Cycloisomerization of olefins in water

Abstract

Preparative reactions that occur efficiently under dilute, buffered, aqueous conditions in the presence of biomolecules find application in ligation, peptide synthesis, polynucleotide synthesis and sequencing. However, the identification of functional groups or reagents that are mutually reactive with one another, but unreactive with biopolymers and water, is challenging. Here we show that cobalt catalysts will react with alkenes under dilute, aqueous, buffered conditions and promote efficient cycloisomerization and formal Friedel-Crafts reactions. We find the constraining conditions of biorthogonal chemistry to be beneficial for reaction efficiency as we obtain superior conversion at low catalyst concentration and maintain competent rates in dilute conditions. Efficiency at high dilution in the presence of buffer and nucleobases suggests that these conditions may find broad application.

Graphical Abstract

Virtues we write in water. Cobalt-catalyzed cycloisomerization occurs with efficiency in aqueous buffer at high dilution. A variety of putative metal hydride atom transfer (MHAT) reactions, including deallylation, proceed under these stringent conditions and even on DNA.

Introduction

Organic reactions that are compatible with biomolecules and aqueous conditions find application in the labeling of macromolecules,¹ uncaging and delivery of metabolites to specific targets,² and metabolic engineering for the synthesis of small molecules.³ The application of standard reactions in these settings, however, is hampered by the instability of polar intermediates in aqueous environments, the sensitivity of transition metals to Lewis bases, or by overall slow kinetics exacerbated by dilute conditions.

Co(salen)-catalyzed radical hydrofunctionalization of olefins ⁴ has emerged as a useful method to functionalize unbiased alkenes with branch selectivity. Under mild conditions⁵ olefins can be appended with small functional groups or larger scaffolds by either direct radical addition or cross-coupling via a second transition metal.⁶ Both types of reactions increase sp³ character and tend not to affect other functional groups. Consequently, these reactions have enabled the functionalization of olefins in the presence of Lewis or Brønsted acids and bases with minimal to no loss in reactivity.⁷ A notable aspect of these reactions is their generally high rates and occasional tolerance of water co-solvent.^8,9,10,11

We previously reported the Co(salen) catalyzed isomerization/ cycloisomerization of alkenes under anhydrous conditions in non-polar solvent.^12,13 The reaction displayed a tolerance to Lewis basic functional groups, consistent with our proposed outer-sphere mechanism of metal-hydride hydrogen atom transfer (MHAT). We thought this tolerance might be useful for the synthesis of partially saturated heterocycles. Such cycloisomers benefit from fewer rotatable bonds and a higher fraction of sp³-hybridized atoms (Fsp³) compared to their linear, sp²-rich precursors.¹⁴

We were interested to explore whether this reaction could be rendered biocompatible, and thus open olefin hydrofunctionalization as a biorthogonal tool. Although carbon-centered radicals are unreactive with water, we did not know whether the putative metal hydrides of MHAT reactions would be sufficiently reactive under highly dilute, buffered, aqueous conditions, or active in the presence of biomolecules.

Here we report that 1) MHAT hydrofunctionalization is compatible with dilute aqueous conditions; 2) catalysis and cycloisomerization are actually optimal under these restrictive conditions; 3) MHAT reactions can occur in the presence of / on DNA with no degradation of the DNA and no apparent inhibition of the catalytic cycle; and 4) similar reaction conditions can be applied to other MHAT reactions under these conditions. Taken together, MHAT catalysts are shown to be additional tools for the community to develop novel biorthogonal applications.

Results and Discussion

Adaptation of our MHAT cycloisomerization conditions for use in biocompatible settings required us to address three major challenges: 1) the desired reaction would require fast kinetics relative to catalyst deactivation under dilute conditions, 2) cyclization (cycloisomerization) needed to outcompete other pathways such as linear isomerization, hydrogenation or hydration; and 3) the conditions would need to be compatible with aqueous, buffered solvent and with biomolecules.

Initial studies of feasibility and rate under dilute conditions indicated that although the reactions appeared to be slow at a 1 mM concentration (1 equiv. 1, 0.3 equiv. catalyst), the rate increased in acetone/ water mixtures (Figure 2).^15,16,17 Our lab previously reported a Co/Ni bimetallic hydroarylation of olefins, where Co(salen)BF₄ appeared to be superior to Co(salen)Cl.^18,19 We found that at high dilution Co(salen)BF₄ catalyzes cycloisomerization significantly faster than the initially reported Co(salen)Cl. Water, however, had a leveling effect on this cationic complex, possibly acting as an apical ligand, so that both complexes reached roughly equal rates (Figure 2).

Figure 2:

Kinetic trace of the cyclization reaction with Co(salen)X under different conditions at 1 mM. Solid dots represent acquired data. Lines included for better visualization.

Next we explored the effect of concentration on rate. As shown in Figure 3a, the initial rate decreased only 9-fold upon 50X dilution of all reagents (from 50 to 1 mM 1, Figure 3a). Consistent with these observations, the reaction reached completion faster under dilute conditions (Figure 3b), measured by product formation and substrate consumption. ²⁰ Two causes appeared responsible for this counterintuitive effect: competing pathways and catalyst decomposition. First, the intermediate carbon-centered radical has several pathways available (Figure 4). By analogy to the MHAT paradigm interrogated by Norton²¹ and others²² the radical (from Path 1) can react with [Co⁺²] to return a hydrogen atom adjacent to the radical (reverse of Path 1), cyclize onto a pendant acceptor—the desired pathway of cycloisomerization (Path 2), react with Co⁺² to afford an isomeric product (Path 3), or react with an equivalent of a [Co–H] to afford hydrogenation of the double bond. Unimolecular radical cyclization would not be affected by dilution, whereas bimolecular back-HAT would (Path 2 vs. 1, 3 or 4).

Figure 3:

Kinetic trace of the cyclization reaction at different overall concentrations. Traces overlayed with a logarithmic trendline for visualization.

Figure 4:

Pathways available for a radical after MHAT.

Additionally, higher concentration would favor catalyst decomposition through hydrogen evolution (2Co⁺³–H → 2Co⁺² + H₂), consistent with rapid rate decline at 50mM. Indeed, bubbling occurs at high concentration, and paramagnetic cobalt species are observed at the end of the reaction, frustrating NMR analysis of crude mixtures. Similar behavior has been observed in a Co-catalyzed olefin isomerization, where bimolecular hydrogen transfer between two Co species outcompetes a unimolecular pathway of isomerization at higher concentrations.²³

Given the prevalence of hydrogen evolution examples from similar cobalt hydride complexes,²⁴ and the calculated low BDE of iron hydrides that undergo MHAT with alkyl substituted olefins,²⁵ this background decomposition may be a general problem. Indeed, we have found N-fluorocollidinium oxidants to rescue off-pathway Co⁺² species via reoxidation to Co⁺³.¹⁸ Importantly, rates at high dilution (1 mM) allowed completion of the reaction within only 25 minutes. Initial rates indicated the bimolecular MHAT step (path 1, Figure 4) must possess a rate constant of at least 76 M⁻¹s⁻¹, corresponding to an energy barrier of at most 14.9 kcal/mol.²⁶ Since catalysis occurred at high rates under high dilution, we interrogated the selectivity for cyclization over competing pathways. The first three pathways might be suppressed by low concentration of the metal species, whereas unimolecular cyclization should be [Co] independent. Studying the reaction at different concentrations gave more weight to this hypothesis.

Reaction of 3 with catalytic Co(salen)BF₄ and stoichiometric silane at varying concentrations indicated an increase in the ratio of cycloisomerization to isomerization products as dilution increased. At a 100 mM concentration, linear isomerization was favored 1.2:1, whereas at 10 mM and at 1 mM cycloisomerization predominated in ratios of 3:1 and >10:1 respectively. It is noteworthy that for change in ratio to occur upon dilution, escape from the initial Co(II) / alkyl-radical cage should be faster than the rate of isomerization (i.e. fractional cage-recombination efficiency is probably low).²⁷ In other words, the relevant pathway for isomerization is bimolecular, inconsistent with reaction from the initial cage pair, despite a required movement of only a few angstroms and H• abstraction. Bimolecular catalyst death (hydrogen evolution) may produce Co(II) species capable of back-HAT, which reinforces this correlation between linear isomerization and concentration.

Notably, hydrogenation is a minor pathway under all conditions, despite the fast reaction of open-shell radicals towards hydrogen atom transfer²⁸ and the use of excess silane. This may be attributed to a kinetic barrier of polarity mismatch between the silane and the putative alkyl radicals,²⁹ or to a low concentration of the Co—H (or both). In summary, both relative rate of product formation and preference for on-pathway cycloisomerization improve under dilute, aqueous conditions.

Finally, the reaction proceeds in polar solvent (N-methyl-2-pyrrolidinone, NMP, entry 1; or others, Table 1), as well as NMP/water mixtures buffered with bis-(2-hydroxy-ethyl)-amino-tris(hydroxymethyl)-propane (Bis-Tris), a metal-binding aminopolyol. Even in equivolume mixtures of NMP and Bis-Tris buffer solutions, the reaction proceeded to completion at pH 7. Deviations at pH 9 and 5, however, prevented reaction.

Table 1:

Yield of cyclization under different conditions.

Entry	Concentration (NMP / Buffer ratio)	8 (%)a
1	0.1 M (1 / 0)	76
2b	0.1 M (4 / 1)	72
3b	0.001 M (4 / 1)	70
4 c,d	0.001 M (4 / 1), 30 mol% of [Co] and 3 equiv. of PhSiH3, Bis-Tris Buffer pH 7	72
5 c,d	0.001 M (1 / 1), 30 mol% of [Co] and 3 equiv. of PhSiH3, Bis-Tris Buffer pH 7	71
6 c,d	0.001 M (1 / 1), 30 mol% of [Co] and 3 equiv. of PhSiH3, Bis-Tris Buffer pH 9	<10
7c,d	0.001 M (1 / 1), 30 mol% of [Co] and 3 equiv. of PhSiH3, Citric acid Buffer pH 5	–

^aYield determined by LCMS or NMR with an internal standard. 200 mM Bis-Tris Buffer pH adjusted with HCl.

^bDeionized water was used instead of the buffer.

^c3 equiv. of PhSiH₃ instead of 1 equiv.

^d30 mol% of Co(salen)Cl instead of 10 mol%.

A variety of tertiary amine-containing substrates were investigated under these conditions to study the scope of the transformation in aqueous co-solvent, since a variety of semisaturated heterocycles of varying substitution patterns could be constructed (Figure 6, 9–28). One such example is the tetrahydroisoquinoline scaffold, considered a privileged scaffold in medicinal chemistry.³⁰

Figure 6:

Scope under dilute aqueous conditions.^aReaction from the internal olefin. ^bMajor regiosiomer. See Supporting Information for regioselectivities in organic and aqueous conditions. ^cOnly PhH as the solvent. Yields at 1mM were determined by ¹H NMR using an internal standard or by LCMS using a calibrated internal standard. Boc = t-butyloxycarbonyl; Bn = benzyl

Formation of 5- and 6-membered rings proved facile, but seven membered rings did not cyclize.³¹ Electron-rich (22, 23) and electron poor arenes (17–21) were engaged, but excessively electron rich substrates, such as trimethoxybenzene derivative 28, performed poorly. Various heterocycles could be employed; substrates that could generate mixtures of regioisomers tended to react at the most electron deficient site (18–20). Imidazole derivatives like 20 have raised interest as bacterial gyrase inhibitors,³² and furan-tetrahydroisoquinolines such as 22 show activity against psychological disorders.³³

Although the reaction performed optimally under dilute, aqueous conditions, two questions remained unanswered: would the reaction function once appended to biomolecules and would these reaction conditions compromise the integrity of a biomolecule? This last point is crucial, given reports that similar cobalt complexes can degrade nucleoside strands (as is observed for radical carriers).³⁴

This project evolved from exploration of DNA-encoded library (DEL)-compatible reactions, so we explored cycloisomerization in the presence of a relevant DNA fragment. Addition of acetylated (5′-/5Phos/GAGTCA/iSp9/iUni AmM/iSp9/TGACTCCC-3′) (32)³⁵ to the conditions for MHAT cycloisomerization resulted in no change in the yield of product formation, consistent with the high functional group tolerance of MHAT radical reactions. Quantitative analysis of the reaction components on DNA amplifiability³⁶ revealed no significant degradation of the DNA, demonstrating reaction compatibility. This assay exposes dsDNA-functionalized magnetic beads (sensor beads) to the subject reaction conditions, followed by separation of the beads, and quantitation of the PCR-amplifiable DNA for comparison with suitable positive and negative controls. Addition of the sensor beads to a 20% trifluoroacetic acid / dichloromethane solution served as a positive control for DNA degradation. These conditions caused extensive degradation within 15 min corresponding to only a 6% recovery (positive control), whereas overnight storage of the DNA strand in aqueous Bis-Tris buffer at room temperature led to complete DNA recovery (negative control). The stability of DNA to high volume fractions of NMP was tested to probe the limits of organic solvent tolerance. Most of the DNA (80%) was recovered after standing overnight, well within the standard guidelines for DEL which suggest more than 30% of the DNA should be intact for amplification by PCR.

Finally, the effect of MHAT cycloisomerization on DNA was analyzed using 4:1 NMP/ aqueous Bis-Tris buffer, conditions that helped dissolve the Co(salen)Cl precatalyst. Nearly the same DNA recovery as NMP solvent alone was achieved. PCR amplification of the remaining DNA indicated that after running our reaction, at least 77% of the DNA remained intact. We concluded that the DNA tag and the conditions of MHAT cycloisomerization—Co⁺³ precatalyst, phenylsilane reductant, the intermediacy of carbon-centered radicals—were mutually compatible. We did not know, however, if the reaction would still occur on DNA, a potential problem given that MHAT cycloisomerization involves no change to mass, frustrating analysis.

Attempts to detect retention time shifts were unsuccessful. A ¹³C label on the alkene was explored to differentiate cycloisomer from byproduct, but high dilution and small scale of the DEL protocol led to insufficient NMR signal-to-noise ratios. Instead, we pursued a cleavable linker strategy that would allow quality control sampling of any given reaction. Cleavage of the scaffold from DNA would allow us to monitor the conversion and distribution of products by LCMS.

High-yielding ester scaffold 3 provided the best LC separation among starting alkene, cycloisomer and byproducts like linear isomer and hydrate. An α-hydroxyl linker could be easily installed and cleaved using basic conditions. We were pleased to find that the reaction performed well on DNA and produced the cycloisomer as the major product (Figure 6). Combined with its preservation of DNA in the headpiece, we concluded that this cobalt catalyzed reaction is competent under the demanding conditions required for DEL library construction and may extend to similarly stringent applications. Its general applicability stems from the competency of multiple related reactions.

Several experiments involving intermolecular couplings were undertaken to probe feasibility for related applications (Figure 7). The related reactions of hydrothiolation (33), imine coupling (34), and the alkylation of olefins with electron-deficient alkenes (35) were successful. All reactions occurred under aqueous conditions with moderate yields and outcompeted undesired pathways such as imine hydrolysis. In a potentially useful deprotection reaction, amides (36) could be deallylated via isomerization to afford the secondary amide (37). Allyl deprotections that are compatible with aqueous systems have found application in, for example, ³⁷ DNA sequencing with chain-terminating nucleotides.

Figure 7:

Intermolecular examples and deallylation of amides. Yield determined by ¹H NMR of the crude reaction mixture. For procedures, see Supporting Information. ^aReaction ran at 1 mM. ^bReaction ran at 10 mM. SOAr = (R)-mesitylsulfinyl; Cy = cyclohexyl

Conclusion

We have shown that Co(salen)⁺³-catalyzed cycloisomerizations and related reactions—thiolation, imine addition, Giese addition—function under the dilute, aqueous, buffered conditions required for biocompatible reactions, a challenge even for simple allyl deprotections. The cycloisomerization reaction appears to work even more efficiently under these stringent conditions. The likelihood that each reaction reported here involves a metal hydride atom transfer elementary step is significant since, to the best of our knowledge, this fundamental mode of reactivity has not been demonstrated in biorthogonal or DNA-related chemistry. The methods presented here may prove useful for the synthesis of partially saturated heterocycles as well as the diversification of partially unsaturated scaffolds. More generally, the feasibility of an additional fundamental organometallic elementary step under dilute, buffered conditions in the presence of complex biomolecules promises to deliver many useful applications.

Figure 1:

MHAT hydrofunctionalization as a biocompatible reaction.

Table 2:

Reactions in the presence of DNA and DNA crash test.

Entry	Incubation Conditions	Remaining DNA (%)a
1	20% TFA/DCM (v/v), 15 min, rt	6
2	Bis-Tris Buffer, ON, rt	100
3	80% NMP (v/v) in Bis-Tris Buffer, ON, rt	80
4	Standard HAT conditions, ON, rt	77

Figure 5.

Effect of reaction concentration on the yield of the cyclic isomer vs. linear isomer. ^aYield determined by ¹H NMR of the crude reaction mixture. ND = Not Determined.