Discovery of 23 Natural Tubulysins from Angiococcus disciformis An d48 and Cystobacter SBCb004

Article abstract of DOI:10.1016/j.chembiol.2010.01.016

The tubulysins are a family of complex peptides with promising cytotoxic activity against multi-drug-resistant tumors. To date, ten tubulysins have been described from the myxobacterial strains Angiococcus disciformis An d48 and Archangium gephyra Ar 315. We report here a third producing strain, Cystobacter sp. SBCb004. Comparison of the tubulysin biosynthetic gene clusters in SBCb004 and An d48 reveals a conserved architecture, allowing the assignment of cluster boundaries. A SBCb004 strain containing a mutant in the putative cyclodeaminase gene tubZ accumulates pretubulysin A, the proposed first enzyme-free intermediate in the pathway, whose structure we confirm by NMR. We further show, using a combination of feeding studies and structure elucidation by NMR and high-resolution tandem mass spectrometry, that SBCb004 and An d48 together biosynthesize 22 additional tubulysin derivatives. These data reveal the inherently diversity-oriented nature of the tubulysin biosynthetic pathway.

Full text of DOI:10.1016/j.chembiol.2010.01.016

Chemistry & Biology
Article
Discovery of 23 Natural Tubulysins
from Angiococcus disciformis An d48
and Cystobacter SBCb004
Yi Chai,1,4 Dominik Pistorius,1,4 Angelika Ullrich,2 Kira J. Weissman,1 Uli Kazmaier,2 and Rolf Mu¨ ller1,3,
1Department of Pharmaceutical Biotechnology, Saarland University, P.O. Box 151150, 66041 Saarbru¨ cken, Germany
2Institute for Organic Chemistry, Saarland University, 66123 Saarbru¨ cken, Germany
*
3Helmholtz Institute for Pharmaceutical Research Saarland (HIPS), Helmholtz Center for Infection Research (HZI), Saarland University,
Campus C2 3, 66123 Saarbru¨ cken, Germany
4These authors contributed equally to this work
*Correspondence: rom@mx.uni-saarland.de
DOI 10.1016/j.chembiol.2010.01.016
SUMMARY
gephyra Ar 315, using bioactivity-guided screening (Sasse
et al., 2000; Steinmetz et al., 2004). The shared molecular core
The tubulysins are a family of complex peptides with comprises five amino acids (N-methyl pipecolic acid [Mep],
isoleucine [Ile], valine, cysteine, and phenylalanine or tyrosine),
and two units of acetate (Figure 1D). In the synthetic literature,
promising cytotoxic activity against multi-drug-
resistant tumors. To date, ten tubulysins have been
valine, cysteine, and one unit of acetate have been grouped
together into a structural unit referred to as tubuvaline (Tuv),
whereas the acetate chain-extended forms of phenylalanine
described from the myxobacterial strains Angiococ-
cus disciformis An d48 and Archangium gephyra Ar
315. We report here a third producing strain, Cysto-
bacter sp. SBCb004. Comparison of the tubulysin
or tyrosine are designated as tubuphenylalaline (Tup) and tubu-
tyrosine (Tut), respectively. All of the structures include an ace-
biosynthetic gene clusters in SBCb004 and An d48
reveals a conserved architecture, allowing the
assignment of cluster boundaries. A SBCb004 strain
containing a mutant in the putative cyclodeaminase
toxy moiety, but exhibit variable functionality (R2) within the
Tuv bis-acyl N,O-acetal substituent (Figure 1D). SAR studies
using synthetic analogs of the most potent tubulysin, tubulysin
D (1), have begun to define the structural features underlying
gene tubZ accumulates pretubulysin A, the proposed the compound’s cytotoxicity (Patterson et al., 2007; Balasubra-
manian et al., 2008; Raghavan et al., 2008; Wang et al., 2007; Pat-
terson et al., 2008). Taken together, these data reveal a surpris-
first enzyme-free intermediate in the pathway, whose
structure we confirm by NMR. We further show, using
a combination of feeding studies and structure eluci-
dation by NMR and high-resolution tandem mass
spectrometry, that SBCb004 and An d48 together
biosynthesize 22 additional tubulysin derivatives.
These data reveal the inherently diversity-oriented
nature of the tubulysin biosynthetic pathway.
ing tolerance to structural modification, encouraging efforts to
synthesize simplified variants of the tubulysins for evaluation as
anticancer agents. For example, replacing the chemically labile
N,O-acetal and stereogenic acetate groups of tubuvaline with
stable alternatives, resulted in only a minor loss of potency
(Raghavan et al., 2008). As a complement to chemical synthesis,
we have aimed to enable the genetic engineering of additional
tubulysin analogs by detailed investigation of the underlying
biosynthetic machinery (Sandmann et al., 2004; Wenzel and
INTRODUCTION
Mu¨ ller, 2005; Bode and Mu¨ ller, 2006; Wenzel and Mu¨ ller, 2007).
Angiococcus disciformis An d48 is known to produce four
The tubulysins are highly cytotoxic compounds, whose growth tubulysins, tubulysins D (1), E (2), F (3), and H (4) (Steinmetz
inhibition potential exceeds that of the anticancer drugs epothi- et al., 2004). Sequencing of the ca. 40 kb tubulysin gene cluster
lone, vinblastine, and taxol, by 20–1000 fold (Steinmetz et al., within the strain revealed that tubulysin assembly is catalyzed
2004). The metabolites are active against a wide variety of by a molecular ‘‘assembly line’’ consisting of polyketide syn-
cancer cell lines (ovarian, breast, prostate, colon, lung, and thase (PKS) and nonribosomal peptide synthetase (NRPS)
leukemia) within the NIC-60 cell line panel, including multi- multienzymes, and a hybrid PKS-NRPS subunit (Figure 1A)
¨
drug-resistant tumors (Domling and Richter, 2005). Mode-of- (Sandmann et al., 2004). Although the gene set does not incorpo-
action studies have demonstrated that the tubulysins disrupt rate any obvious functions for carrying out the required postas-
microtubule assembly (Khalil et al., 2006), and also show antian- sembly line oxidation and acyl transfer reactions, we identified
giogenic effects (Kaur et al., 2006). Together, these properties a gene encoding a lysine cyclodeaminase (tubZ), which was
make the tubulysins attractive lead structures for development assumed to participate in provision of the starter unit pipecolic
as antineoplastic agents.
acid (Khaw et al., 1998; Gatto et al., 2006). To directly probe
¨
The nine-membered tubulysin family was discovered by Hofle, the function of tubZ in An d48, we inactivated the gene by inser-
Reichenbach, and coworkers from several myxobacterial strains tional mutagenesis (Ullrich et al., 2009). The resulting mutant,
including Angiococcus disciformis An d48, and Archangium An d48-tubZꢀ, was expected to be deficient in tubulysin
296 Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved

Chemistry & Biology
Discovery of 23 Tubulysins
Figure 1. Comparison of Tubulysin Biosynthetic Gene Clusters in An d48 and SBCb004
(A) Genetic organization of the tubulysin gene cluster in Angiococcus disciformis An d48. Based upon the newly determined cluster boundaries, we have
redesignated gene tubG (Sandmann et al., 2004) as orf25, reflecting the fact that it is not part of the tubulysin pathway.
(B) Genetic organization of the tubulysin gene cluster in Cystobacter sp. SBCb004.
(C) Genetic organization of the tubulysin gene cluster fragment in Stigmatella aurantiaca DW 4/3-1.
(D) Model for tubulysin biosynthesis, showing the domain organization of the PKS-NRPS assembly line and the structures of the chain extension intermediates.
The structures of the ten known tubulysins (1–10) are shown, as well as that of the newly discovered compound, pretubulysin A (11).
biosynthesis, but instead generated tubulysin D at approxi- its intermediacy in the biosynthesis. Although yields of pretubu-
mately 3% of wild-type levels. The continued availability of pipe- lysin D were insufficient to enable unambiguous confirmation of
colic acid to the pathway was attributed to the presence of addi- its structure by nuclear magnetic resonance (NMR), the com-
tional, unidentified lysine cyclodeaminase function(s) in A. pound was shown to retain the high tubulin-degrading activity
disciformis (Ullrich et al., 2009). In addition, analysis by high- of its more complex tubulysin relatives (Ullrich et al., 2009).
resolution tandem mass spectrometry and comparison with
To obtain further insights into tubulysin biosynthesis via
authentic, synthetic standard, showed that An d48-tubZꢀ ap- comparative cluster analysis, we aimed to identify the gene
peared to accumulate the presumed first enzyme-free interme- cluster in a second myxobacterial strain, Cystobacter sp.
diate in the tubulysin pathway, pretubulysin D (5). The same SBCb004 (Figure 1B). In common with Archangium gephyra
metabolite was also detected in the wild-type strain, supporting Ar315, Cystobacter SBCb004 generates tubulysins A (6), B (7),
Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved 297

Chemistry & Biology
Discovery of 23 Tubulysins
Table 1. Proteins Encoded in the Sequenced Region in the SBCb004 Tubulysin Biosynthetic Gene Cluster and their Putative Functions
NRPS/PKS Portion of the Gene Cluster
Protein
aa
Proposed Function
(protein domains with their position in the sequence)a
TubB
TubC
1520
2628
NRPS domains: C (62ꢀ493), A (494ꢀ1416), N-MT (971ꢀ1368), PCP (1432ꢀ1500)
NRPS domains: C (77ꢀ515), A (516ꢀ1027), PCP (1044ꢀ1112), C (1136ꢀ1569), A (1570ꢀ2499), N-MT (2042ꢀ2451),
PCP (2514ꢀ2582)
TubD
3511
PKS domains: KS (11ꢀ441), AT (527ꢀ852), KR (939ꢀ1370), DH (1371ꢀ1652), ER (1678ꢀ1980), ACP (2062ꢀ2157)
NRPS domains: HC (2190ꢀ2624), A (2625ꢀ3138), PCP (3153ꢀ3223), Ox (3231ꢀ3509)
TubE
TubF
1161
2843
NRPS domains: C (47ꢀ482), A (483ꢀ1034), PCP (1051ꢀ1121)
PKS domains: KS (5ꢀ436), AT (522ꢀ847), KR (925ꢀ1721), C-MT (1309ꢀ1540), DH (1722ꢀ1997), ER (2023ꢀ2322),
ACP (2404ꢀ2506), TE (2541ꢀ2815)
ORFs Encoded Upstream and Downstream of tubB–tubF
Protein
aa
Proposed Function of the
Homologous Protein
Source of the Homologous Protein
Similarity/
Identity
GenBank Accession
Number
OrfC
735
Outer membrane protein, OMP85
family, putative
Stigmatella aurantiaca
71%/57%
EAU64371.1
OrfB
OrfA
TubA
Orf2
1549
273
432
219
386
394
337
368
124
119
271
Hypothetical protein MXAN_6465
a/b hydrolase fold
Myxococcus xanthus
Geobacter uraniireducens
Stigmatella aurantiaca
Stigmatella aurantiaca
Stigmatella aurantiaca
Angiococcus disciformis
Angiococcus disciformis
Angiococcus disciformis
Pennisetum glaucum
Stigmatella aurantiaca
Stigmatella aurantiaca
65%/48%
91%/81%
89%/81%
86%/74%
91%/86%
90%/78%
77%/62%
82%/68%
50%/33%
93%/89%
49%/33%
ABF91418.1
ABQ27606.1
EAU66910.1
EAU66945.1
EAU66930.1
CAF05646.1
CAF05653.1
CAF05653.1
ABP65327.1
EAU64035.1
EAU63389.1
TubA
Conserved hypothetical protein
TubZ
TubZ
Orf1
Hypothetical protein (Orf1)
Patatin-like protein (Orf18)
Patatin-like protein (Orf18)
Chloroplast heat shock protein 70
ChpK
Orf17
Orf18
OrfD
OrfE
OrfF
Protein kinase
For the corresponding proteins in A. disciformis An d48 gene cluster, please see (Sandmann et al., 2004).
a Domain boundaries were assigned, when possible, based on the solved structures of the individual domains.
C (8), G (9), and I (10) (Figure 1D), incorporating tubutyrosine Genes within an 82 kb region were then functionally annotated.
instead of the tubuphenylalanine of tubulysins D, E, F and H. By identification of the gene set common to both SBCb004
The SBCb004 tubulysin gene cluster closely resembles that in and An d48, we were able for the first time to define the cluster
An d48 (compare Figures 1A and 1B), allowing straightforward boundaries in both organisms. The clusters each incorporate
determination of the cluster boundaries in both myxobacteria. 11 genes, which are present in the same order and orientation
Like its An d48 counterpart, no genes encoding post-PKS (Figures 1A and 1B and Table 1). The core biosynthetic genes
tailoring functions were present in the SBCb004 tubulysin (tubB–tubF) encode three NRPS subunits, one PKS multien-
cluster. As anticipated, a tubZ disruption mutant in SBCb004 zyme, and a hybrid PKS-NRPS protein. Two conserved open
accumulated the tubutyrosine analog of pretubulysin D, pretubu- reading frames (orfs) (orf17 and orf18) are present downstream
lysin A (11) (Figure 1D). In this case, yields were sufficient to allow of gene tubF, whereas tubB is preceded by genes tubA, orf2,
verification of the structure for the first time by NMR. In addition, tubZ, and orf1. Gene orf1 is predicted to encode an anion-
on the basis of NMR, feeding experiments and high-resolution transporting ATPase, whereas the orf2 gene product exhibits
mass spectrometry (MS) analysis, we report the existence of similarity to hypothetical proteins. Mutual sequence homology
22 additional novel tubulysin derivatives from both SBCb004 between the SBCb004 and An d48 proteins is high (Orf1: 78%
and An d48.
identity/90% similarity; Orf2: 69% identity/82% similarity versus
68%–73% identity/78%–81% similarity for the PKS/NRPS
proteins), which suggests evolutionary pressure to maintain their
functions. However, neither Orf1 nor Orf2 has an obvious role to
play in tubulysin biosynthesis.
RESULTS AND DISCUSSION
Discovery and Annotation of the Tubulysin Biosynthetic
Gene Cluster in Cystobacter SBCb004
The product of gene tubA shows similarity to NRPS C domains
To locate the tubulysin gene cluster in Cystobacter SBCb004, we (Sandmann et al., 2004). Notably, a sequence (DHxxxDx) within
shot-gun sequenced the genome to 276 contigs using 454 core motif C3 is closely similar to a proposed signature sequence
technology (Margulies et al., 2005; Rothberg and Leamon, (HHxxxDG) for enzymes that effect acyl transfer, such as
2008), and then analyzed the data by BLAST (Altschul et al., chloramphenicol acyltransferase and dihyrolipoamide acyltrans-
1997) using tub genes from An d48 (Sandmann et al., 2004). ferase (De Crecy-Lagard et al., 1995). Thus, TubA may serve
298 Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved

Chemistry & Biology
Discovery of 23 Tubulysins
Figure 2. Characterization of TubZ Cyclo-
deaminase Activity
(A) SDS-PAGE analysis of GST-TubZ (lane 1)
and discrete TubZ (lane 3). Lane
2 contains
molecular weight markers, with the indicated
masses. The calculated molecular weight of
GST-TubZ is 67.698 kDa, whereas that of TubZ
is 40.501 kDa.
(B) Chiral HPLC-MS analysis (extracted ion chro-
matogram [EIC], m/z [M+H]+ = 352.2) of a 1:1
mixture of Fmoc-derivatized, commercially avail-
able L- and D-pipecolic acid. Peak identities
were assigned by comparison to analysis of
Fmoc-derivatized D-pipecolic acid alone. Inset
is the mass spectrum of Fmoc-derivatized L-pipe-
colic acid (calculated m/z [M+H]+ = 352.16).
(C) Chiral HPLC-MS (EIC) analysis of the Fmoc-
derivatized commercially available D-pipecolic
acid alone.
(D) Chiral HPLC-MS (EIC) analysis of the TubZ reaction carried out in the presence of D-lysine.
(E) Chiral HPLC-MS (EIC) analysis of the TubZ reaction carried out in the presence of L-lysine.
(F) Mixture of the assay shown in (E) and Fmoc-derivatized D-pipecolic acid (C).
(G) Chiral HPLC-MS (EIC) analysis of a representative negative control reaction carried out with L-lysine in the absence of TubZ.
as one of the missing acyl transferases in the tubulysin pathway.
Analysis by BLAST also revealed a portion of the tubulysin
In support of a shared role for the TubA proteins in the two gene cluster comprising tubA, orf2, tubZ, and fragments of
strains, the SBCb004 and An d48 homologs show a high orf1 and tubF, in the publically available genome of a third
degree of sequence conservation (56% identity/73% similarity). myxobacterial strain, Stigmatella aurantiaca DW 4/3-1 (Fig-
However, the proposed functions remain to be directly demon- ure 1C). This architecture may have arisen from deletion of the
strated. orf17 and orf18 both encode patatin-like proteins, which intervening genes (tubB–tubE) from an ancestral cluster. Inter-
show high mutual sequence homology (60% identity/70% estingly, the Cystobacter TubA, Orf2, and TubZ proteins show
similarity in the case of the SBCb004 proteins), suggesting higher sequence identity to their Stigmatella homologs than to
evolution by gene duplication. Interestingly, the closest homolog those of Angiococcus (Table 1), despite the closer evolutionary
to both Orf17 and Orf18 in SBCb004 is Orf18 of An d48, which relationship between Cystobacter and Angiococcus (as judged
may imply that orf17 of An d48 has diverged from the other on the basis of 16S rDNA) (Figure S1). Consistent with the fact
sequences. Patatin-like proteins exhibit lipid acyl hydrolase that only part of the gene cluster is present in S. aurantiaca,
activity (Moraleda-Munoz and Shimkets, 2007; Banerji and the strain is not known to produce any of the tubulysins.
Flieger, 2004), a function apparently not required in the tubulysin
pathway. Intriguingly, however, genes encoding patatin-like Investigation of the Stereochemistry of the Pipecolic
proteins are also present in the DKxanthene gene clusters of Acid Starter Unit
the myxobacteria Myxococcus xanthus DK1622 and Stigmatella The final configuration of the N-methyl pipecolic acid in the
aurantiaca DW4/3-1 (Meiser et al., 2008), suggesting that the tubulysins is D (Steinmetz et al., 2004), and therefore we hypoth-
enzymes may play a previously unsuspected role in natural esized that free D-pipecolic acid (either generated directly by the
product biosynthesis.
cyclodeaminase TubZ, or by epimerization of an initially-formed
Predictably from the defined boundaries, orfs upstream of L-isomer), might be selected as a substrate by the tubulysin
tubA and downstream of orf18 in both organisms have no PKS-NRPS. Interestingly, the lysine cyclodeaminases RapL
obvious function in the tubulysin pathway. Thus, the oxidase from the rapamycin pathway (40% identity/53% similarity to
activities, and possibly a second acyl transferase enzyme in TubZ) and Pip from friulimicin biosynthesis (45% identity/56%
addition to TubA, are apparently lacking in both clusters, and similarity to TubZ) have been shown to generate exclusively
so are presumed to be located elsewhere in the genome. Such L-pipecolic acid from L-lysine (Khaw et al., 1998; Gatto et al.,
split cluster PKS/NRPS architecture is unusual for Strepto- 2006; Mu¨ ller et al., 2007). To address the stereospecificity of
myces, but has precedent in the myxobacteria (Carvalho et al., TubZ, we expressed the An d48 enzyme in recombinant form
2005; Kopp et al., 2005; Perlova et al., 2006). The fact that this as a C-terminal translational fusion with glutathione-S-trans-
split organization is conserved in two different species argues ferase (GST) (see Experimental Procedures), and confirmed the
for the presence of the tubulysin genes in an ancestor common identity of the protein by matrix assisted laser desorption ioniza-
to both Cystobacter SBCb004 and Angiococcus An d48—that is, tion-time-of-flight mass spectrometry (MALDI-TOF-MS) anal-
that the cluster evolved into its present form prior to the specia- ysis. TubZ was purified as its GST fusion, and also as a discrete
tion event within the Cystobacterineae suborder. Phylogenetic protein by on-column cleavage of GST using PreScission
analysis based on 16S rDNA supports this hypothesis, because Proteaseꢀ (Figure 2A). The recombinant TubZ proteins (1.5 mM
An d48 and SBCb004 belong to the same family, sharing GST-TubZ or 0.6 mM TubZ) were then incubated with either
membership with another potential tubulysin producing-strain L- or D-lysine (1 mM), NAD+ (100 mM), and BSA (14 mM) at
Archangium gephyra Ar g1 (see Figure S1 available online).
30ꢁC for 24 hr (Gatto et al., 2006). The reactions were quenched
Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved 299

Chemistry & Biology
Discovery of 23 Tubulysins
Figure 3. Production of Tubulysin A and
Pretubulysin A by SBCb004 Wild-Type and
Mutant SBCb004-tubZꢀ
(A) HPLC-MS analysis (base peak chromatogram
[BPC]) of wild-type SBCb004. Peaks correspond-
ing to tubulysin A (6) and pretubulysin A (11) are
indicated.
(B) HPLC-MS analysis (BPC) of extracts of mutant
SBCb004-tubZꢀ. Peaks corresponding to tubuly-
sin A (6) and pretubulysin A (11) are indicated.
(C) [1H, 1H]-COSY (bold bonds) and HMBC
(arrows) couplings determined for pretubulysin A
(11). Only a subset of HMBC couplings have
been included within the tyrosine aromatic ring.
incorporated in either case, strongly
suggesting that epimerization occurs
following attachment of the amino acid
to the multienzyme. However, module
TubB (Figure 1) lacks a classical epimeri-
zation (E) domain (Luo et al., 2002; Sta-
chelhaus and Walsh, 2000) to generate
the required stereochemistry. Further
support for the lack of epimerization by
TubB comes from phylogenetic analysis
of the TubC condensation domain that
accepts the N-methyl pipecolic acid as
substrate: the C domain clusters together
with other C domains with demonstrated
specificity for L-amino acids at both their
by the addition of acetonitrile. Prior to analysis by high-perfor- donor and acceptor sites (so-called LCL domains) (Rausch et al.,
mance liquid chromatography-mass spectrometry (HPLC-MS), 2007) (Figure S3). This analysis also argues against the possi-
the assay mixtures were derivatized with 9-fluorenylmethyl bility that the TubC C domain functions as a condensase/
chloroformate (Fmoc-Cl). Crucially, we were able to separate epimerase (Balibar et al., 2005), a dual-function domain that
Fmoc-derivatized, commercially available L- and D-pipecolic has been shown in other NRPS pathways to convert L-amino
acid from each other by chiral chromatography (Figures 2B acids to their D-isomers. Thus, epimerization apparently occurs
and C). Assays with the lysine substrates and GST-TubZ or downstream of the first round of chain extension and is likely to
discrete TubZ gave identical results. In each case, analysis of be catalyzed by an external racemase (Dubern et al., 2008).
the assay containing D-lysine did not reveal any evidence for Intriguingly, only the PCP domain of TubD incorporates a
formation of the product pipecolic acid (Figure 2D), relative to sequence motif at the phosphopantetheine attachment site
controls performed with L-/D-lysine in the absence of TubZ (GXDSX), which resembles the GGDSI motif previously impli-
(Figure 2G). In contrast, pipecolic acid was detected in the cated in recognition between E and PCP domains in cis (Linne
assays with L-lysine (Figure 2E), and was shown to be exclu- et al., 2001), perhaps indicating that the epimerization in trans
sively the L-isomer by spiking the assay mixture with authentic, may occur at this stage of chain extension. However, further
derivatized D-pipecolic acid (Figure 2F). Thus, TubZ shows the studies will be required to elucidate the precise timing of this
same substrate and product stereospecificities as RapL and Pip. cryptic reaction.
At this stage, it remained a possibility that the L-pipecolic acid
generated by TubZ was epimerized to the D-isomer before Generation of a tubZ Mutant of SBCb004 to Enable
activation by TubB by an as yet unidentified epimerase. To probe Structure Elucidation of Pretubulysin A
this question directly, we synthesized both enantiomers of pipe- Although analysis by mass spectrometry of pretubulysin D (5)
colic acid in 4,5-di-deuterium-labeled form via ring closing from An d48 strongly supported its proposed structure (Ullrich
metathesis (Grubbs et al., 1995; Schuster and Blechert, 1997) et al., 2009), we aimed to obtain sufficient quantities of its pretu-
of an N-allylated g,d-unsaturated amino acid, followed by cata- bulysin A analog to allow structural proof by NMR. A compound
lytic deuteration (Schneider and Kazmaier, 1998; Kazmaier and with a mass (m/z 686.4 [M+H]+) consistent with pretubulysin A
Schneider, 1998; Zumpe and Kazmaier, 1998, 1999; Miller (11) was identified in extracts of the Cystobacter sp. SBCb004
et al., 1998). The enantiomerically pure materials were then wild-type strain, at 42% yield relative to tubulysin A (6) (Fig-
administered to cultures of both wild-type An d48 and the An ure 3A). Because we had previously observed that the An
d48-tubZꢀ mutant (Ullrich et al., 2009). Analysis by HPLC-MS d48-tubZꢀ mutant produced increased amounts of pretubulysin
(Figure S2) showed that only the L-isomer of pipecolate was D relative to wild-type A. disciformis (Ullrich et al., 2009), we
300 Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved

Chemistry & Biology
Discovery of 23 Tubulysins
hoped to boost the titer of pretubulysin A in SBCb004 by engi- was barely visible in the 13C NMR spectrum, and therefore
neering the equivalent tubZ mutation (Figure S4). HPLC-MS the 6.5 mg purified pretubulysin A was insufficient to assign
analysis of the resulting mutant, SBCb004-tubZꢀ, revealed that this peak. We also confirmed the correct sequence of building
the strain accumulated pretubulysin A at approximately 5-fold blocks from the [1H, 13C]-HMBC connectivities (Figure 3C). The
wild-type levels, whereas the production of tubulysin A was only connection which could not be established in this way
decreased by 20-fold (Figure 3B). The availability of the genome was that between tubuvaline and tubutyrosine, due to the
sequence of SBCb004 allowed us to probe for the presence of absence of a signal for the C6 carbonyl (this connectivity was
additional lysine cyclodeaminases in the strain, which could also missing in the HMBC of synthetic pretubulysin D). Nonethe-
complement the inactivated TubZ. BLAST analysis revealed less, taken together, the combined data from the 1D and 2D
two genes with homology to ornithine cyclodeaminases (20% NMR experiments strongly support our proposed structure for
and 25% sequence identity to TubZ). Because RapL exhibits pretubulysin A.
ornithine cyclodeaminase activity as a side reaction (Gatto
et al., 2006), it seems reasonable that the detected ornithine cy- Analysis of the Two Strains for Novel Metabolites
clodeaminases may catalyze low levels of lysine cyclodeamina- The presence of the pretubulysins in the wild-type strains
tion, explaining the continued presence of pipecolic acid in the prompted us to reanalyze the extracts for additional metabolites,
SBCb004-tubZꢀ mutant. Given the evolutionary relationship which may have escaped detection in earlier experiments. In
between Cystobacter SBCb004 and Angiococcus disciformis fact, we found 11 further candidates for novel compounds in
An d48 (Figure S1), it is likely that An d48 harbors homologs of A. disciformis An d48, with molecular masses m/z [M+H]+
=
628.4, 656.4, 670.4 (different retention time than that of pretubu-
these two genes.
lysin D), 672.4, 686.4 (3 2), 700.4, 714.4, 744.4, 772.4, and 786.4
(Figure 4). Although most of these metabolites appeared to be
minor in comparison with tubulysin D (see Figure 4 for yield
Purification of Pretubulysin A and NMR Structure
Elucidation
To generate enough pretubulysin A to enable purification, estimates), compound 670.4 was present in significant amounts
SBCb004-tubZꢀ was grown on the 26 L scale in the presence (as judged by relative peak area in the HPLC-MS chromato-
of XAD adsorber resin. Pretubulysin A was then purified from gram). In each case, a likely relationship to the tubulysins was
a methanolic resin extract by silica gel chromatography, fol- established by feeding deuterium-labeled L-pipecolic acid,
lowed by preparative HPLC, yielding 6.5 mg material. High-reso- valine, or both and by comparison of the incorporation pattern
lution electrospray ionization mass spectrometry (HR-ESI-MS) to that of pretubulysin D (5) (Figure S2). Comparison of the frag-
revealed a monoisotopic mass (m/z [M+H]+ = 686.3941 Da) mentation patterns of metabolites 772.4 and 786.4 with those of
corresponding, as predicted, to
a molecular formula of the tubulysins D–F and H (Figure S6) allowed us to assign the
C36H56N5O6S. Structure elucidation was accomplished using structures to desacetyl tubulysin E (12) and desacetyl tubulysin
combined one- and two-dimensional (1D and 2D) NMR D (13), respectively (Figure 4). Although biosynthetic consider-
experiments (1H, 13C, [1H, 1H]-COSY, [1H, 13C]-HSQC, and ations (i.e., the defined stereospecificities of adenylation (A)
[1H, 13C]-HMBC) (Figure S5). As the interpretation of the NMR and C-methyltransferase (MT) domains) strongly suggest that
spectra of the tubulysins is complicated by missing and overlap- the stereochemistry of these metabolites should be shared
ping signals (Steinmetz et al., 2004), reference spectra of with the parent compounds, we cannot make any definitive
synthetic pretubulysin D (5) (Ullrich et al., 2009) were recorded statements on this issue in the absence of NMR data.
in the same solvent and on the same instrument, in order to allow
for direct comparison of the data (Table 2).
We next obtained compound 670.4 at high purity (1.3 mg), and
determined its accurate mass. The most probable molecular
This approach enabled the assignment of the majority of formula, C35H51N5O6S, differs from pretubulysin D by –CH4
pretubulysin A signals, which were in good agreement with those and +O, demonstrating that the new compound is not simply
observed from synthetic pretubulysin D. The exceptions (19-H, a stereoisomer of pretubulysin D. We hypothesized that the
23-H and 25-H) are likely due to the fact that pretubulysin A metabolite might have lost a methyl group relative to pretubuly-
was only partially protonated at N24 under the purification condi- sin D (accounting for –CH2), and that it would incorporate an
tions, while pretubulysin D was analyzed as its trifluoroacetic extra oxygen (+O) and a site of unsaturation (–2H) (e.g.,
acid (TFA) salt, which likely resulted in a higher degree of proton- a carbonyl functionality). To confirm these expectations, cultures
ation at this position (Steinmetz et al., 2004). Pretubulysin A of wild-type A. disciformis were grown in the presence of
signals that could not be assigned with confidence from the 1H d3-methionine, in order to generate a pool of [methyl-d3]-S-ad-
NMR spectrum included most of the methylene groups, the enosylmethionine ([methyl-d3]-SAM) in the cell. Analysis of the
proton at C13, and the protons within the piperidine ring of pipe- resulting extracts showed that while tubulysin D incorporated
colic acid (Figure 1D). The exact chemical shifts for all of these three SAM-derived methyl groups (maximum mass shift of +8),
protons were therefore deduced from [1H,1H]-COSY data. The compound 670.4 contained only two (+6) (compare Figures 5A
COSY couplings were also used to verify the structures of the and 5B). To determine the location of the missing methyl group,
individual building blocks in the molecule, Mep, Ile, Tuv, and we elucidated the fragmentation pattern of 670.4 by comparison
Tut (Figure 1D). All carbon signals with the exception of the C6 to that of pretubulysin D (Figure S7). This analysis revealed
carbonyl were assigned for the biosynthetic compound by that deuterium labeling occurred exclusively on the N-methyl
[1H, 13C]-HSQC, and [1H, 13C]-HMBC, and showed good agree- of pipecolic acid, and/or the methyl of tubuphenylalanine (Fig-
ment with the corresponding data for the synthetic material. ure 5C). We therefore concluded that compound 670.4 lacks
Even with 24 mg synthetic pretubulysin D, the C6 carbonyl signal the N-methyl group of valine. Further support for this hypothesis
Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved 301

Chemistry & Biology
Discovery of 23 Tubulysins
Table 2. Comparison of 1H NMR (500 MHz, CD3OD), 13C NMR (125 MHz, CD3OD) and 2D NMR (HSQC, HMBC) Spectroscopic
Assignments of Synthetic Pretubulysin D and Biosynthetic Pretubulysin A
Synthetic Pretubulysin Da
Isolated Pretubulysin A
Position
(C)
Position
(H)
dC (ppm)
dH (ppm)
HMBC
(1H to 13C)
dC (ppm)
dH (ppm)
HMBC
(1H /13C)
(m, J [Hz])
(m, J [Hz])
1
2
3
180.2
37.4
39.0
180.3
38.0
39.1
2-H
2.55 (m)
1.71 (m)
2.01 (m)
4.36 (m)
2.55 (m)
1.68 (m)
2.01 (m)
4.30 (m)
3a-H
3b-H
4-H
4
50.4
50.8
-
6
163.2
150.7
123.7
172.0
30.8
7
150.5
123.8
171.7
30.8
8
8-H
7.96 (s)
7, 10
7.96 (s)
7, 10
10
11
11a-H
11b-H
12a-H
12b-H
13-H
2.86 (m)
2.95 (m)
1.99 (m)
2.18 (m)
4.37 (m)
2.85 (m)
2.95 (m)
1.99 (m)
2.17 (m)
4.33 (m)
12
12
12
30.0
30.0
13
15
16
18
19
20
60.3
174.7
55.7
169.4
67.8
29.9
60.4
174.5
55.6
170.2
68.3
30.1
16-H
4.70 (d, 8.3)
15, 18, 40, 41
4.73 (d, 8.4)
15, 18
19-H
3.77 (dd, 12.0, 2.8)
1.80 (m)
3.58 (m)
1.78 (m)
2.06 (m)
1.54 (m)
1.76 (m)
1.85 (m)
2.85 (m)
3.40 (brs)
2.63 (s)
20a-H
20b-H
21-H
2.15 (m)
21
22
22.3
23.6
1.60 (m)
22.4
24.2
22a-H
22b-H
23a-H
23b-H
25-H3
26-H3
27-H
1.78 (m)
1.92 (m)
23
56.0
3.07 (brs)
3.49 (d, 12.3 Hz)
2.74 (s)
56.2
25
26
27
28
29
30
31
32
33
36
37
38
39
40
41
42.6
19, 23
43.2
19, 23
1, 2, 3
3, 4
18.2
1.16 (d, 7.2)
2.89 (m)
1, 2, 3
18.3
1.16 (d, 7.2)
2.81 (m)
42.0
3, 4, 28, 29, 33
41.2
139.7
130.6
129.5
127.6
129.5
130.6
31.2
130.3
131.3
116.2
156.8
116.2
131.3
31.2
29-H
30-H
7.22 (m)
27, 28
27, 28
30, 32
27, 28
27, 28
7.03 (d, 8.3)
6.68 (m)
27,30, 31, 32, 33
28, 31, 32
7.22 (m)
7.15 (m)
32-H
7.22 (m)
6.68 (m)
28, 30, 31
33-H
7.22 (m)
7.03 (d, 8.3)
1.80 (m)
27, 29, 30, 31, 32
36-H
1.80 (m)
37-H3
38-H3
39-H3
40-H
20.0
0.80 (d, 6.5)
0.98 (d, 6.5)
3.10 (s)
13, 36, 38
13, 36, 38
13, 15
20.1
0.79 (d, 6.5)
0.97 (d, 6.5)
3.07 (s)
13, 36, 38
13, 36, 37
13, 15
20.3
20.2
30.1
30.4
37.2
1.92 (m)
37.6
1.92 (m)
41a-H
41b-H
42-H3
43-H3
25.3
1.24 (m)
25.3
1.23 (m)
1.61 (m)
1.61 (m)
42
43
11.0
15.8
0.92 (t, 7.4)
40, 41
11.1
15.9
0.90 (t, 7.4)
1.00 (d, 6.8)
41
1.02 (d, 6.8)
16, 40, 41
16, 41
a Pretubulysin D was analyzed as its trifluoroacetic acid (TFA) salt.
was obtained by NMR analysis. The 1H NMR spectra of pretubly- with the C39 N-methyl signal (Figure 1) appearing downfield of
sins D (5) and A (11) (Figure S5) exhibit two clear singlet signals the C25 methyl (d 3.07 and 2.63 ppm respectively, for pretubuly-
corresponding to the N-methyl groups (Table 2 and Figure 5D), sin A). Inspection of the spectrum of metabolite 14 (Figures 5E
302 Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved

Chemistry & Biology
Discovery of 23 Tubulysins
Figure 4. Novel Tubulysins Produced by A. disciformis An d48
(A) HPLC-MS analysis (BPC) of wild-type extracts. Peaks corresponding to the five known (1–5) and eleven novel tubulysins (12–22) are indicated.
(B) Proposed structures of compounds 12–22.
and S5) reveals a single N-methyl signal, which on the basis of its b-carbonyl of the chain extension intermediate to a methylene,
chemical shift (d 2.35 ppm), can be assigned to the C25 methyl. and the respective carbons, C3 and C12. Based on the appear-
To account for the introduction of a carbonyl group, we consid- ance of new fragment peaks at mass 361.1 and 343.1 Da, corre-
ered a biosynthetic mechanism in which an oxygen atom which sponding to cleavage adjacent to C12 (Figure S6), we identified
is normally lost could be retained in the structure. This focused this carbon as the most probable site of modification. Thus,
attention on the two PKS modules which normally process the 670.4 is proposed to be N-desmethyl 12-keto pretubulysin D (14).
Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved 303

Chemistry & Biology
Discovery of 23 Tubulysins
Figure 5. Identification of N-Desmethyl 12-keto Pretubulysin (14)
(A) High-resolution MS analysis of 14 from cultures supplemented with d3-methionine. Labeling occurred at either or both of the C25 and C26 methyl groups,
generating mass shifts of +3 and +6, respectively.
(B) High-resolution MS analysis of tubulysin D (1) from cultures supplemented with d3-methionine. Labeling occurred at either or both of the C25 and C26 methyl
groups and/or at the C39 methylene, giving rise to the characteristic +2, +3, +5, +6, and +8 labeling pattern.
304 Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved

Chemistry & Biology
Discovery of 23 Tubulysins
With this structure in hand, it was possible to assign chemistry) that can result from the tubulysin pathway. Impor-
compounds 656.4 and 672.4, to N-desmethyl pretubulysin tantly, in every case, our proposed structures can be rationalized
analogs in which the C12 carbonyl group had alternately been in terms of reasonable variations in the action of a specific
reduced to a hydroxyl (672.4; N-desmethyl 12-hydroxy pretubu- domain or domains within the assembly lines.
lysin D (15)), or to a methylene functionality (656.4; N-desmethyl
pretubulysin D (16)) (Figure 4; Figures S6 and S7); in the case of
compound 15, however, we cannot rule out the alternative
that the hydroxyl group is instead located at the adjacent C11
position, consistent with post-assembly line hydroxylation of
metabolite 16. In any case, a compound corresponding to the
intermediate dehydrated analog was not detected. In addition,
one of the two compounds with mass m/z [M+H]+ = 686.4 (17)
appears to be the 11-hydroxy version of metabolite 14. By MS
analysis, it was also possible to putatively identify compound
714.4 as N-desmethyl 11-acetoxy pretubulysin D (18) (Figure 4;
Figures S6 and S7), referred to in earlier synthesis literature as
Conclusions
The identification of 23 novel tubulysins in A. disciformis and
Cystobacter reveals new aspects of the biosynthesis, and
supports earlier proposals. For example, our results strongly
suggest that pretubulysin D/A (5, 11) are the products of the
hybrid PKS-NRPS in each strain. More generally, our data
show that several steps in the pathway do not go to completion.
The presence in the extracts of pretubulysins, and compounds
12, 13, 19, 20 and 23, 24, 30, 31 missing one or more of the
oxidation and acylation reactions, demonstrates that post-
assembly line modification is relatively inefficient, even in the
wild-type strains. This observation is consistent with the
absence of obvious tailoring genes (oxidases and perhaps an
acyltransferase) within the clusters, arguing for a lack of coevo-
lution of this portion of the pathways. The N-methyltransferase
domain of subunit TubC also operates only imperfectly—in
fact, the N-desmethyl compounds (14–18) and (25–29) consti-
tute major metabolites of the strain. Apparently, the lack of
N-methylation somehow induces skipping of some, if not all of
the reductive steps carried out by the adjacent PKS module.
However, these activities do occasionally function, because
the corresponding hydroxyl (15, 26) and methylene (16, 27)
analogs are produced at low levels. Processing by post-
assembly line enzymes at C11 can also occur, and is more
efficient when full reduction has taken place at the adjacent
C12 position. This mechanism is inconsistent, however, with
the discovery in extracts of N-hydroxymethyl 12-keto pretubuly-
sins (21, 32), because the reductive steps have been bypassed
even though the N-methyl group was added. One possible
explanation is that, in a small number of cases, the N-methyl is
hydroxylated while the intermediate is attached to the assembly
line, and this modification also disrupts reductive processing by
the PKS module. Finally, the truncated analogs (22, 33) appear to
arise from failure of the final chain extension step with malonate,
catalyzed by TubF. At this stage it is not clear whether premature
release from the assembly line is spontaneous due to stalling of
the intermediate at the TubE/TubF interface, or alternatively
enzyme-catalyzed, for example by the TE domain, as demon-
strated in pikromycin biosynthesis (Kittendorf et al., 2007).
It is intriguing to consider whether the apparent ‘‘failure’’ of
various enzymes in the tubulysin pathway is in fact favored by
evolution to increase the diversity of structures generated by the
strain, a proposal known as the ‘‘screening hypothesis’’ (Firn and
Jones, 2003). Indeed, the production of sometimes large families
of metabolites that differ in their core structures appears to be a
general feature of myxobacterial secondary metabolism, in which
post-assembly line processing is relatively minimal (Weissman
and Mu¨ ller, 2008). However, relevant information on the biolog-
ical roles of the various tubulysins for the producing organisms
is lacking at present. Clearly, though, the PKS-NRPS can process
¨
tubulysin U (Domling et al., 2006; Sani et al., 2007; Balasubrama-
nian et al., 2009), demonstrating that at least two of the normal
post-assembly line modifications can occur even in the absence
of N-methylation. By similar logic, the second peak at 686.4 was
attributed to N-hydroxymethyl pretubulysin (19) and that at 744.4
to its 11-acetoxy analog (20) (Figure 4; Figures S6 and S7).
Analysis of the compound 700.4 revealed fragments consistent
with the presence of an N-hydroxymethyl functionality and a
12-keto group, leading to its identification as N-hydroxymethyl
12-keto pretubulysin D (21). Finally, metabolite 628.4 appears
to comprise a truncated pretubulysin derivative in which tubu-
phenylalanine has been replaced by phenylalanine (22) (Figure 4;
Figures S6 and S7). As further support for these assignments,
incubation in the presence of d3-methionine revealed the incor-
poration of two or three labeled methyl groups, as appropriate,
into metabolites 5, 14–16, 18, 20–22 (Figure S2). We next
analyzed extracts of wild-type SBCb004 for the corresponding
novel tubulysins, using the same analytical methodology. This
comparison revealed a tubutyrosine analog for each of the new
compounds (compounds 23–33) (Table S1 and Figure S8). As
with the tubulysin D series, the amounts of these compounds
relative to tubulysin A were uniformly low; however the specific
yield profiles for the various analogs differ between the strains.
Although the amounts of the new metabolites enabled
structure elucidation by mass spectrometry, the yields were
insufficient to allow full structure proof by NMR, or evaluation
of cytotoxicity relative to the parent compounds. Nonetheless,
the high-resolution MS analysis of the parent ions and all frag-
ments is fully consistent with the suggested structures, there is
strong internal concordance among the proposed fragmentation
patterns (Figure S6), and the structural assignments are in
complete agreement with the feeding studies. In further support
of our MS-based methodology, the structure hypothesized for
pretubulysin on the basis of the MS data (Ullrich et al., 2009)
was subsequently shown to be correct by NMR, and all available
NMR evidence suggests that the structure of N-desmethyl
12-keto pretubulysin D was also correctly elucidated by MS
analysis. In addition, the mode of biosynthesis significantly
restricts the number of possible structures (including stereo-
(C) Determination of the location of the methyl groups in 14, based upon analysis of deuterium-labeled fragments (compare to Figures S6 and S7).
(D) Portion of the 1H NMR spectrum of pretubulysin D (5). Signals arising from aliphatic and N-methyl groups are indicated.
(E) Portion of the 1H NMR spectrum of N-desmethyl 12-keto pretubulysin (14). Comparison with (D) shows that the metabolite is lacking the C39 N-methyl group.
Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved 305

Chemistry & Biology
Discovery of 23 Tubulysins
electroporation, as follows: SBCb004 wild-type cells were grown to an
OD600 0.8–1.0, and cells were harvested from 2 ml culture by centrifugation
at room temperature. The cell pellet was then washed twice with an equal
amount of washing buffer (5 mM HEPES [pH 7.2], 0.5 mM CaCl2), and resus-
pended in 50 ml ddH2O. Plasmid pTOPO-tubZ (1–3 mg) was then introduced
into the cell suspension, and the mixture was electroporated in a 0.1 cm
cuvette (200 U, 25 mF, and 0.9 kV/cm). M-medium (1 ml) was added to the
cells. The suspension was transferred into a 2 ml centrifuge tube and cultivated
for 2 hr at 30ꢁC and 800 rpm on a thermomixer. The cells were then mixed with
1 ml M-medium, and plated on M-agar (M-medium with 1.5% agar) containing
kanamycin (100 mg/ml). The plates were incubated at 30ꢁC for 1–2 weeks until
colonies became visible. Correct integration of plasmid pTOPO–tubZ into the
chromosome of SBCb004 was confirmed using three different PCR reactions
with primers pTOPO-In (50-CCTCTAGATGCATGCTCGAGC-30), pTOPO-Out
(50-TTGGTACCGAGCTCGGATCC-30), primer A (50-GTCGGCGAGGGATCAC
TCGCGGGCTGC-30; binding in gene orf2) and primer B (50-TGCCTGAAAG
AAGGGTTACCGC-30; binding in gene orf1).
a range of structurally divergent intermediates, encouraging the
view that the assembly line could be tailored by genetic engi-
neering to produce further variants. Similar observations have
been made for the epothilone assembly line, from which multiple
epothilone derivatives have been generated by directed modifi-
cation (Altmann et al., 2009). In the meantime, the novel analogs
can be targeted for total synthesis, in order to obtain sufficient
materials to extend the ongoing structure-activity studies of the
tubulysin metabolites. It is already promising that pretubulysin
D (5) exhibits activity similar to that of tubulysin A.
SIGNIFICANCE
Small molecules that interact with the eukaryotic cytoskel-
eton are promising candidates for development as novel
antineoplastic agents. This fact explains the significant
interest in the tubulysin family of natural products, tubulin
modifiers that are active in the low picomolar range against
a range of cancer cell lines. Until 2009, nine tubulysins were
known from the myxobacterial strains Angiococcus discifor-
mis An d48 and Archangium gephyra Ar 315. We expanded
this family recently with the discovery in A. disciformis of
pretubulysin D, the presumed first enzyme-free intermediate
in the tubulysin pathway. We now report that A. disciformis
and a third myxobacterial strain Cystobacter sp. SBCb004
together produce a further 23 novel tubulysins. Analysis
of the molecular basis for this diversity-oriented biosyn-
thesis provides new insights into the function of polyketide
synthase (PKS) and non-ribosomal peptide synthetase
(NRPS) natural product ‘‘assembly lines.’’ We have also
shown that pretubulysin D retains much of the potency of
its more structurally elaborate relatives. Thus, the 23 addi-
tional simplified analogs are attractive targets for chemical
synthesis, with the aim of deepening our understanding of
structure-activity relationships within the tubulysin family
of compounds.
Chromatography and Mass Spectrometry
Standard analysis of crude extracts was performed on an HPLC-DAD system
from the Agilent 1100 series, coupled to a Brucker Daltonics HCTultra ESI-MS
ion trap instrument operating in positive ionization mode. Compounds were
separated on a Luna RP-C18 column (125 3 2 mm; 2.5 mm particle diameter;
flow rate 0.4 ml/min, Phenomenex), with a mobile phase of water/acetonitrile
each containing 0.1% formic acid, using a gradient from 5%–95% acetonitrile
over 20 min. Detection was by both diode array and ESI-MS. High-resolution
mass spectrometry was performed onan Accela UPLC-system(Thermo-Fisher)
coupled to a linear trap-FT-Orbitrap combination (LTQ-Orbitrap), operating in
positive ionization mode. Separation was achieved using a BEH RP-C18 column
(5032mm;1.7mmparticlediameter;flowrate0.6ml/min,Waters),withamobile
phase of water/acetonitrile each containing 0.1% formic acid, using a gradient
from 5%–95% acetonitrile over 9 min. The UPLC-system was coupled to
the LTQ-Orbitrap by a Triversa Nanomate (Advion), a chip-based nano-ESI
interface. In certainexperiments, theextractswere fractionated, andsingle frac-
tions were analyzed by direct infusion. The yield of tubulysin D (Figure 4) was
determined from the peak area (base peak chromatogram) by reference to
a standard curve generated with authentic, synthetic material (a kind gift of
H. Steinmetz). Estimates of the yields of all other metabolites were obtained
by comparison of the peak areas to those of tubulysin D.
Cloning, Expression, and Purification of TubZ
The tubZ gene was amplified from cosmid F5 (Sandmann et al., 2004) contain-
ing the Angiococcus disciformis An d48 tubulysin gene cluster (primer pair:
50-GGAGCCGAATTCGTGAATGCAGTG-30 (f) and 50-GGCGGCCGCTACGGC
TGGGGC-30 (r)), introducing flanking EcoRI and NotI sites. The resulting
product was then ligated with EcoRI and NotI, and ligated into pGEX-6P-1,
previously digested with both EcoRI and NotI. The fidelity of the PCR reaction
was then confirmed by sequencing. The resulting expression construct pGEX-
6P-1-TubZ was transformed into E. coli strain BL21(DE3)pLyS. The cells were
grown in LB medium (1 l) containing ampicillin (100 mg/ml) at 30ꢁC until an
OD600 of 0.6–0.7 was reached, and then expression was induced by addition
of 0.1 mM IPTG. The cultivation was continued at 16ꢁC overnight, and then the
cells were harvested by centrifugation. The cell pellets were resuspended
in ice-cold lysis buffer (25 ml; 1 3 PBS buffer [10 mM Na2HPO4, 1.8 mM
KH2PO4 (pH 7.3), 140 mM NaCl, 2.7 mM KCl], containing 5 mM DTT and
0.5% Triton X-100) and lysed by French Press. The lysis solution was centri-
fuged (14,000 g) at 4ꢁC for 30 min, and then the clarified lysate was mixed
with a 50% slurry of glutathione Sepharose 4B beads (2 ml; GE Healthcare),
and incubated at room temperature for 30 min, with gentle agitation. The
beads were then washed with buffer (1 3 PBS buffer; 10 column volumes).
In the case of the GST-TubZ fusion protein, elution was performed directly
with elution buffer (50 mM Tris-HCl [pH 8.0], 10 mM reduced glutathione);
the protein was eluted in three fractions (10 min incubation). Discrete TubZ
was obtained by on-column digestion of the bound GST-TubZ fusion protein.
For this, PreScission Proteaseꢀ (5 units) was added to the beads in protease
buffer (50 mM Tris-HCl [pH 7.0], 150 mM NaCl, 1 mM EDTA, 1 mM DTT), and
incubated at 4ꢁC overnight. The purified proteins were flash frozen in liquid
nitrogen, and stored at ꢀ80ꢁC until use.
EXPERIMENTAL PROCEDURES
Culture Conditions
Cystobacter SBCb004 was cultivated at 30ꢁC and 170 rpm in M-medium (1%
papaic digest of soybean meal, 1% maltose, 0.1% CaCl2$2H2O, 0.1%
MgSO4$7H2O, 8 mg/l NaFe-EDTA, 1.19% HEPES, adjusted to pH 7.2 with
10 N KOH). Mutant SBCb004-tubZꢀ was grown in the presence of kanamycin
(100 mg/ml). For analysis of secondary metabolite production, the wild-type
and mutant strains were grown at 30ꢁC for 5–7 days in 50 ml M-medium
containing 1% adsorber resin Amberlite XAD-16. The resin was separated
from the medium by sieving, and then extracted with 40 ml of methanol.
The extracts were evaporated to dryness and resuspended in 1 ml methanol
before analysis by mass spectrometry. The cultivation and extraction condi-
tions of An d48 wild-type and mutant An d48-tubZꢀ were as reported earlier
(Ullrich et al., 2009).
Inactivation of tubZ in SBCb004
The construction of gene tubZSBCb004 knockout in SBCb004 was accom-
plished by homologous recombination in the encoding regions. Primer pairs,
Cbt-Z1 (50-GAAGTGACTCATGTTGAGCACGAGGGTG-30) and Cbt-Z2 (50-GA
GCGCACCAGCTCGAAGAG-30), were designed to amplify the knockout
region (864 bp) in the middle of target gene tubZSBCb004, and introduce a
stop codon. The amplified DNA fragment was then cloned into pCR
2.1-TOPO vector (Invitrogen) to obtain the inactivation vector pTOPO-tubZ.
The tubZSBCb004 inactivation mutant SBCb004-tubZꢀ was generated by
306 Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved

Chemistry & Biology
Discovery of 23 Tubulysins
Characterization of TubZ Cyclodeaminase Activity
Synthesis of L-4,5-Dideutero-pipecolic acid-hydrochloride
(L-d2-Pip) and D-4,5-Dideutero-pipecolic acid-hydrochloride
(D-d2-Pip)
Assays contained TubZ–GST fusion protein (1.5 mM) or discrete TubZ protein
(0.6 mM), L- or D-lysine (1 mM), NAD+ (100 mM), and BSA (1 mg/ml) were
performed in buffer (5 mM Tris-HCl [pH 8]) in a total volume of 100 ml. Control
incubations were performed with L- or D-lysine in the absence of added TubZ.
After 24 hr incubation at room temperature, the reactions were quenched by
the addition of acetonitrile (200 ml). The assay mixture was chilled at ꢀ20ꢁC
for at least 10 min, and then centrifuged (16,060 g) at 4ꢁC for 10 min. Deriva-
tization was carried out on the assay supernatant (240 ml), using Fmoc-Cl
(10 mM; 20 ml) in borate buffer (10 mM; 80 ml). The reaction mixture was
incubated at room temperature for 30 min, and then quenched by addition
of 2 volumes of pure acetic acid. The same derivatization procedure was
performed on a commercially available racemic mixture of L-/D-pipecolic
acid, and enantiomerically pure D-pipecolic acid. Following centrifugation,
the supernatants were analyzed using the HPLC-DAD system (see above)
at 25ꢁC. A Nucleodex b-OH column (200 3 4 mm) (Machery-Nagel) was
employed for separation with a solvent system consisting of H2O and acetoni-
trile, each containing 0.1% formic acid. The following gradient was applied:
30%–95% acetonitrile over 35 min, at a flow rate 0.5 ml/min.
A solution of HCl in dioxane (4 M, 4 ml, 16 mmol) was added to L-d2-Boc-Pip
(185 mg, 0.803 mmol). When deprotection was complete, the suspension was
centrifuged (6000 rpm) for 15 min and the HCl-dioxane solution was decanted
off. The product was dried under high-vacuum to yield L-d2-Pip (107 mg,
0.642 mmol, 78% yield) as a white solid, m. p. 240ꢁC (decomp.). D-d2-Pip
was likewise obtained as a white solid (0.654 mmol, 81% yield). L-d2-Pip:
20
20
½aꢂD = –6.3 (c 1.0, MeOH), D-d2-Pip: ½aꢂD = +6.2 (c 1.0, MeOH). 1H-NMR
(400 MHz, D2O): d 1.58–1.73 (m, 1 H), 1.78 (m, 1 H), 1.87–1.98 (m, 1 H), 2.31
(dt, J = 14.5 Hz, 3.9 Hz, 1 H), 3.06 (m, 1 H), 3.48 (dd, J = 12.8 Hz, 3.8 Hz,
1 H), 3.96 (dd, J = 11.8 Hz, 3.5 Hz, 1 H); 13C-NMR (100 MHz, D2O): d 21.0,
25.7, 43.8, 57.0, 172.0; HRMS (CI) m/z [M+H]+ 132.1002 (calculated for
C6H10D2NO2,132.0993).
Feeding Studies
Feeding studies of An d48 wild-type and An d48-tubZꢀ mutant were
performed in 50 ml M7 culture (0.5% probion, 0.1% CaCl2$2H2O, 0.1%
MgSO4$7H2O, 0.1% yeast extract, 0.5% soluble starch, 1% HEPES
[pH 7.4]), with synthetic deuterium-labeled D-/L-pipecolic acid (1 mg), d8
L-valine (2.5 mg) and d3 L-methionine (25 mg). The compounds were dissolved
in water, sterile filtered and administered in three portions to one-day old
cultures in three parts (100 ml every 12 hr). The cells were cultivated for an
additional day. 1% Amberlite XAD absorber resin was then added to the
culture, and the cultivation was continued for a further 5 days. The adsorber
resin and cell clumps were separated from the medium by sieving, and then
extracted with 40 ml of a methanol/acetone mixture (1:1 v/v). The extracts
were evaporated to dryness and resuspended in 0.5 ml methanol before
analysis by HPLC-MS or LTQ-Orbitrap.
Isolation, Purification and Structure Elucidation of Pretubulysin A
The SBCb004-tubZꢀ mutant was cultivated in M-medium in the presence of
2% XAD adsorber resin at 30ꢁC for 6 days. Cultivation was carried out in
a 17 l fermenter and shaking flasks (9 l; 150 rpm), for a total volume of 26 l.
Fermentation was performed with stirring at 300 rpm, an oxygen saturation
of 35%, and at a pH of 7.2–7.4. The XAD was separated from the supernatant
by sieving. The XAD was then extracted batchwise with a total volume of 6 l
methanol. The resulting extracts were combined and evaporated. Before initial
purification, the methanolic raw extracts were degreased by counter extrac-
tion with heptane. This procedure afforded 18.4 g crude extract.
Initial purification was performed by vacuum liquid chromatography on silica
gel, using a stepwise gradient from nonpolar to polar solvents: hexane,
chloroform, chloroform/methanol (3:1, 2:1 and 1:1), and methanol. Fractions
containing significant amounts of pretubulysin A were combined and evapo-
rated to dryness. Pure pretubulysin A was obtained by purification on a prepar-
ative HPLC instrument (Waters), equipped with a 2545 binary gradient module,
a 515 HPLC pump, a SFO system fluidics organizer, a 2998 PDA, a 3100 mass
detector, and a 2767 sample manager. An X-Bridge RP-C18 column (19 3
100 mm, 5 mm) (Waters) was used for separation with a solvent system consist-
ing of water (A) and methanol (B), each containing 0.1% formic acid. The
following gradient was applied: 0–0.8 min, 35% B; 0.8–6.8 min 35%–55% B;
6.8–10 min 55%–67.3% B. Fractions were collected using mass-guided
fraction collection, with a trigger mass of 686.5 Da. This procedure yielded
approximately 6.5 mg of purified pretubulysin A. NMR spectra were recorded
on a Bruker Avance 500 spectrometer using CD3OD as solvent and as an
internal standard. For 1H and 13C NMR data, see Table 2.
Synthesis of L-N-tert-butoxycarbonyl-4,5-dideutero-pipecolic
acid (L-d2-Boc-Pip) and D-N-tert-butoxycarbonyl-4,5-dideutero-
pipecolic acid (D-d2-Boc-Pip)
A solution of N-Boc-protected L-4,5-dehydropipecolic acid (186 mg, 0.818
mmol) and Pd/C (10%) catalyst (24 mg) in methanol (10 ml) was stirred under
adeuteriumatmosphere. Whenthereactionwas finished(TLC control), thecata-
lyst was filtered away using a pad of celite, and the methanol was evaporated to
afford L-d2-Boc-Pip (185 mg, 0.803 mmol, 98% yield) as a white solid (m. p.
104ꢁC). D-d2-Boc-Pip was obtained by the same method, in quantitative yield.
20
L-d2-Boc-Pip: ½aꢂD = –45.5 (c 1.0, MeOH) [Lit (Johnson et al., 1986):
24
½aꢂD = –45.1 (c 1.0, MeOH, L-Boc-Pip)].
20
D-d2-Boc-Pip: ½aꢂD = +41.2 (c 1.0, MeOH) [Lit (Garvey et al., 1992):
23
½aꢂD = +43.6 (c 0.96, MeOH, D-Boc-Pip)].
1H-NMR analysis at room temperature showed a 1.7:1 mixture of rotamers.
The NMR spectra and mass analysis also indicated that deuteration had
occurred to varying extents, yielding a mixture of tri-deuterated (15%),
di-deuterated (60%), mono-deuterated (20%), and undeuterated pipecolic
acid (7%), in each case. 1H-NMR (400 MHz, CDCl3) (di-deuterated): d = 1.27
(m, 0.5 H), 1.41 (m, 0.5 H), 1.43 (s, 9 H), 1.56–1.73 (m, 2 H), 2.19 (bs, 1 H),
2.93 (bs, 1 H), 3.90 (d, J = 10.8 Hz, 1 H), 4.90 (s, 1 H), 10.00 (bs, 1 H);
13C-NMR (100 MHz, CDCl3): d = 20.4, 24.3, 26.5, 28.3, 42.0, 53.5, 80.3,
156.2, 177.5; characteristic signals of the minor rotamer: 1H-NMR (400 MHz,
CDCl3): d = 2.89 (bs, 1 H), 3.97 (bs, 1 H), 4.74 (bs, 1 H); 13C-NMR (100 MHz,
CDCl3): d 41.0, 54.7, 155.5, 177.4; HRMS (CI) m/z [M+H]+ 232.1519 (calculated
for C11H18D2NO4, 232.1518).
Isolation of N-Desmethyl 12-keto pretubulysin (14)
Wild-type An d48A fermentation was grown in M7-medium (8.5 l) in the
presence of 2% Amberlite XAD-16. Fermentation was carried out at 30ꢁC
(pH 7.2–7.4, O2 saturation of 35%, 250 rpm). By HPLC-MS analysis, produc-
tion of 14 was observed to plateau at day 8, at which point the culture was
harvested. The XAD and the cells were separated from the medium by centri-
fugation, and then stepwise extracted with acetone (1.5 l total) and methanol
(3.0 l total). The extracts were filtered, and the solvent removed under vacuum.
The remaining 0.3 l aqueous solution was extracted three times with an equal
amount of heptane. The water was then removed by freeze-drying to yield
11.2 g of crude extract. The crude extract was fractionated by silica gel-VLC
chromatography using a stepwise gradient from nonpolar to polar solvents
(hexane, hexane-chloroform (1:1), chloroform, chloroform-methanol (9:1, 3:1,
1:1), methanol, acetonitrile. The resulting fractions were dried under vacuum
and analyzed by HPLC-MS. Fractions containing N-desmethyl 12-keto pretu-
bulysin (14) were pooled, and 14 was purified further by chromatography on
a Sephadex LH20 column (100 3 2.5 cm; flow rate 0.083 L/h over 24 h) using
Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved 307

Chemistry & Biology
Discovery of 23 Tubulysins
methanol. The fraction containing 14 was applied to a semipreparative HPLC
system (Dionex P680 HPLC pump connected to a PDA-100 photo diode array
detector; X-Bridge RP-C18 column (19 3 100 mm; 5 mm particle diameter,
Waters) in a water/acetonitrile gradient (both solvents contained 0.1% formic
acid). Chromatographic conditions were: flow rate 10 ml/min; detection
wavelengths 210, 254, 366 nm; mobile phase: 35% acetonitrile, 5 min, gradient
from 35%–45% acetonitrile, 6 min. This step yielded 1.3 mg of N-desmethyl
12-keto pretubulysin (14) as a slightly beige amorphous solid.
Do¨ mling, A., Beck, B., Eichelberger, U., Sakamuri, S., Menon, S., Chen, Q.Z.,
Lu, Y., and Wessjohann, L.A. (2006). Total synthesis of tubulysin U and V.
Angew. Chem. Int. Ed. Engl. 45, 7235–7239.
Dubern, J.F., Coppoolse, E.R., Stiekema, W.J., and Bloemberg, G.V. (2008).
Genetic and functional characterization of the gene cluster directing the
biosynthesis of putisolvin I and II in Pseudomonas putida strain PCL1445.
Microbiology 154, 2070–2083.
Firn, R.D., and Jones, C.G. (2003). Natural products – a simple model to
explain chemical diversity. Nat. Prod. Rep. 20, 382–391.
ACCESSION NUMBERS
Garvey, D.S., Wasicak, J.T., Chung, J.Y.L., Shue, Y.K., Carrera, G.M., May,
P.D., Mckinney, M.M., Anderson, D., Cadman, E., Vellarountree, L., et al.
(1992). Synthesis and in vitro characterization of novel amino terminally
modified oxotremorine derivatives for brain muscarinic receptors. J. Med.
Chem. 35, 1550–1557.
The sequence of the tubulysin biosynthetic gene cluster in Cystobacter
sp. SBCb004 has been deposited in GenBank under accession number
GU0002154.
Gatto, G.J., Boyne, M.T., Kelleher, N.L., and Walsh, C.T. (2006). Biosynthesis
of pipecolic acid by RapL, a lysine cyclodeaminase encoded in the rapamycin
gene cluster. J. Am. Chem. Soc. 128, 3838–3847.
SUPPLEMENTAL INFORMATION
Supplemental Information includes eight figures and one table and can be
found with this article online at doi:10.1016/j.chembiol.2010.01.016.
Grubbs, R.H., Miller, S.J., and Fu, G.C. (1995). Ring-closing metathesis and
related processes in organic synthesis. Acc. Chem. Res. 28, 446–452.
Johnson, R.L., Rajakumar, G., Yu, K.L., and Mishra, R.K. (1986). Synthesis of
Pro-Leu-Gly-NH2 analogs modified at the prolyl residue and evaluation of their
effects on the receptor-binding activity of the central dopamine receptor
agonist, ADTN. J. Med. Chem. 29, 2104–2107.
ACKNOWLEDGMENTS
The authors would like to thank Eva Luxenburger, Daniel Krug, and Ole Rever-
mann from the Department of Pharmaceutical Biotechnology for skillful help
with various analytical technologies, and Axel Sandmann for assistance with
the analysis in vitro of TubZ. We thank Heinrich Steinmetz for provision of
synthetic tubulysin D. Work in R.M.’s laboratory is funded by the Bundesminis-
terium fu¨ r Bildung und Forschung and the Deutsche Forschungsgemeinschaft.
Kaur, G., Hollingshead, M., Holbeck, S., Schauer-Vukasinovic, V., Camalier,
R.F., Domling, A., and Agarwal, S. (2006). Biological evaluation of tubulysin
A: a potential anticancer and antiangiogenic natural product. Biochem. J.
396, 235–242.
Kazmaier, U., and Schneider, C. (1998). Application of the asymmetric chelate
enolate Claisen rearrangement to the synthesis of unsaturated polyhydroxy-
lated amino acids. Synthesis, 1321–1326.
Received: September 30, 2009
Revised: January 21, 2010
Accepted: January 28, 2010
Published: March 25, 2010
Khalil, M.W., Sasse, F., Lu¨ nsdorf, H., Elnakady, Y.A., and Reichenbach, H.
(2006). Mechanism of action of tubulysin, an antimitotic peptide from
myxobacteria. ChemBioChem 7, 678–683.
REFERENCES
Khaw, L.E., Bohm, G.A., Metcalfe, S., Staunton, J., and Leadlay, P.F. (1998).
Mutational biosynthesis of novel rapamycins by a strain of Streptomyces
hygroscopicus NRRL 5491 disrupted in rapL, encoding a putative lysine
cyclodeaminase. J. Bacteriol. 180, 809–814.
¨
¨
Altmann, K.H., Hofle, G., Muller, R., Mulzer, J., and Prantz, K. (2009). The
Epothilones – An Outstanding Family of Anti-Tumour Agents, A.D. Kinghorn,
H. Falk, and J. Kobayashi, eds. (New York: Springer).
Kittendorf, J.D., Beck, B.J., Buchholz, T.J., Seufert, W., and Sherman, D.H.
(2007). Interrogating the molecular basis for multiple macrolactone ring forma-
tion by the pikromycin polyketide synthase. Chem. Biol. 14, 944–954.
Altschul, S.F., Madden, T.L., Schaffer, A.A., Zhang, J.H., Zhang, Z., Miller, W.,
and Lipman, D.J. (1997). Gapped BLAST and PSI-BLAST: a new generation of
protein database search programs. Nucleic Acids Res. 25, 3389–3402.
Kopp, M., Irschik, H., Pradella, S., and Mu¨ ller, R. (2005). Production of the
tubulin destabilizer disorazol in Sorangium cellulosum: biosynthetic machinery
and regulatory genes. ChemBioChem 6, 1277–1286.
Balasubramanian, R., Raghavan, B., Steele, J.C., Sackett, D.L., and Fecik,
R.A. (2008). Tubulysin analogs incorporating desmethyl and dimethyl tubuphe-
nylalanine derivatives. Bioorg. Med. Chem. Lett. 18, 2996–2999.
Linne, U., Doekel, S., and Marahiel, M.A. (2001). Portability of epimerization
domain and role of peptidyl carrier protein on epimerization activity in nonribo-
somal peptide synthetases. Biochemistry 40, 15824–15834.
Balasubramanian, R., Raghavan, B., Begaye, A., Sackett, D.L., and Fecik, R.A.
(2009). Total synthesis and biological evaluation of tubulysin U, tubulysin V,
and their analogues. J. Med. Chem. 52, 238–240.
Luo, L., Kohli, R.M., Onishi, M., Linne, U., Marahiel, M.A., and Walsh, C.T.
(2002). Timing of epimerization and condensation reactions in nonribosomal
peptide assembly lines: kinetic analysis of phenylalanine activating elongation
modules of tyrocidine synthetase B. Biochemistry 41, 9184–9196.
Balibar, C.J., Vaillancourt, F.H., and Walsh, C.T. (2005). Generation of D-amino
acid residues in assembly of arthrofactin by dual condensation/epimerization
domains. Chem. Biol. 12, 1189–1200.
Banerji, S., and Flieger, A. (2004). Patatin-like proteins: a new family of lipolytic
Margulies, M., Egholm, M., Altman, W.E., Attiya, S., Bader, J.S., Bemben, L.A.,
Berka, J., Braverman, M.S., Chen, Y.J., Chen, Z., et al. (2005). Genome
sequencing in microfabricated high-density picolitre reactors. Nature 437,
376–380.
enzymes present in bacteria? Microbiology 150, 522–525.
Bode, H.B., and Mu¨ ller, R. (2006). Analysis of myxobacterial secondary
metabolism goes molecular. J. Ind. Microbiol. Biotechnol. 33, 577–588.
Meiser, P., Weissman, K.J., Bode, H.B., Krug, D., Dickschat, J.S., Sandmann,
A., and Mu¨ ller, R. (2008). DKxanthene biosynthesis – understanding the basis
for diversity-oriented synthesis in myxobacterial secondary metabolism.
Chem. Biol. 15, 771–781.
Carvalho, R., Reid, R., Viswanathan, N., Gramajo, H., and Julien, B. (2005). The
biosynthetic genes for disorazoles, potent cytotoxic compounds that disrupt
microtubule formation. Gene 359, 91–98.
De Crecy-Lagard, V., Marliere, P., and Saurin, W. (1995). Multienzymatic non
ribosomal peptide biosynthesis: identification of the functional domains
catalysing peptide elongation and epimerisation. C. R. Acad. Sci. III 318,
927–936.
Miller, J.F., Termin, A., Koch, K., and Piscopio, A.D. (1998). Stereoselective
synthesis of functionalized carbocycles and heterocycles via an ester enolate
Claisen/ring-closing metathesis manifold. J. Org. Chem. 63, 3158–3159.
¨
Domling, A., and Richter, W. (2005). Myxobacterial epothilones and tubulysins
Moraleda-Munoz, A., and Shimkets, L.J. (2007). Lipolytic enzymes in
as promising anticancer agents. Mol. Divers. 9, 141–147.
Myxococcus xanthus. J. Bacteriol. 189, 3072–3080.
308 Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved

Chemistry & Biology
Discovery of 23 Tubulysins
Mu¨ ller, C., Nolden, S., Gebhardt, P., Heinzelmann, E., Lange, C., Puk, O.,
Welzel, K., Wohlleben, W., and Schwartz, D. (2007). Sequencing and analysis
of the biosynthetic gene cluster of the lipopeptide antibiotic friulimicin in
Actinoplanes friuliensis. Antimicrob. Agents Chemother. 51, 1028–1037.
Schneider, C., and Kazmaier, U. (1998). Asymmetric syntheses of chiral allylic
alcohols. Synthesis, 1314–1320.
Schuster, M., and Blechert, S. (1997). Olefin metathesis in organic chemistry.
Angew. Chem. Int. Ed. Engl. 36, 2037–2056.
Patterson, A.W., Peltier, H.M., Sasse, F., and Ellman, J.A. (2007). Design,
synthesis, and biological properties of highly potent tubulysin D analogues.
Chemistry 13, 9534–9541.
Stachelhaus, T., and Walsh, C.T. (2000). Mutational analysis of the epimeriza-
tion domain in the initiation module PheATE of gramicidin S synthetase.
Biochemistry 39, 5775–5787.
Steinmetz, H., Glaser, N., Herdtweck, E., Sasse, F., Reichenbach, H., and
Patterson, A.W., Peltier, H.M., and Ellman, J.A. (2008). Expedient synthesis
of N-methyl tubulysin analogues with high cytotoxicity. J. Org. Chem. 73,
4362–4369.
¨
Hofle, G. (2004). Isolation, crystal and solution structure determination, and
biosynthesis of tubulysins—powerful inhibitors of tubulin polymerization from
myxobacteria. Angew. Chem. Int. Ed. Engl. 43, 4888–4892.
Perlova, O., Gerth, K., Hans, A., Kaiser, O., and Mu¨ ller, R. (2006). Identification
and analysis of the chivosazol biosynthetic gene cluster from the myxobacte-
rial model strain Sorangium cellulosum So ce56. J. Biotechnol. 121, 174–191.
Ullrich, A., Chai, Y., Pistorius, D., Elnakady, Y.A., Herrmann, J.E., Weissman,
K.J., Kazmaier, U., and Mu¨ ller, R. (2009). Pretubulysin, a potent and chemi-
cally-accessible tubulysin precursor from Angiococcus disciformis. Angew.
Chem. Int. Ed. Engl. 48, 4422–4425.
Raghavan, B., Balasubramanian, R., Steele, J.C., Sackett, D.L., and Fecik,
R.A. (2008). Cytotoxic simplified tubulysin analogues. J. Med. Chem. 51,
1530–1533.
Wang, Z., McPherson, P.A., Raccor, B.S., Balachandran, R., Zhu, G.Y., Day,
B.W., Vogt, A., and Wipf, P. (2007). Structure-activity and high-content
imaging analyses of novel tubulysins. Chem. Biol. Drug Des. 70, 75–86.
Rausch, C., Hoof, I., Weber, T., Wohlleben, W., and Huson, D.H. (2007).
Phylogenetic analysis of condensation domains in NRPS sheds light on their
functional evolution. BMC Evol. Biol. 7, 78–92.
Weissman, K.J., and Mu¨ ller, R. (2008). A brief tour of myxobacterial secondary
metabolism. Bioorg. Med. Chem. 17, 2121–2136.
Rothberg, J.M., and Leamon, J.H. (2008). The development and impact of 454
Wenzel, S.C., and Mu¨ ller, R. (2005). Recent developments towards the
heterologous expression of complex bacterial natural product biosynthetic
pathways. Curr. Opin. Biotechnol. 16, 594–606.
sequencing. Nat. Biotechnol. 26, 1117–1124.
Sandmann, A., Sasse, F., and Mu¨ ller, R. (2004). Identification and analysis of
the core biosynthetic machinery of tubulysin, a potent cytotoxin with potential
anticancer activity. Chem. Biol. 11, 1071–1079.
Wenzel, S.C., and Mu¨ ller, R. (2007). Myxobacterial natural product assembly
lines: fascinating examples of curious biochemistry. Nat. Prod. Rep. 24,
1211–1224.
Sani, M., Fossati, G., Huguenot, F., and Zanda, M. (2007). Total synthesis of
tubulysins U and V. Angew. Chem. Int. Ed. Engl. 46, 3526–3529.
Zumpe, F.L., and Kazmaier, U. (1998). A mild palladium catalyzed N-allylation
of amino acids and peptides. Synlett, 1199–1200.
¨
Sasse, F., Steinmetz, H., Heil, J., Hofle, G., and Reichenbach, H. (2000).
Tubulysins, new cytostatic peptides from myxobacteria acting on microtubuli:
production, isolation, physico-chemical and biological properties. J. Antibiot.
(Tokyo) 53, 879–885.
Zumpe, F.L., and Kazmaier, U. (1999). Application of the palladium catalyzed
N-allylation to the modification of amino acids and peptides. Synthesis,
1785–1791.
Chemistry & Biology 17, 296–309, March 26, 2010 ª2010 Elsevier Ltd All rights reserved 309