An Assessment of Photoreleasable Linkers and Light Capturing Antennas on a Photoresponsive Cobalamin Scaffold

Abstract

Cobalamin has shown promise as a light-sensitive drug delivery platform owing to its ease of modification and the high quantum yields for drug photorelease. However, studies to date on the general photochemistry of alkyl-cobalamins have primarily focused on methyl and adenosyl-substituted derivatives, the natural cofactors present in various enzymatic species. We describe the synthesis and photolytic behavior of cobalamin conjugates comprised of different combinations of fluorophores and β-axial ligands. In general, cobalamin conjugates containing β-axial alkyl substituents undergo efficient photolysis under aqueous conditions, with quantum yields up to >40%. However, substituents that are large and hydrophobic, or unable to readily support the presumed radical intermediate, suffer less efficient photolysis (<15%) than smaller, water soluble, analogs. By contrast, quantum yields improve by two-fold in DMF for cobalamins containing large hydrophobic ß-axial substituents. This suggests that drug release from carriers comprised of membranous compartments, such as liposomes, may be significantly more efficient than the corresponding photorelease in an aqueous environment. Finally, we explored the impact of fluorophores on the photolysis of alkyl-cobalamins under tissue mimetic conditions. Cobalamins substituted with efficient photon-capturing fluorophores display up to 4-fold enhancements in photolysis relative to unsubstituted derivatives. In summary, we’ve shown that the photosensitivity of alkyl-cobalamin conjugates can be tuned by altering the Co-appended alkyl moiety, modulating the polarity of the environment (solvent), and installing photon-capturing fluorophores onto the cobalamin framework.

Graphical Abstract

INTRODUCTION

Light-responsive agents have found application in a variety of areas, including polymer synthesis,¹ sensor design,² manipulation of intracellular biochemistry,^{3, 4} and drug delivery^5-7. For example, light triggered drug delivery offers the promise of diseased site targeting and consequent reduced side-effects.^{6, 8, 9} Nitrobenzyl,^{10, 11} hydroxyphenacyl,^{11, 12} coumarin-4-yl,^{11, 12} and a wide assortment of other derivatives¹³ have long served in the capacity of light-absorbing/photorelease scaffolds. However, within the last decade, attention has shifted to the design of molecular scaffolds that absorb long wavelength visible and near infrared light, which has the advantage of deeper tissue penetration than that of shorter wavelengths.^{5, 14} These scaffolds include BODIPY analogs,^15-18 cyanine derivatives,^19-23 [Ru(bpy)₃]³⁺ complexes,^24-26 and others.^{5, 14} In this regard, we have found that cobalamin (Cbl; vitamin B12) serves as an unusually flexible molecular launch pad for therapeutic agents that can be tuned to wavelengths in the red, far red, and near IR.²⁷ Alkyl substituents attached to Co of Cbl have long been known to be photolabile at wavelengths absorbed by the corrin ring of Cbl (i.e., 360 - 550 nm).²⁸ Furthermore, a variety of drugs have been covalently appended to short linkers attached to the corrin ring-constrained Co atom and subsequently photoreleased.^29-31 However, the range of photoreleasable linkers attached to Co has not been methodically investigated. The linkers assessed to date primarily include members of the standard methyl, ethyl, and propyl series^32-38 and, in the case of phototherapeutics,^29-31 ethyl and propyl linkers in Drug-CO-HNCH₂CH₂-Co and Drug-NH-COCH₂CH₂CH₂-Co, respectively (Figure 1). In addition, although alkyl-Cbls undergo photolysis at the short wavelengths (<550 nm) absorbed by the corrin ring system, the extinction coefficients are modest (ε < 10,000 M⁻¹ cm⁻¹). This renders the photocleavage of ligands appended to Co, dependent upon a robust photon flux, which can be challenging to achieve with tissue embedded phototherapeutics. Consequently, we have investigated whether the photosensitivity of the C-Co bond is affected by the nature of the alkyl substituent by preparing a small structurally diverse library of alkyl-Cbl derivatives. In addition, we have explored whether the photosensitivity of the C-Co bond is enhanced via appended photon capturing antennas, thereby promoting photocleavage under otherwise challenging tissue embedded conditions.

Figure 1.

Structure and schematic of Cbl (R = CN) and the various alkylated derivatives (R = alkyl) employed in this study. The highlighted ribose hydroxyl moiety is used as an attachment site for fluorophores. Photolysis of alkyl-Cbls (1) generates H₂O-Cbl (2) and the photoreleased alkyl substituent.

EXPERIMENTAL SECTION

General Synthesis of Alkyl-Cbl conjugates (Scheme S-1).

Alkylations were performed in a low light environment, and alkyl-Cbl conjugates were always handled under low light conditions (Scheme S1). For each alkylation, 50 mg (0.036 mmol) of cyanocobalamin was dissolved in 5 mL MeOH in a 50 mL plastic conical centrifuge tube to give a red solution. To this solution was added 250 mg (5% w/v) powdered NH₄Br. The solution was then vortexed until the NH₄Br was completely dissolved. The tube was covered in a fitted aluminum foil sleeve to protect the contents from ambient light. Next, 231 mg (3.6 mmol) of Zn powder was added to the solution and the tube was shaken by hand for 30 s, then placed on a shake plate for 30 min. Reduction of the Cbl from Co^III to the supernucleophilic Co^I state caused the solution to change from red to dark gray/black. To the reduced solution, 4 equivalents of the alkyl halide were added, and the solution was allowed to shake an additional 4 h. The conical tube was centrifuged at high speed in a tabletop centrifuge for 4 min to pellet the Zn, and the supernatant was decanted. Successful alkylation was indicated by a solution with a translucent orange color. The methanolic reaction mixture was filtered through a Sfär 12g column sacrificial loading segment that had been conditioned with 100% MeOH, then diluted with deionized (DI) water (to 40% methanol). The solution was loaded onto a Sfär reverse phase 12 g column containing a sacrificial loading section forming a thin red band near the top of the section. The column was conditioned with a 4:1 ratio of DI water: MeOH with 0.1% TFA. The conjugate was then purified using a H₂O:MeOH gradient from 9:1 H₂O:MeOH to 2:8 H₂O:MeOH on a Biotage Isolera One 3.0. Fractions were collected in 5 mL aliquots, based on their absorption at 360 nm as measured by the Biotage. Fractions were analyzed for purity on an LC-MS, and pure fractions were combined and concentrated under reduced pressure. The concentrated solution was frozen in a 50 mL plastic centrifuge tube and lyophilized for 3 days to remove all remaining solvent. The tube was then weighed to determine amount of Alkyl-Cbl product and yield.

Photolysis of Alkyl-Cbl Conjugates

Photolysis of the alkyl-Cbl derivatives was performed at 360 nm, which falls within the wavelength range most commonly employed for photolytic studies. The absorbance at 360 nm (27290 M⁻¹ cm⁻¹)³⁹ is significantly greater than that at the longer wavelengths [440 nm (3039 M⁻¹ cm⁻¹) and 550 nm (8739 M⁻¹ cm⁻¹)]³⁹ that also elicit photocleavage.⁴⁰ We, as well as others,^{38, 41} have found that the absorbance of alkyl-Cbl derivatives are nearly identical in the UV range. The LED light source (360 nm) continuously illuminated the sample as the spectrometer rapidly scanned between 325 nm and 380 nm every 6 s for a total of 30 scans. Photolysis experiments were performed at room temperature (23 – 25 °C). During the photolysis there was a growth of a new feature with an absorption maximum of 352 nm. Some complexes displayed two isosbestic points of 337 nm and 369 nm. Single wavelength kinetics were evaluated at 352 nm and normalized traces could be satisfactorily fit with a biexponential function in equation 1. To compare kinetics between complexes with a single “effective” lifetime, a weighted average lifetime ⟨τ⟩ was calculated using equation 2. Three trials were performed to attain an average effective lifetime and deviation from the mean. The rate constant k was calculated as the inverse of ⟨τ⟩.

$$Ab{s}_{352nm}(t)=A_1{e}^{{}^t∕_{{\tau }_1}}+A_2{e}^{{}^t∕_{{\tau }_2}}$$ (1)

$$\langle \tau \rangle =\frac{A_1{\tau }_1+A_2{\tau }_2}{A_1+A_2}$$ (2)

Synthesis of the Ethyl-Cbl-Linker-Fluorophore Conjugates (Scheme S2)

Cbl-Linker-Fmoc Conjugates.

Cyanocobalamin (1 g, 0.7 mmol) was dissolved in 10 mL dry DMSO in a 100 mL round bottom flask to form a dark red solution. To the solution was added 2.8 mmol of 1,1-carbonyldiiimidazole to activate the 5’ ribose hydroxyl group of the nucleotide loop. The solution was allowed to stir for 4 h, at which time 5.8 mmol of the diamine linker (EDA or DACH) was added to the solution. The reaction mixture was stirred for an additional 4 h. Upon LC-MS confirmation of the formation of the desired amide bond, the solution was separated into two 5 mL aliquots in 50 mL plastic centrifuge tubes. The DMSO-containing aliquots were diluted to 50 mL with a 2:1 solution of diethyl ether:DCM, which induced precipitation of the cobalamin derivative. The tubes were centrifuged for 4 min to pellet the modified cobalamin, decanted and allowed to dry. The pellets were dissolved in a 10 mL dry DMF in a 50 mL conical centrifuge tube. 982 mg (3 mmol) 9-fluorenylmethyl N-succinimidyl carbonate (Fmoc-OSu) was added to the solution and stirred for 4 h. The solution was split into two 5 mL aliquots and both aliquots were diluted to 50 mL with diethyl ether. The tubes were centrifuged to pellet the Cbl-linker-Fmoc and decanted, leaving red pellets. The pellets were allowed to dry, dissolved in 10 mL MeOH each, and re-centrifuged. Unreacted Fmoc-OSu and trace cobalamin formed light pink pellets. The MeOH supernatants containing the majority of the Cbl conjugate were decanted, combined, and diluted to 40% MeOH with DI water. The Cbl-linker-Fmoc conjugates were purified using reverse phase C18 ultra column from Biotage. Products were purified using a MeOH:H₂O (0.1%TFA) gradient from 20 to 80% MeOH. Fractions were analyzed for purity via LC-MS and combined, concentrated under reduced pressure, and lyophilized to yield powdered red solids. CN-Cbl-EDA-Fmoc (yield: 65.8%) and CN-Cbl-DACH-Fmoc (yield: 62.8%).

Ethyl-Cbl-linker-Fmoc Conjugates.

CN-Cbl-linker-Fmoc (0.06 mmol) was dissolved 10 mL MeOH in a 50 mL conical centrifuge tube. NH₄Br (0.5 g, 5% w/v) was added to the solution and vortexed to dissolve. Zinc (0.006 mol) was added to the tube, and it was shaken for 30 sec then put on a shake plate for 30 min. After 30 min, the solution turned black, indicating reduction of the cobalt. Iodoethane (0.24 mmol) as added the solution, and the tube was returned to the shake plate for a further 4 h. The solution was centrifuged to pellet the zinc, then decanted. LC-MS showed quantitative conversion to the ethyl-Cbl-linker-Fmoc conjugate. The solution was diluted to 50 mL with diethyl ether and centrifuged to precipitate the cobalamin conjugate. The supernatant was decanted, and the pellet was allowed to dry overnight. The pellet was dissolved in 10 mL dry DMF in a 50 mL centrifuge tube, and 0.42 mmol piperidine was added. The solution was placed on a shake plate for 4 h. After 4 h, LC-MS showed deprotection of the Fmoc group, so the reaction solution was diluted with diethyl ether and centrifuged to pellet the cobalamin conjugate. The resulting pellet was allowed to dry for 1 h, then dissolved in 10 mL MeOH and diluted to 40 mL with H₂O. The conjugate was loaded onto a Sfär reverse phase 30 g column from Biotage and purified with a MeOH:H₂O gradient from 5 to 80% MeOH with 0.1% TFA. Aliquots were collected based on absorption at 360 nm. Purity of the fractions was assessed via LC-MS. Pure fractions were combined and concentrated under reduced atmosphere. The ethyl-Cbl-linker conjugates were frozen and remaining solvent was removed via lyophilization.

Ethyl-Cbl-linker-Fluorophore Conjugates.

The BODIPY 505/515 variant and Cy3 were synthesized as previously described.^{42, 43} In 5 mL dry DMF, 0.019 mmol of the fluorophore (TAMRA, BODIPY, or Cy3) was dissolved in a 50 mL plastic centrifuge tube. To the solution was added 0.141 mmol (18.2 mg) DIPEA and 0.018 mmol (6.7 mg) HATU. The solution was allowed to incubate for 5 min to allow for activation of the carboxylic acid on the fluorophore. Under low light conditions, 0.017 mmol of the ethyl-Cbl-linker-Fmoc conjugate was added to the solution. The reaction was placed on a shake plate for 2 h. Upon confirmation of product formation via LC-MS, the reaction was diluted to 50 mL with diethyl ether and centrifuged to pellet the cobalamin conjugate. The supernatant was decanted, the pellet allowed to dry, subsequently dissolved in 5 mL MeOH, and diluted to 15 mL with DI water. The solution was loaded onto a Sfär reverse phase 12 g column from Biotage and purified with a MeOH:H₂O gradient from 20 to 80% MeOH with 0.1% TFA. Aliquots were collected based on absorption at 360 nm while absorption at the longer wavelength (e.g., TAMRA at 550 nm) was monitored to identify the fluorophore conjugated Cbl. Purity of the fractions was assessed via LC-MS. Pure fractions were combined and concentrated under reduced atmosphere. The ethyl-Cbl-linker-fluorophore conjugates were frozen and remaining solvent was removed via lyophilization.

Photolysis of Ethyl-Cbl-Linker-Fluorophore Conjugates

Photolysis of ethyl-Cbl-linker-fluorophore conjugates (40 μM) was carried out in 1X PBS at pH 7.0. Illumination was performed using an Oriel Xe flash as the light source. Optical filters purchased from Newport Inc. were used to block out unwanted wavelengths to accurately assess the impact of the fluorophore antennas on photolysis rate. Filters used were 505 ± 5 nm and 550 ± 5 nm. The lamp bulb had a reported wattage of 60 W, and the output of the lamp at the above wavelengths was approximately 0.025 mW. Samples were illuminated for 5, 10, 15, then 30 min intervals. After each time point, 40 μL of the ethyl-Cbl-linker-fluorophore solution was withdrawn, diluted with 40 μL of DMF, and injected into a 384 well plate for automated LC-MS analysis. The same sample size (40 μL) was injected for all samples and care was taken to ensure Photolysis of the ethyl-Co bond generated (H₂O)-Cbl-linker-fluorophore, which was well separated from the unphotolyzed starting material on reverse phase LC-MS using an H₂O:MeOH with 0.1% formic acid gradient. The extent of photolysis was quantified by assessing the area under the peak of the photolyzed (P) and unphotolyzed (U) Cbl.

$$\frac{P}{U+P}=\%photolyzedconjugate$$ (3)

Ethyl-Cbl and ethyl-Cbl-EDA-TAMRA photolysis rates were further compared via illumination through phantom solutions mimicking the three skin tones represented by the Fitzpatrick scale.⁴⁴ Eumelanin samples of various concentrations (0.0088 mg/mL, 0.066 mg/mL, and 0.13 mg/mL) were used to mimic different skin colors, from lightest to darkest, according to the Fitzpatrick scale (I-II, III-IV, and V-VI).⁴⁵ In addition, the phantom solutions contain 3% w/v of intralipid, which mimics the light scattering properties of skin. The phantom solutions (5 mL) were placed in a 2 cm long quartz cuvette, which was inserted between the Xe lamp light source and the cuvette containing the photosensitive cobalamin solution. The light source was placed in a cardboard housing with a cutout to ensure that all light from the light source was filtered through the phantom solution. Photolyzed Ethyl-Cbl and ethyl-Cbl-EDA-TAMRA solutions were sampled at 0, 10, 20, and 30 min and analyzed using the same methodology described above.

Potential Hazards.

Eye protection is required when working with high intensity light sources.

RESULTS AND DISCUSSION

Synthesis and Photolytic Behavior of Alkyl-Cbls.

We prepared a series of alkyl-Cbl conjugates, including linear, cyclic, and heteroatom-containing derivatives (Chart 1). All of these conjugates were synthesized by exposing cyano-Cbl to Zn in methanol, which converts the starting Co^III species to the supernucleophilic Co^I state. The latter was subsequently treated with 4 equivalents of the appropriate alkylhalide at room temperature for 4 h to furnish the desired alkyl-Cbls upon workup and purification. Yields, which were not optimized, range from a low of 23% (propyl-Cbl) to a high of 60% (5-amino-5-oxopenyl-Cbl). The alkyl-Cbl derivatives are protected from light during synthesis, purification, and subsequent storage (see experimental details).

Chart 1.

Structures of alkyl-Cbls, including acyclic (3 – 7) and cyclic hydrocarbons (8 – 12), as well as heteroatom containing substituents (13 – 18).

The photolysis rates, quantum yields, and calculated octanol-water partition coefficients (K_OW) are furnished in Table 1 for compounds 3 – 18. The quantum yield (0.35) for methyl-Cbl has been previously reported³⁸ and served as the actinometer for the alkyl-Cbls in Table 1. Photolysis was performed in a spectrophotometer outfitted with an LED (5 mW at 360 nm) that is oriented perpendicular to the detector to avoid interference with the Cbl spectrum (325 – 380 nm), which was recorded every 5 s (see Figure 2 and supporting information). An increase in the absorbance at 352 nm (γ-band of H₂O-Cbl byproduct) was used to monitor the progress of photolysis. Using the biexponential fit of the change in absorbance, rate constants (k) were obtained for each compound. Single exponential fits were possible for some compounds, but double exponential fits were more suitable for many of the conjugates, so a biexponential fit was used for all alkyl-Cbl conjugates for the sake of consistency. Sension and colleagues also reported the use of multiexponential fits in their calculations of the quantum yields for ultrafast measurements of propyl- and ethyl-Cbl photolysis.⁴⁶ The competition between radical escape and recombination is one possible reason that single exponential fits are often unsuitable for alkyl-Cbl photolysis fitting. However, complications associated with longer time scale aerobic photolysis (e.g., formation of a R-O-O-Co intermediate) may also be responsible for the multiexponential behavior.⁴⁷ We determined quantum yields by comparing the ratio of rate constant and quantum yield of methyl-Cbl to the rate constant of the other alkyl-Cbl conjugates.

Table 1.

Photolysis (5 mW LED at 365 nm) of alkyl-Cbls in phosphate buffered saline.

	ß-axial Substituent	K OW a	k (s−1)b	Φc
3	methyl	1.09	0.10 ± 0.01	0.35 ± 0.02
4	ethyl	1.33	0.13 ± 0.02	0.44 ± 0.05
5	propyl	1.75	0.087 ± 0.006	0.31 ± 0.02
6	isobutyl	2.08	0.084 ± 0.007	0.29 ± 0.02
7	decyl	4.67	0.04 ± 0.02	0.14 ± 0.06
8	cyclopropyl	1.25	0.026 ± 0.003	0.09 ± 0.01
9	cyclopropylmethyl	1.58	0.099 ± 0.004	0.34 ± 0.01
10	cyclobutyl	1.67	0.064 ± 0.006	0.22 ± 0.02
11	cyclobutylmethyl	2.00	0.083 ± 0.006	0.29 ± 0.01
12	cyclohexylmethyl	2.83	0.045 ± 0.004	0.16 ± 0.01
13	1-aminopropyl	0.17	0.079 ± 0.002	0.27 ± 0.01
14	1-carboxyethyl	0.35	0.10 ± 0.01	0.34 ± 0.05
15	1-amino-1-carboxypropyl	-0.41	0.066 ± 0.005	0.23 ± 0.02
16	1-methoxy-1-oxopentyl	1.44	0.074 ± 0.008	0.26 ± 0.03
17	1-carboxypentyl	1.60	0.07 ± 0.01	0.23 ± 0.04
18	1-amino-1-oxohexyl	0.95	0.07 ± 0.01	0.23 ± 0.04

^aThe octanol water partition coefficient was calculated using ChemDraw 20.0.

^bPhotolysis rate from the average of three trials per conjugate.

^cThe quantum yields were calculated using methyl-Cbl as the actinometer.³⁸

Figure 2.

Representative photolysis of alkyl-Cbl derivatives. (A) Photolysis of cyclopropyl-Cbl (8) was performed using a 5 mW LED (360 nm) and assessed by the change in absorbance at 352 nm, which was measured at 5 s intervals. (B) The rate of photolysis was quantified in triplicate. The change in absorbance at 352 nm for each photolysis experiment was fit to the biexponential function given in equation 1.

The acyclic hydrocarbon ligands 4 – 7 display a decrease in quantum yield with increasing size and increasing K_ow. As the partition coefficient between octanol and water, K_ow gives the ratio of the concentration of a compound in octanol to the concentration of the compound in water. A large K_ow indicates a compound will be found in higher concentration in the octanol than in the water, generally indicating hydrophobicity. Based on our results, it is tempting ascribe the observed inverse relationship between quantum yield and alkyl size and/or lipophilicity to the ability of the alkyl radical and Cbl(Co^II) pair to become solvent separated.⁴⁸ For example, the comparatively large and lipophilic decyl radical likely remains proximal to the corrin ring system due to its reduced ability to diffuse into the water network. Close proximity to the Cbl(Co^II) should promote recombination, nullifying photolysis, and reducing the effective quantum yield. Indeed, Sension and her colleagues employed ps-resolved transient spectroscopy to demonstrate that the rate of escape of the solvent caged alkyl radical correlates with size, notably methyl radical > ethyl radical > propyl radical > adenosyl radical.^{41, 46} For example, in this study, the quantum yield of ethyl-Cbl photolysis is nearly four times greater than that of decyl-Cbl. We do note that, in terms of quantum yield, methyl-Cbl is an outlier in the observed inverse trend with increasing alkyl chain size/lipophilicity (Table 1). This may be due to the lower stability of the methyl radical relative to its ethyl and longer chain counterparts,⁴⁹ or a higher barrier of methyl-Cbl to photolysis^{46, 48}.

In addition to linear hydrocarbons, we explored the photolysis of several cyclic hydrocarbon containing Cbl conjugates, including the cyclopropyl (8) and cyclobutyl (10) derivatives, in which the ring is directly appended to the Co. Furthermore, we synthesized the corresponding cyclopropylmethyl (8), cyclobutylmethyl (11), and methylcyclohexyl (12) analogs to assess whether the methylene bridge has any influence on photolysis rate of these cyclic species. The photolytic quantum yield for cyclopropyl-Cbl is modest (0.09) relative to other alkyl-Cbl derivatives [e.g., cyclobutyl-Cbl (0.22)]. In addition, the corresponding methylcycloalkyl derivatives 9, 11, and 12 are all more efficiently photolyzed than cyclopropyl-Cbl. Furthermore, the trend for 9, 11, and 12 is analogous to that seen with the linear alkyl series (4 - 7) namely, with increasing size, there is an increase in both K_OW and a decrease in the quantum yield of photolysis. The poor quantum yield associated with cyclopropyl-Cbl photolysis is consistent with the significant ring strain inherent in the cyclopropyl radical [radical stabilization energy (RSE) = 5.2 kcal/mol] relative to that of other cyclic derivatives, including the cyclobutyl radical (RSE = −3.7 kcal/mol).⁴⁹ However, other factors may contribute to the reduced photosensitivity of cyclopropyl-Cbl. In particular, the behavior of phenyl-Cbl and ethynyl-Cbl derivatives may be informative. The former is dramatically less photolytically sensitive (quantum yield < 1%) than alkyl-Cbls and the latter is completely photostable.⁵⁰ Ethynyl-Cbl photo-resistance has been ascribed to the significant S character (sp) associated with the carbon attached to Co as well as to the interaction between the ethynyl and corrin π systems (which shortens and strengthens the C-Co bond).⁵¹ Intriguingly, in the cyclopropyl case, the C-H bonds (and presumably the C-Co bond) likewise possess substantial S character (sp²),⁵² whereas the cyclopropyl C-C bonds display double bond-like behavior⁵³.

The relatively low quantum yields for the cyclopropyl- and decyl-Cbl derivatives are of potential therapeutic utility since it may be desirable to temper the light sensitivity of phototherapeutics for certain indications (e.g., surface versus deep tissue therapies). Ultimately, it may be possible to further modulate photosensitivity by combining different features (e.g., hybridization and lipophilicity) to create a series of differentially light responsive constructs. Finally, we note that ligand substitution at the Co is not the only means to modulate the photochemistry of alkyl-Cbls. In particular, several studies have demonstrated that the photoreactivity of adenosyl-Cbl is altered when bound to the photoreceptor CarH.^54-56

In addition to alkyl and cycloalkyl moieties, we examined a series of β-axial ligands that contain various heteroatoms with a K_OW range that extends from strongly water soluble (K_OW = −0.41) to a significant octanol preference (K_OW = 2.46). Functional groups include carboxylate (14, 15, 17), amine (13, 15), ester (16), and amide (18) derivatives. In contrast to the hydrocarbon ligands, substituents containing oxygen, nitrogen and fluorine atoms display quantum yields that lie within a relatively tight range (0.23 – 0.34), and do not approach the extremes exhibited by certain hydrocarbon ligands. These quantum yields are smaller than that displayed by ethyl-Cbl (0.44), which indicates that enhanced hydrophilicity alone is insufficient for promoting photolytic release similar to or greater than that of ethyl-Cbl. Rather, derivatives 14 – 18 are of a similar size, which appears to support the narrative that the size of the light-generated alkyl radical plays an important role in determining the overall quantum yield.⁴⁸

Phototherapeutics embedded within non-aqueous regions of water-soluble carriers (e.g., liposomes) offer a number of potential therapeutic advantages relative to freely circulating light-responsive drugs.⁵ Consequently, we examined the photolysis of several of the alkyl-Cbl conjugates in DMF in order to assess the impact of a non-aqueous solvent on photolysis rates (Table 2). The conjugates analyzed include aliphatic acyclic (4, 7) and cyclic derivatives (8, 12), as well as amino acid (15) and fluorinated (16) analogs. The ethyl-Cbl conjugate behaves similarly in aqueous and organic media, likely due to the relatively simple ethyl ligand having few interactions with the surrounding solvent in either system. By contrast, the quantum yield associated with decyl-Cbl photolysis is dramatically enhanced in DMF (0.29) relative to that in aqueous solution (0.14). This difference supports the notion that photolyzed hydrophobic ligands are slow to escape from the aqueous solvent cage that encompass the initial alkyl radical and Cbl(Co^II) photoproducts. This trend holds for cyclohexylmethyl-Cbl (12) as well, which displays a quantum yield of 0.40 in DMF compared to 0.16 in aqueous solution. These results suggest that carriers comprised of membranous compartments, such as liposomes, may promote enhanced photolytic quantum yields for lipophilic ligands. Finally, the quantum yield of the cyclopropyl derivative in DMF is somewhat larger than that in water but remains modest compared to other hydrocarbon-based ligands.

Table 2.

Photolysis (5 mW LED at 365 nm) of alkyl-Cbls in DMF.

	ß-axial Substituent	K OW a	k (s−1)b	Φc
4	ethyl	1.33	0.14 ± 0.02	0.46 ± 0.08
7	decyl	4.67	0.09 ± 0.02	0.29 ± 0.05
8	cyclopropyl	1.25	0.057 ± 0.009	0.19 ± 0.03
12	cyclohexylmethyl	2.83	0.12 ± 0.01	0.40 ± 0.01
15	1-amino-1-carboxypropyl	−0.41	0.036 ± 0.005	0.12 ± 0.02

^aThe octanol water partition coefficient was calculated using ChemDraw 20.0.

^bInitial rate of photolysis from the average of three photolysis trials per conjugate.

^cThe quantum yields were calculated using methyl-Cbl as the actinometer.³⁸

The photolysis of Cbl conjugates containing ligands with N, O, and/or F was also examined in DMF. In contrast to the ligands composed solely of hydrocarbons, the Co-substituents containing heteroatoms display lower quantum yields in DMF than in aqueous solution. For example, amino acid substituted Cbl 15 exhibits a quantum yield of 0.12 in DMF, approximately half of its aqueous quantum yield (0.23). The results imply that the aqueous environment assists with the solvent separation of the initially formed photoproducts more readily than that of DMF.

Our results emphasize the differential impact that ligand radical stability, size, and lipophilicity can have on quantum yield. Furthermore, the photolytic quantum yields for some alkyl-Cbls (e.g., methyl)^{38, 57} are known to be dependent upon the illumination wavelength, whereas others (e.g., adenosyl)⁵⁷ are wavelength independent. In addition the nature of the ligand and wavelength of illumination, the ability of the corrin ring system to absorb photons also contributes to photolytic efficiency.⁵⁸ Indeed, the latter is given by the product of ε and Φ,⁵⁹ where ε is the molar absorptivity of the chromophore. The sensitivity of a phototherapeutic’s response to light is a critical parameter for in vivo applications. For example, although Cbl absorbs well into the green (500 – 565 nm), there are endogenous tissue-containing chromophores, including hemoglobin and melanin, that absorb in this wavelength range. In particular, the ε at 550 nm for Cbl is 8750 M⁻¹ cm⁻¹, but the presence of oxygenated hemoglobin (ε = 43000 M⁻¹ cm⁻¹) and melanin (ε ~2000 M⁻¹ cm⁻¹) will efficiently compete for the 550 nm photons.⁶⁰ As a consequence, the ability of green light to pass through tissue is significantly compromised by endogenous biological chromophores. A possible strategy to surmount this difficulty is the use of fluorophores as antennas to assist the alkyl-Cbl in capturing photons of the appropriate wavelength. The premise for this approach is based on the observation that fluorophores appended to Cbl are fluorescently quenched, implying an electronic interaction between the fluorescent partner and the corrin ring system.^{61, 62} Jacobson et al referred to this behavior as “cryptofluorescence” and ascribed it to a non-radiative energy transfer between fluorophore and cobalamin.⁶³ Lee and Grissom, working with different analogs of cobalamin, observed a partial restoration of fluorescence upon insertion of a rigid cyclic linker between the fluorophore and cobalamin, suggestive of a contact quenching mechanism.⁶¹ Alternatively, fluorescent quenching may be due to photoinduced electron transfer between the excited state fluorophore and cobalamin. The latter is especially intriguing since the reduction of Co(III) to Co(II)⁶⁴ results in Co-C bond cleavage in Cbl and analogs.⁶⁵ In addition, the corresponding oxidation of Co(III) to Co(IV)⁶⁶ been asserted to result in Co-C bond cleavage as well. However, a recent study by Kozlowski and colleagues has called into question the existence of a formal Co(IV) oxidation state and, instead, raises the prospect of a mixture of Co(IV)/Co(III) states.⁶⁷ Although we have not investigated which mechanism is in play, the crosstalk between fluorophore and cobalamin supports the possibility that a strongly absorbent fluorophore could boost the photosensitivity of alkyl-Cbl conjugates. Consequently, we examined the photolysis of several ethyl-Cbl-linker-fluorophore derivatives using green light (505 and 550 nm) filtered through a tissue mimetic replicating the absorbance properties of darkly pigmented skin (Fitzpatrick series V – VI).⁴⁴

Three green light capturing fluorophores were explored: a Bodipy derivative (λ_max = 505 nm, ε = 80,000 M⁻¹ cm⁻¹),⁴² Cy3 (λ_max = 550 nm, ε = 125,000 M⁻¹ cm⁻¹), and TAMRA (λ_max = 550 nm, ε = 85,000 M⁻¹ cm⁻¹) (Chart 2). In addition, we installed two structurally distinct linkers between the primary ribose hydroxyl moiety of the Cbl and the fluorophores: the relatively flexible ethylene diamine (EDA) and a more conformationally rigid counterpart, 1,4-trans-diaminocyclohexane (DACH). The latter was included in this study due to the observation that DACH partially disrupts the electronic interaction between the corrin ring and appended fluorophores (fluorescein and rhodamine 6G),⁶¹ suggesting that the conjugate containing this linker might prove less effective in promoting photolysis. In all cases, we photolyzed ethyl-Cbl-linker-fluorophore derivatives and quantified the rate constants to compare the impact of the structural diversity on the rate of photolysis. The unmodified ethyl-Cbl (6) served as a control in these experiments.

Chart 2.

Linkers (EDA and DACH) and fluorophores (Bodipy, Cy3, and TAMRA) employed in the green light (505 and 550 nm) photolysis of ethyl-Cbl-linker-fluorophore derivatives.

Several conclusions are readily drawn from the photolysis of the ethyl-Cbl-linker-fluorophore derivatives (Table 3). First, at equivalent concentrations, all fluorophore-substituted Cbl derivatives display between a 2-to-4-fold enhancement in photolysis rate relative to the parent ethyl-Cbl. This supports the conjecture that the addition of an antenna, in the form of a strong light absorbing fluorophore, promotes the rate of photolysis of the ethyl-Co bond under tissue mimetic conditions. Second, in all cases, the more flexible linker (EDA) furnishes a greater photolytic rate then the corresponding conformationally constrained (DACH) derivative. This is consistent with the notion that the quenching mechanism is supported by some form of crosstalk between the fluorophore and the Cbl.⁶¹ Finally, we decided to explore, in greater detail, the photolytic rate differences between ethyl-Cbl and ethyl-Cbl-EDA-TAMRA as a function of the tissue skin mimetic Fitzpatrick series. The Fitzpatrick series of phantoms (a mix of varying concentrations of melanin, intralipid)⁵⁸ are designed to replicate the properties of pale (I-II), medium (III-IV), and dark (V-VI) skin color.⁴⁵ The ethyl-Cbl and ethyl-Cbl-EDA-TAMRA constructs were exposed to 550 nm light filtered through a 2 cm deep solution of pale, medium, and dark skin mimetic. In all three cases, the ethyl-Cbl remained almost completely structurally intact over the course of 30 min of illumination. By contrast, ethyl-Cbl-EDA-TAMRA underwent photolysis at least 4 times more rapidly than its fluorophore-free counterpart. Finally, as expected, the TAMRA derivative does show a decrease in photolytic rate as a function of increasing skin mimetic darkness.

Table 3.

Photolysis (0.025 mW output Xe arc lamp at 505 or 550 nm) of ethyl-Cbl and ethyl-Cbl-linker-fluorophore in phosphate buffered saline.

Linker	Fluorophore	Relative Rate	Rate (min−1) Fitzpatrick I/IIc	Rate (min−1) Fitzpatrick III/IVc	Rate (min−1) Fitzpatrick V/VIc
-	-	1	0.13 ± 0.01	0.09 ± 0.01	0.11 ± 0.02
EDA	Bodipy a	4.3 ± 0.7	-	-	-
DACH	Bodipy a	3.1 ± 0.5	-	-	-
EDA	Cy3 b	4.5 ± 0.6	-	-	-
DACH	Cy3 b	3.1 ± 0.3	-	-	-
EDA	TAMRA b	4.0 ± 0.4	0.56 ± 0.08	0.44 ± 0.05	0.32 ± 0.02
DACH	TAMRA b	2.8 ± 0.4	-	-	-

^aPhotolysis performed at 505 nm.

^bPhotolysis performed at 550 nm.

^cLight from the Xe arc lamp source was pre-filtered through 2 cm of a Fitzpatrick phantom solution.

CONCLUSIONS

We have examined the photolytic behavior of a structural array of alkyl-Cbls. The majority of these derivatives undergo efficient photolysis with high quantum yields. However, a few analogs, such as cyclopropyl-Cbl and decyl-Cbl, display a significant reduction in photolysis rate relative to the standard ligands utilized for Cbl-based phototherapeutics.^{30, 31, 68} By contrast, the installation of fluorophores on the ribose 5’-hydroxyl moiety, in the form of photon-capturing “antennas”, enhance the photolysis rate of alkyl-Cbls under challenging tissue mimetic conditions. In short, the ability to increase or decrease the photosensitivity of Drug-linker-Cbl derivatives potentially enables phototherapeutics to be photochemically tuned to the required application (e.g., surface versus deep tissue therapy). We note that the chemistry of Cbls is extraordinarily rich⁶⁹ and modified derivatives are well tolerated by proteins involved in Cbl uptake and transport. This suggests that Cbl conjugates structurally related to the derivatives in this study could be readily absorbed (e.g., sublingually)⁷⁰, transported (in the circulatory system)⁷¹ and delivered to diseased sites⁷². These studies are currently ongoing in our lab.