Solid-Phase Synthesis and Combinatorial Technologies

a:6 selection/amplification cycles using substrate 1 (cycles 1,4), 2 (cycles 2,5), 3 (cycles 3,6) ;

b:7 cycles with heavily mutagenic PCR conditions.

CONSENSUS SEQUENCE:

29-mer sequence (from final clones) UGUGGUGAAAAAAUAGAAUCGAUCUUGUC

containing also the GAAA-based Mn++-dependent RNA-cleavage motif.

10.38

small 29-mer ribozyme dual catalytic activity

pHand divalent ion-dependent activity switch

Figure 10.33 Selection of artificial ribozymes with dual, switchable RNA cleavage and ligation activity from the biosynthetic ON ribozyme library L23: the selection/amplification process and the structure of the optimized ribozyme 10.38.

ribozymes with broad specificity. Seven further, more focused cycles were performed using heavily mutagenic PCR amplification of selected sequences to produce a family of related catalytic RNA molecules (step b, Fig. 10.33). Comparison of their sequences highlighted a highly conserved, 29-mer catalytic core (10.38, Fig. 10.33) that was prepared separately and showed similar, Mg2+-dependent activity to the integral sequences selected from round 63. Surprisingly, this 29-mer sequence also contained a 7-mer that was already known as a Mn2+-dependent RNA cleavage motif (Fig. 10.33) previously observed in naturally occurring ribozymes and that also occurred in the longer sequences (213, 214). Further studies highlighted the dual nature of 10.38, which, according to the pH and the presence of different divalent cations, was able to switch from ligation of RNA to self-cleavage.

The existence of a coordination site for divalent cations and the possibility for the ribozyme to adopt two different conformations, one suited to ligation, the other to self-cleavage, were possible explanations for this dual activity. The occurrence of a parallel, catalytic activity that was selective for Mn2+, even though this cation was never used in the in vitro selection experiments, hinted at the flexibility of ribozymes able to adapt their properties to different conditions and that could allow the evolution of new catalytic activities. Other artificial ribozymes selected in vitro have been reported by Vaish et al. (hammerhead-like) (215), Yu et al. (hairpin-like) (216), Robertson and Ellington (allosteric ligase activated by ON effectors) (217), and

544 BIOSYNTHETIC COMBINATORIAL LIBRARIES

Santoro and Joyce (218). The last authors cited reported the selection of multipurpose RNA-cleaving DNAzymes with potential applications in molecular biology. The elusive RNA replicases, which should be able to replicate RNA structures including their own to sustain the RNA world hypothesis, are a common target for research; a recent review (219) summarized the efforts in this field.

A primordial RNA world would have required the efficient assembly of single nucleotides into ON chains. Sugar phosphates (220) and nucleobases (221) have been prepared under conditions that supposedly mimic those found in this primordial world, but the assembly of the two nucleotide constituents remained a challenge. Unrau and Bartel (222) studied the nucleophilic attack of activated ribose (pRpp, 10.39) on uracil (10.40) to give uridine 5′-phosphate (UMP, 10.41, Fig. 10.34), a reaction that is catalyzed by the enzyme uracylphosphoribosyl transferase (UPRT). A 1.5 × 1015- member library of 294-mer ONs, L24, with 228 random positions was prepared using standard synthetic protocols and was submitted to the in vitro selection strategy depicted in Fig. 10.35. Activated ribose 10.39 was first condensed with activated adenosine 10.42 (step a); then the resulting compound 10.43 was condensed onto L24 (step b) to give 10.44. Compound 10.44 was incubated in the presence of 4-thiouracil

10.45(step c, 8 mM) (223) for 18 h in the first six cycles, gradually decreasing the concentration of 10.45 to 40 µM and the incubation time to 7.5 min for five additional

rounds. The presence of a 4-thio group on 10.46 allowed the separation of the reacted RNA structures from the inactive components (step a, Fig. 10.36). The reaction of

10.46with iodoacetyl biotin to give 10.47 (step b) allowed the separation of the reacted RNA by capture with supported streptavidin (step c, Fig. 10.36). The bound sequences were eluted and submitted to standard reverse transcription, PCR, and transcription protocols (steps d–g). The new pools of RNA were submitted to further selection cycles (step h, Fig. 10.36). Error-prone PCR (224) was used for amplification of individuals from rounds 4–6 to introduce mutations that were absent from L24 and to include additional structural diversity. After 11 selection cycles 35 clones were randomly

selected, and three families of ribozymes were identified. Representative examples of these families showed significant catalytic activity, with up to 107 times enhancement

compared with the noncatalyzed reaction and selectivity toward 10.45, as only uracil was slowly condensed with 3′-derivatized 10.44 while the other thionucleobases did

Figure 10.34 Uracylphosphoribosyltransferase (UPRT)-catalyzed synthesis of uridine-5- phosphate 10.41.

546 BIOSYNTHETIC COMBINATORIAL LIBRARIES

OH OH

10.47

a: separation by two-dimensional TLC; b: aq. buffer, DMF, rt, 3 hrs; c: capture of 10.47 with streptavidin-coated beads; d: elution of ribozymes; e: reverse transcription to DNA; f: PCR amplification; g: transcription into RNA; h: 11 iterative cycles and final selection.

3 REPRESENTATIVE RIBOZYME FAMILIES:

up to 107 enhancement of nucleotide synthesis compared to uncatalyzed reaction; specificity for 10.46, as no reaction was observed with 2-thiouracil, 2,4-thiouracil, 2-thiocytosine, 2-thiopyrimidine, 2-thiopyridine or 5-carboxy-2-thiouracil.

Figure 10.36 Selection of artificial ribozymes with nucleotide-assembling properties from the biosynthetic ON ribozyme library L24: the selection/amplification process from 10.46 to three ribozyme families with nucleotide synthesis properties.

selection substrate 10.50 (AMP-Met-biotin, Fig. 10.37). The in vitro selection strategy adopted is shown in Fig. 10.38. Incubation of L25 with 10.50 (8 mM) for 20 h at 25°C (step a) in the presence of Mg2+ (50 mM) induced the nucleophilic attack of the free NH2 onto the ester bond of 10.50, releasing AMP and linking the biotinyl-Met moiety to the active RNAs 10.51. The library was eluted through a streptavidin column that complexed the biotin-containing ribozymes (step b). Treatment with dithiothreitol (DTT, step c) reduced the disulfide bonds and eluted the selected free 5′-GMPS-RNAs 10.52 that had been selected (Fig. 10.38). After standard reverse transcription and PCR amplification, the transcription of dsDNA was again performed in the presence of GMPS (step d). The RNA pool was converted to 5′-Phe-S-S-RNA as described in Fig. 10.37, and a new selection cycle was started. The first nine selection rounds used the same reaction conditions (step e), and then the concentration of 10.50 (up to 200 µM) and the incubation time (up to 1 h) were reduced to increase the stringency (step f, Fig. 10.38). The RNA population selected after 19 cycles, corresponding to a >10,000-fold increase in catalytic efficiency, was processed individually, and 75 sequences were cloned. Of these, nine sequences were found to be more active than the selection pool and could be divided into two main structural families. The most active RNA sequence (10.53, Fig. 10.38) was fully characterized and showed a markedly enhanced catalytic activity (106 times the uncatalyzed peptide bond formation). This activity was lost

10.50

O O OH

NH

BIOTIN

S

Figure 10.37 Selection of artificial ribozymes with peptide-assembling properties from the biosynthetic ON riboyzyme labeled library L25: structure of the key intermediates 10.49 and

10.50.

upon deletion of both the 3′- and 5′-ends and removal of Mg2+ ions and the amide link between 10.50 and 5′-Phe-S-S-RNA. The activity was also diminished when the linker between Phe and GMPS was shortened. The ribozyme 10.53 also showed a broader substrate specificity and was able to use AMP-Leu-biotin and, to a lesser extent, phenylalanine and lysine-containing substrates as donors. Even though the activity of 10.53 was significantly lower than the natural peptide–peptide bond-forming machinery of the ribosome, 10.53 successfully demonstrated the feasibility of peptide bondforming ribozymes able to act on multiple substrates. A few similar reports (226–229) have confirmed the activity of families of ribozyme in acylations and peptide bond formation.

Naturally occurring and synthetic ribozymes often require metal ion cofactors (usually Mg2+) to be effective, as demonstrated in the previous examples. However, the involvement of nucleotide-like cofactors (230) and allosteric regulatory mecha-

548 BIOSYNTHETIC COMBINATORIAL LIBRARIES

S

a:incubation with 10.50 (8 mM), 25°C, 20 hrs; b: absorption on a streptavidin column;

c:DTT, rt; d: standard reverse transcription, PCR amplification, transcription protocols;

e:eight iterative selection/amplification cycles; f: ten iterative cycles with decreasing 10.50

(up to 200 M) and decreasing incubation time (up to 1 hr).

MOST ACTIVE RIBOZYME: 10.53

>106 increase of catalytic efficiency compared to the uncatalyzed reaction; broad substrate specificity.

Figure 10.38 Selection of artificial ribozymes with peptide-assembling properties from the biosynthetic ON ribozyme labeled library L25: the selection/amplification process using 10.50 as the selection substrate to the most active ribozyme 10.53.

nisms (217, 231) has also been reported and illustrates the high degree of versatility found among the ribozymes. Roth and Breaker (232) have described the selection of L-histidine-dependent RNA cleavage DNAzymes in vitro from a 2 × 1013-member library of >100-mer modified ODNs, L26, in which a biotin moiety was attached onto the 5′-position. The library contained a single oligonucleotide as the self-cleavage site and a 40-mer random sequence from which to select the active DNAzymes (Fig. 10.39). L26 was absorbed onto a streptavidin-derivatized column matrix (step a) via its biotinylated end, so that self-cleavage of the active library individuals caused their release during incubation (vide infra) while the unreactive DNAs remained anchored onto the column. The column was first washed with aqueous buffer to discard non-histidine-dependent DNAzymes (step b) and was then eluted with three portions

L26

40-mer inner randomized sequences 5'-biotinylated individuals

a: adsorption onto a streptavidin column; b: elution with buffers, to waste; c: elution with L-histidine (50 mM) and EDTA, pH 7.5, 23°C, 1 hr; d: standard protocols for PCR amplification;

e: reconstruction of biotinylated end; f: six repeated a-e cycles; g: 11 iterative cycles with shorter elution times (up to 15').

REPRESENTATIVE DNAzyme: 10.54

39-mer selected sequence (deletion) significant but low His-dependent, RNA-cleavage activity

h

L27

10.54-focused library

1013-member ODNs up to 7 mutated residues

in the 39-mer catalytic sequence

i

10.55 and 10.56

100-fold increased activity in respect to 10.54

strong dependence from L-His strong cofactor selectivity:

lack of activity for D-His, modified histidines and other AAs.

h: synthesis of L27 (0.21 degeneracy of each position in the 39-mer sequence); i: standard selection/amplification protocols, five cycles.

Figure 10.39 Selection of artificial ribozymes with amino acidic cofactors from the biosynthetic modified on DNAzyme libraries L26 and L27: the selection/amplification process to the L-his dependent ribozymes 10.55 and 10.56.

of 50 mM L-histidine solutions for 1 h at pH 7.5 and 23°C (step c) to select for the L-His-dependent ribozymes. The solutions also contained the metal chelator ethylenediaminetetraacetic acid (EDTA) to prevent the selection of metal cofactor-dependent DNAzymes. Selected DNA sequences were PCR amplified and the proper biotinylated 5′-ends were reconstructed to provide a new pool of DNA for another selection cycle (steps d and e). Six cycles were performed as above (step f); then an additional four followed with shorter elution times (15–25 min) to increase stringency (step g). The DNA population after the 11th cycle was characterized, and a family of six sequences

550 BIOSYNTHETIC COMBINATORIAL LIBRARIES

of L-histidine-dependent DNAzymes was obtained. These sequences contained 39 ONs in the randomized sequence, probably due to a single deletion event during in vitro selection. Their activity, although significant, was ~1,000-fold lower than most natural self-cleaving ribozymes. For this reason, the structure of the ribozyme 10.54 (Fig. 10.39) was used to prepare a focused library of ODNs L27 in which all the individuals containing up to seven modifications with respect to 10.54 were represented. Five rounds of reselection with lower L-histidine concentrations (5 mM) produced a further 100-fold increase in catalytic efficiency (step i, Fig. 10.39). Two isolated DNAzymes, 10.55 and 10.56 (Fig. 10.39), were fully characterized and found to have a strict requirement for L-His, as similar amino acids showed little or no cofactor activity apart from L-histidine methyl ester. The selection of cofactordependent ribozymes could increase the versatility and possible applications for catalytic ON sequences.

Ribozyme-catalyzed reactions involving C–C bond formations have also been reported. Seelig and Jaschke (233) presented the in vitro selection of ribozyme catalysts for the Diels–Alder reaction between maleimide and anthracene, employing a 2 × 1014-member library of 160-mer modified ONs (L28) with 120 randomized positions. The selection strategy used is shown in Fig. 10.40. Library L28 was prepared from the corresponding dsDNA sequences, and transcription initiation was performed in the presence of ternary complexes between guanosine monophosphate (10.57), PEG (10.58), and anthracene (10.59, step a, Fig. 10.40). The library obtained contained a 5′-anthracene–PEG appendage and was incubated with biotin-modified maleimide

10.60(25 µM) as the Diels–Alder substrate in the presence of various metal cations at 25°C for 1 h (step b). The reaction caused the attachment of the biotin-bearing adduct

10.61onto the active RNAs, and immobilization with streptavidin-functionalized agarose (step c) selected the Diels–Alder catalysts from L28 (Fig. 10.40). Their elution using standard protocols freed the selected ribozymes (step d, Fig. 10.40). Reverse

transcription, PCR amplification, and modified transcription (234) yielded another population of 5′-modified RNAs (step a, Fig. 10.41) that were submitted to a new round

of selection. The first five cycles were performed as above (step b), while stringency was then increased between the 6th and 10th cycles by reducing the concentration of

10.61to 2.5 µM and the incubation time to 1 min, step d. After round 10, the selected population was characterized, and 35 different sequences were obtained, among which

10.62(Fig. 10.41) showed an ~20,000-fold increase in efficiency with respect to the uncatalyzed Diels–Alder reaction. Further studies identified the minimal structural requirements for catalysis and allowed the preparation of several active truncated ribozymes (10.63–10.66, from 39to 57-mers, Fig. 10.41) that maintained a similar level of activity as the original ribozymes. The specificity for this reaction was high, as several related Diels–Alder substrates did not react with the ribozyme 10.66. The length of the tether between the anthracene and RNA was not important, as three PEGs with various lengths (7, 10, and 16 ethylene glycol units compared with the 13-unit construct originally employed) were inserted into L28 and gave similar reactions with

The results from this and another similar study (235) could be the precursors to the identification of novel, potent ribozymes able to catalyze organic reactions and to

552 BIOSYNTHETIC COMBINATORIAL LIBRARIES

a: reverse transcription, PCR amplification and transcription as in

standard protocols; b: four iterative cycles; c: five additional cycles with reduced concentrations of 10.60 (up to 2.5 M) and incubation times (up to 1').

BEST SELECTED INDIVIDUAL FROM L28 10.62

around 20,000-fold increase of catalytic efficiency compared to the uncatalyzed reaction

d

10.63-10.66

four truncated (39-mer to 57-mer) ribozymes with efficiency similar to 10.62

and high specificity for 10.61

d: truncation of ON sequences unnecessary for catalytic activity.

Figure 10.41 Selection of artificial ribozymes with catalytic properties for a Diels–Alder reaction from the biosynthetic ON modified ribozyme library L28: the selection/amplification process from 10.61 to the optimized ribozymes 10.63–10.66.

ested reader could further expand the knowledge of this area by consulting this relevant paper.

10.3 COMBINATORIAL BIOSYNTHESIS OF NATURAL PRODUCTS

10.3.1 General Considerations

Natural products (NPs) have long been a source of biologically active compounds, and their extraction and synthetic modification, especially for pharmaceutical purposes (238, 239), have been well-studied. Their structures exhibit a wide degree of diversity and levels of complexity that are seldom attained in totally synthetic structures, and this is reflected in the wide range of biological properties they display.

Recently, combinatorial technologies and libraries have somewhat replaced NPs as a source of diversity in pharmaceutical research. A comparison of combinatorial technologies with the fermentation of an NP-producing organism show that they both produce a library of compounds. While the former library is structurally determined a priori either by the selected synthetic scheme or by recombinant genetic information in the case of biosynthetic peptide and ON libraries, the latter is the result of the metabolic complexity of the producing strain and has to be deconvoluted in order to