelty” display a hit rate of 9.11 ventionality but high tail novelty display hit rates
cator for whether the paper’s median z score is in out of 100 papers, or nearly twice the background of 5.33 out of 100 papers. Finally, papers low
the upper or lower half of all median z scores) rate of 5 out of 100 papers. All other categories on both dimensions have hit rates of just 2.05
out of 100.
Further analyses suggest a universality of
these relationships for scientific work across time
and fields. We considered the same relationships
for different time periods (fig. S4), for different
definitions of high-impact papers (fig. S5), and
for each of 243 fields of science (fig. S6 and
table S2). These analyses confirmed the findings
above. Thus, novelty and conventionality are not
opposing factors in the production of science;
rather, papers with an injection of novelty into an
otherwise exceptionally familiar mass of prior
work are unusually likely to have high impact.
Collaboration is often claimed to produce
more novel combinations of ideas (10–14), but
the extent to which teams incorporate novel
combinations across the universe of fields is un-
known. Team-authored papers were more likely
Fig. 2. The probability of a “hit” paper, conditional on novelty and conventionality. This figure to show atypical combinations than were single-
presents the probability of a paper being in the top 5% of the citation distribution conditional on or pair-authored papers. Figure 3A shows that
two dimensions: whether a paper exhibits (i) high or low median conventionality and (ii) high or
the distribution of 10th-percentile z scores shifted
low tail novelty, as defined in the text. Papers that combine high median conventionality and high
significantly leftward as the number of authors
tail novelty are hits in 9.11 out of 100 papers, a rate nearly double the background rate of 5%.
Papers that are high on one dimension only (high median conventionality or high tail novelty but increased [Kolmogorov-Smirnov (K-S) tests in-
not both) have hit rates about half as large. Papers with low median conventionality and low tail dicate solo versus pair P = 0.016, pair versus team
novelty have hit rates of only 2.05 out of 100 papers. The sample includes all papers published in P = 0.001, team versus solo P < 0.001]. Papers
the WOS from 1990 to 2000. The supplementary materials show similar findings when considering written by one, two, or three or more authors
(i) all other decades from 1950 to 2000; (ii) hit papers defined as the top 1 or 10% by citations; showed high tail novelty in 36.1, 39.8, and 49.7%
and (iii) analyses controlling for field and other observable differences across papers, hinting at a of cases, respectively, indicating that papers with
universality of these relationships for scientific work. The difference in the hit probabilities for each three or more authors showed an increased fre-
category is statistically significant (P < 0.00001). The percentage of WOS papers in each category are: quency of high tail novelty over the solo-author
6.7% (green bar), 23% (gold bar), 26% (red bar), and 44% (blue bar). rate by 37.7%.

Fig. 3. Authorship structure, novelty, and conventionality. Team- distinct (solo versus pair P = 0.016, pair versus team P = 0.001, team
authored papers are more likely to incorporate tail novelty but without versus solo P < 0.001). In contrast, each team size shows similar distribu-
sacrificing a central tendency for high conventionality. Papers introduce tions for median conventionality [(B), K-S tests indicate no statistically
tail novelty (a 10th-percentile z score less than 0) in 36.2, 39.9, and 49.7% significant differences]. These findings suggest that a distinguishing feature
of cases for solo authors, dual authors, and three or more authors, re- of teamwork, and teams’ exceptional impact, reflects a tendency to incor-
spectively (A). K-S tests confirm that the distributions of tail novelty are porate novelty.

470 25 OCTOBER 2013 VOL 342 SCIENCE www.sciencemag.org

 REPORTS

Fig. 4. Novel and conventional combinations in the production of 85th to 95th percentile of median conventionality, after which the relationship
science. (A to C) The interplay between tail novelty, median conventionality, reverses. Third, larger teams obtain higher impact given the right mix of tail
and hit paper probabilities shows remarkable empirical regularities. First, high novelty and median conventionality. Nonetheless, at low levels of median
tail novelty papers have higher impact than low tail novelty papers at (i) any convention and tail novelty, even teams have low impact, further emphasizing
level of conventionality and (ii) regardless of authorship structure. Second, the fundamental relationship between novelty, conventionality, and impact in
increasing median conventionality is associated with higher impact up to the science.

Teams were neither more nor less likely than science draws on primarily highly conventional between extending technology with atypical com-
single authors or pairs of authors to display high combinations of prior work, with an intrusion of binations of prior ideas while embedding them in
median conventionality. Figure 3B indicates no combinations unlikely to have been joined to- conventional knowledge frames may be critical to
significant statistical difference in the median gether before. These patterns suggest that novelty human progress in many domains. Future research
z-score distributions (K-S tests indicate solo ver- and conventionality are not factors in opposition; questions also arise from our findings. Science is
sus pair P = 0.768, pair versus team P = 0.417, rather, papers that mix high tail novelty with high dynamic, with research areas shifting and new
team versus solo P = 0.164). Teams thus achieve median conventionality have nearly twice the pro- fields arising. Although we find that the regulari-
high tail novelty more often than solo authors. pensity to be unusually highly cited. ties relating novelty, conventionality, and impact
Yet, teams were not simply more novel but rather These findings have implications for theories persist across time and fields, understanding how
displayed a propensity to incorporate high tail about creativity and scientific progress. Combi- research trajectories shift and how new fields are
novelty without giving up a central tendency for nations of existing material are centerpieces in born are questions that measures of novelty and
high conventionality. theories of creativity, whether in the arts, the sci- convention may valuably inform. At root, our work
In our final analysis, we examined the inter- ences, or commercial innovation (2–4, 6–10, 16). suggests that creativity in science appears to be a
play between citation, combination, and collabo- Across the sciences, the propensity for high- nearly universal phenomenon of two extremes. At
ration using regression methods (Fig. 4). Papers impact work is sharply elevated when combina- one extreme is conventionality and at the other is
were binned into 11 equally sized categories of tions of prior work are anchored in substantial novelty. Curiously, notable advances in science
median conventionality. A separate regression conventionality, not novelty, while mixing in a left appear most closely linked not with efforts along
was run for each category of median conven- tail of combinations that are rarely seen together. one boundary or the other but with efforts that
tionality and each team size, with field fixed In part, this pattern may reflect advantages to reach toward both frontiers.
effects. The supplementary materials detail the being within the mainstream of a research trajec-
regression methodology and present additional tory, where scientists are currently focused, while References and Notes
confirmatory tests (figs. S7 to S10). being distinctive in one’s creativity. Combinations 1. Committee on Facilitating Interdisciplinary Research,
Facilitating Interdisciplinary Research (National
There were three primary findings. First, high of prior work also relate to “burden of knowl- Academies Press, Washington, DC, 2004).
tail novelty papers had higher impact than low edge” theory, which emphasizes the growing 2. H. S. Becker, Art Worlds (Univ. of California Press,
tail novelty papers, an impact advantage that oc- knowledge demands on scientists (4, 17, 21). New Berkeley, CA, 1982)
curred at any level of conventionality and regard- articles indexed by the WOS now exceed 1.4 mil- 3. R. Guimerà, B. Uzzi, J. Spiro, L. A. Amaral, Science 308,
697–702 (2005).
less of authorship structure. Second, peak impact lion per year across 251 fields, encouraging spe- 4. B. Jones, Rev. Econ. Stud. 76, 283–317 (2009).
occurred in the 85th to 95th percentile of median cialization and challenging scientists’ capacity to 5. B. F. Jones, S. Wuchty, B. Uzzi, Science 322, 1259–1262
conventionality, an exceptionally high level. This comprehend new thinking across domains. The (2008).
peak and its position appeared irrespective of finding that teams preserve high conventionality 6. J. Schumpeter, Business Cycles (McGraw-Hill, New York,
1939).
tail novelty/no tail novelty or authorship struc- yet introduce tail novelty suggests that teams help
7. A. P. Usher, A History of Mechanical Invention
ture. These generic features suggest fundamental meet the challenge of the burden of knowledge (Harvard Univ. Press, Cambridge, MA, 1954).
underlying rules relating combinations of prior by balancing domain-level depth with a capacity 8. M. L. Weitzman, Q. J. Econ. 113, 331–360 (1998).
work to the highest-impact science. for atypical combinations. 9. M. Schilling, Creat. Res. J. 17, 131–154 (2005).
Finally, Fig. 4 indicates that at virtually all This methodology considered paper and jour- 10. B. Uzzi, J. Spiro, Am. J. Sociol. 111, 447–504 (2005).
11. H. J. Falk-Krzesinski et al., Res. Eval. 20, 145–158 (2011).
mixes of tail novelty and median conventionality, nal pairings but can be applied at the level of 12. S. Wuchty, B. F. Jones, B. Uzzi, Science 316, 1036–1039
larger teams were associated with higher impact. disciplines, papers, or topics within papers, allow- (2007).
Thus, whereas teams incorporated the highest im- ing the examination of combinations of prior 13. D. Stokols, K. L. Hall, B. K. Taylor, R. P. Moser, Am. J.
pact mixes more frequently (Fig. 3), teams also work at different resolutions in future studies of Prev. Med. 35 (suppl.), S77–S89 (2008).
14. S. M. Fiore, Small Group Res. 39, 251–277 (2008).
tended to obtain higher impact for any particular creativity and scientific impact. Beyond science, 15. J. A. Evans, J. G. Foster, Science 331, 721–725 (2011).
mix (Fig. 4). Nonetheless, despite teams’ advan- links between novelty and conventionality in suc- 16. R. Collins, The Sociology of Philosophies: A Global
tage in citations across virtually all fields of sci- cessful innovation also appear. E-books retain Theory of Intellectual Change (Harvard Univ. Press,
ence (12), even teams had low impact at low page-flipping graphics to remind the reader of Cambridge MA, 1998).
17. L. Fleming, Manage. Sci. 47, 117–132 (2001).
levels of median conventionality and tail novelty. physical books, and blue jeans were designed 18. M. Schilling, E. Green, Res. Policy 40, 1321–1331 (2011).
Our analysis of 17.9 million papers across all with a familiar watch pocket to look like conven- 19. R. M. Henderson, K. B. Clark, Admin. Sci. Q. 35, 9–30
scientific fields suggests that the highest-impact tional trousers. From this viewpoint, the balance (1990).

www.sciencemag.org SCIENCE VOL 342 25 OCTOBER 2013 471

REPORTS
20. P. Azoulay, J. G. Zivin, G. Manso, Rand J. Econ. 42, Acknowledgments: Sponsored by the Northwestern raw data on which to run our analytics can obtain it via a
527–554 (2011). University Institute on Complex Systems and by the Army paid subscription to Thomson Reuters.
21. A. Einstein, The World as I See It (Citadel Press, Secaucus Research Laboratory under Cooperative Agreement Number
NJ, 1949). W911NF-09-2-0053 and Defense Advanced Research
22. D. T. Whiteside, J. Hist. Astron. 1, 116–138 (1970). Projects Agency grant BAA-11-64, Social Media in Strategic
Supplementary Materials
www.sciencemag.org/content/342/6157/468/suppl/DC1
23. H. Small, J. Am. Soc. Inf. Sci. 24, 265–269 (1973). Communication. The views and conclusions contained in this
Data and Methods
24. M. J. Stringer, M. Sales-Pardo, L. A. Nunes Amaral, J. Am. document are those of the authors and should not be
Supplementary Text
Soc. Inf. Sci. Technol. 61, 1377–1385 (2010). interpreted as representing the official policies, either
Figs. S1 to S11
25. M. Stringer, M. Sales-Pardo, L. A. Nunes Amaral, expressed or implied, of the Army Research Laboratory or
Tables S1 to S3
PLOS One 3, e1683 (2008). the U.S. government. All our summary statistics and programs
References (27, 28)
26. S. Itzkovitz, R. Milo, N. Kashtan, G. Ziv, U. Alon, are freely available on request. Our access to the WOS
Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 68, 026127 comes through a contract with Thomson Reuters that forbids 14 May 2013; accepted 20 September 2013
(2003). redistribution of their database; researchers who desire the 10.1126/science.1240474

(18, 19). ThiH lacks the second 4Fe-4S binding

A Radical Intermediate in Tyrosine domain of HydG. The electron paramagnetic reso-

Scission to the CO and CN− Ligands

nance (EPR) spectrum of the ThiH SAM-[4Fe-4S]+
cluster is altered by SAM binding but unaffected
by tyrosine incubation. In a proposed mechanism
(18, 19), the initial 5′-dA• abstracts the phenolic
of FeFe Hydrogenase H of the tyrosine substrate, forming a neutral
tyrosine radical, which is then cleaved at the Ca-
Jon M. Kuchenreuther,1∗ William K. Myers,1∗ Troy A. Stich,1 Simon J. George,1 Cb bond. Quantum chemistry calculations favored
Yaser NejatyJahromy,1 James R. Swartz,2 R. David Britt1† homolytic cleavage of this bond to form a tran-
sient glycyl radical, given its lower energy path-
The radical S-adenosylmethionine (SAM) enzyme HydG lyses free L-tyrosine to produce CO and CN− way compared with heterolytic cleavage (19).
for the assembly of the catalytic H cluster of FeFe hydrogenase. We used electron paramagnetic Modeling the differences in reactivity between
resonance spectroscopy to detect and characterize HydG reaction intermediates generated with wild-type HydG and a HydG mutant missing
a set of 2H, 13C, and 15N nuclear spin-labeled tyrosine substrates. We propose a detailed reaction the C-terminal Fe-S cluster (HydGSxxS), which
mechanism in which the radical SAM reaction, initiated at an N-terminal 4Fe-4S cluster, generates produces some CN− but no CO, led Nicolet and
a tyrosine radical bound to a C-terminal 4Fe-4S cluster. Heterolytic cleavage of this tyrosine radical co-workers (20) to propose the same glycyl radical
at the Ca-Cb bond forms a transient 4-oxidobenzyl (4OB• ) radical and a dehydroglycine bound intermediate in the HydG mechanism, building
to the C-terminal 4Fe-4S cluster. Electron and proton transfer to this 4OB• radical forms on the thermodynamic argument for the mecha-
p-cresol, with the conversion of this dehydroglycine ligand to Fe-bound CO and CN−, a key nism in ThiH. However, no radical intermediates
intermediate in the assembly of the 2Fe subunit of the H cluster. have been experimentally characterized for either
enzyme.
icrobial hydrogenase enzymes catalyze cysteine residues coordinate a 4Fe-4S cluster Here, we report on the EPR spectroscopy of

M the redox interconversion of protons

and H2 by using earth-abundant metals
in their catalytic centers, with NiFe, Fe, and
through ligation of three Fe ions. The fourth Fe of
the cluster binds SAM as a N/O chelate, and SAM
is reductively cleaved to produce methionine plus a
wild-type Shewanella oneidensis HydG (HydGWT)
expressed in Escherichia coli. Such expressed
HydG, combined with HydE and HydF, can be
FeFe classes known (1). The FeFe hydrogenases strongly oxidizing 5-deoxyadenosyl (5′-dA• ) rad- used for in vitro synthesis of the 2Fe component
are adept H2 producers, with turnover frequen- ical (8, 9). A diverse array of reactions is powered of the H cluster and concurrent activation of
cies up to 10,000 s−1 (2). Their catalytic H cluster by the H-atom abstraction capability of the result- FeFe hydrogenase apoprotein (14, 21). The use
(Fig. 1) consists of a conventional 4Fe-4S cluster ant 5′-dA•, and current bioinformatics surveys re-
linked to a unique 2Fe cluster that has two CN − veal almost 50,000 members of the radical SAM
ligands, three CO ligands, and a dithiolate bridge enzyme class (10).
with a central atom X, the chemical identity of The focus of this report is the HydG maturase.
which is ambiguous in current x-ray structures This radical SAM enzyme generates CO, CN−,
(3, 4). However, a nitrogen assignment to atom X and p-cresol by using free tyrosine as its substrate
is supported by recent reports of the assembly (11–14). Although HydG has yet to be crystallo-
of active FeFe hydrogenase by incorporation of graphically characterized, biophysical data indi-
a synthetic 2Fe subcluster with an azadithiolate cate that it has two 4Fe-4S clusters (15, 16). The
bridge into apoenzyme (5, 6). SAM–4Fe-4S cluster is bound near the N terminus,
The HydE, HydF, and HydG maturases are whereas the second cluster is modeled as coordi-
Fe-S cluster–containing accessory proteins in- nated to three cysteine residues of a CysX2CysX22Cys
volved in the biological synthesis of the 2Fe sequence near the C terminus. Sequence homol-
component of the H cluster (7). HydE and HydG ogy with other radical SAM enzymes, such as
are members of the ever-growing family of radical biotin synthase and the tyrosine lyase ThiH, point
S-adenosylmethionine (SAM) enzymes, character- to HydG having a triose phosphate isomerase
ized by a CysX3CysX2Cys motif, where the three barrel structure, in which eight a helices surround
1
a ring composed of eight parallel b strands that Fig. 1. The catalytic H cluster of FeFe hydro-
Department of Chemistry, University of California, Davis, Davis, contain a buried active site suitable for small-
CA 95616, USA. 2Department of Chemical Engineering and genases. Ball-and-stick representation (from Pro-
Department of Bioengineering, Stanford University, Stanford, CA molecule substrates (11, 17). tein Data Bank entry 3CY8) was generated by using
94305, USA. Like HydG, ThiH also lyses tyrosine, in this UCSF Chimera (www.cgl.ucsf.edu/chimera/): Fe (brown),
∗
These authors contributed equally to this work. case to generate dehydroglycine (DHG) as an S (yellow), C (gray), O (red), N (blue), and atom X
†Corresponding author. E-mail: [email protected] intermediate in anaerobic thiamine biosynthesis (magenta). H is not shown for simplicity.

472 25 OCTOBER 2013 VOL 342 SCIENCE www.sciencemag.org