Design Strategy and Approach to Increase the Selectivity of 1,2-Bis(Sulfonyl)-1-Alkylhydrazine Warheads in Tumor-Targeted Applications

The 1,2-bis(sulfonyl)hydrazines (BSHs) have very short tunable half-life values, which give them unmatched abilities in terms of their activity confinement to the target site. Furthermore, their therapeutic cytotoxicity, which is delivered as a consequence of their decay, is due to the alkylation of the O-6 position of guanine. This has led to the suggested use of BSHs in tumor targeted applications, in particular to target the subset of tumors lacking O 6 -methylguanine-DNA methyltransferase (MGMT), the sole enzyme involved in the repair of guanine O-6 alkyl lesions. While current BSHs largely owe their therapeutic cytotoxicity to the alkylation of the O-6 position of guanine, there is evidence that alkylation does occur at multiple sites in DNA, and in other biomolecules, as a consequence of alternative decomposition pathways that lead to the production of electrophiles with a preference for other nucleophilic sites. In this paper we discuss strategies to minimize these alternative reaction pathways, thereby further increasing the activity and selectivity of all categories of BSH prodrugs. Design Strategy and Approach to Increase the Selectivity of 1,2-Bis(Sulfonyl)-1-Alkylhydrazine Warheads in Tumor-Targeted


Introduction
The DNA repair protein O 6 -methylguanine-DNA methyltransferase (MGMT) is either absent or expressed in insignificant levels in a small sub-set of many tumor types [1,2]. The size of this subpopulation has been reported to be between 5-30% depending upon the tumor type [1][2][3][4][5]. DNA damage at the O-6 are no approved agents with sufficient DNA guanine O-6 target specificity and potency to effectively accomplish this aim. The proposed improved BSHs can be readily incorporated into a variety of tumor-specific delivery platforms for additional tumor selectivity [2,5]. By exceeding the distinct protection limits provided by MGMT [2,13], tumor-targeted BSHs are expected to function against both moderate MGMT-expressing and non-expressing tumors. Normal tissue MGMT levels would afford protection from the lesser exposures received due to active agent leakage and non-targeted delivery. In targeted applications, the highly tunable half-life of the liberated BSH is also a major advantage since it can be used to limit the spatial volume surrounding the activating cell or region to which the cytotoxic stress is delivered, minimizing off-target cytotoxicity due to leakage from the intended target site [14].
Using existing therapeutics in targeted applications is inherently flawed as their utility as 'free-agents' is due to pharmacokinetic properties that favor rapid distribution and tissue penetration, and these very same properties will enable the escape and systemic distribution of a formerly targeted agent [14][15][16]. The unique properties of BSHs circumvent these limitations and advocate their further development and incorporation into tumor specific delivery platforms [14]. Currently, there are two basic types of approved agents that alkylate the O-6 position of DNA guanine: those which methylate and those that chloroethylate. Unfortunately, the O 6methylguanine lesion is of very low cytotoxicity even in sensitive MGMT-deficient cells, since ~ 5600 lesions/cell are required for a 50% growth inhibition [13], and loss of MMR essentially negates the toxicity of this lesion [17][18][19][20][21]. Despite this limitation, the guanine O-6 methylating agent, temozolomide is likely the most efficacious current agent against MGMT-deficient tumors.
Guanine O-6 chloroethylators behave differently since the initial lesions formed transition into 1-(N 3 -cytosinyl),-2-(N 1 -guaninyl) ethane DNA-DNA interstrand cross-links (G-C ethane cross-links) via an N 1 ,O 6 -ethanoguanine intermediate [22][23][24]. While the O 6 -chloroethylguanine and N 1 ,O 6 -ethanoguanine lesions can be repaired by MGMT, the highly lethal G-C ethane cross-link cannot [22,25]. Thus, for a cell to survive, the MGMT activity must be large enough to produce a repair rate that clears the cross-link precursors before a small but lethal number transition to cross-links (<10 in some clonal survival assays) [23].
A large excess of MGMT relative to the number of O-6 guanine lesion number is therefore required, but little MGMT depletion occurs at cytotoxic doses. This contrasts with O 6 -methylguanine where complete MGMT titration has to occur before toxicity is manifested [13,23]. In addition, loss of MMR does not result in resistance to guanine O 6 -chloroethylation [26]. Cells can repair a very limited number of G-C ethane cross-links using homology directed repair (HDR) and cells lacking both HDR and MGMT are hypersensitive to agents of this type [27]. The higher potency of guanine O-6 chloroethylators (~20-fold) also makes them more attractive than guanine O 6 -methylators for incorporation into tumor delivery platforms, which are frequently limited in their drug loading capacity [28]. Unfortunately, the relatively non-selective

BSHs as O-6 DNA Guanine Targeting Agents
Laromustine (also known as cloretazine, onrigin, VNP40101M, and 101M), a chloroethylating BSH prodrug, was originally designed in our laboratory to eliminate many of the non-efficacious alkylating species generated by the CNUs while retaining the production of the therapeutic electrophiles [29,34,35]. Therefore, hydroxyethylating, vinylating, aminoethylating and direct alkylating activities present in the BCNU were removed, while the generation of oxophilic chloroethylating species favoring the chloroethylation of guanine O-6 (unequivocally required for anticancer activity) and the carbamoylating activities were retained [34,35]. Carbamoylating activity was retained since it had been speculated by several groups that the intracellular release of isocyanates played a role in modulating the biological activity of the CNUs against some specific tumor types [36][37][38][39][40][41][42]. Thus, laromustine was significantly less toxic than BCNU towards mice, yet displayed excellent antitumor activity against MGMT-deficient tumor models, producing 100% cures in many cases [36]. Furthermore, laromustine has a therapeutic index (LD 50 /ED 50 ) against the L1210 leukemia of > 8, more than double that of the best of over 300 nitrosoureas evaluated [36,42].
Although laromustine was the subject of several clinical trials, it was never tested against tumor sub-sets prescreened for low MGMT activity, where a very high response rate would be expected. However, in one Phase I study in patients with refractory leukemia, the MGMT levels in total peripheral blood mononuclear cells (a mixture of peripheral leukemic blast and normal peripheral blood mononucleated cells) were ascertained, and lower MGMT levels were found in responding patients. Median MGMT levels in responding and non-responding patients were 5.73 and 9.21 fmol/ µg of cellular DNA, respectively [43]. A much stronger correlation with therapeutic activity would have been likely if purified leukemic blasts had been assayed rather than total peripheral blood mononuclear cells. Laromustine as a single agent produced an objective response rate (CR, complete remission + CRp, all of the criteria for a CR with platelets less than 100,000/µl) of 32% in de novo AML of poor-risk elderly patients [44]. does result in non-specific toxicity. This accounts for the lower MGMT-dependent cytotoxicity differential seen with laromustine compared to 90CE [32]. An obvious approach to improving the therapeutic index against MGMT-deficient tumors is to replace the methylaminocarbonyl moiety (the source of carbamoylating activity) of laromustine with a cleavable group that does not give rise to a cytotoxic product with no therapeutic benefit.

Some earlier 1,2,2-tris(sulfonyl)-1-(2-chloroethyl)hydrazines
(possessing significantly poorer pharmacokinetics) synthesized in our laboratory already fulfilled this condition and produced 100% cures of the L1210 leukemia with less host toxicity as judged by bodyweight loss [48]. However, the major thrust of this paper is to optimize the 90CE moiety itself. An improved 90CE type 'warhead' can then be exploited by its incorporation into a wide number of agents from spontaneously activated prodrugs such as laromustine [36] to far more complex tumor-selective delivery platforms [2,5,14].
Alkylating agents generate electrophilic species which react with biological nucleophiles, including sites within DNA. These electrophiles can exhibit significant selectivity for their sites of reaction [49]. However, this selectivity is frequently overlooked in the biological literature where the tendency is to assume that all alkylating agents are reactive species that uniformly lack discrimination in their choice of biological nucleophilic targets.
The concept of 'hard' and 'soft' was introduced as a qualitative predictor of electrophile/nucleophile reaction preference [49]. is more a function of the wide range of disparate electrophilic species generated by BCNU rather than the nucleophilic promiscuity of individual electrophilic species [29,49].
Recent studies on the fate of 90CE in different buffer environments have revealed a surprising catalytic effect of Brønsted-Lowry bases on the decomposition pathways taken by 90CE [34] and related compounds. Phosphate and its mono-and diesters were found to be particularly active in this regard [34]. This Brønsted-Lowry base-catalyzed pathway leads to the production of major quantities of a relatively soft unwanted thiophilic, and aromatic nitrogen preferring electrophile [34]. Furthermore, our studies concerning these decomposition mechanisms/routes have indicated how to more precisely tune 90CE analogs to: A. Reduce the production of this unwanted relatively soft electrophile which diverts material from the desired decomposition pathway and likely contributes to MGMTindependent cytotoxicity. Improvements in any of these parameters will further amplify the strong inverse relationship between MGMT activity and cytotoxicity seen in 90CE analogs, leading to lower host toxicities at cytocidal concentrations towards tumor cells lacking MGMT activity.
Prodrug delivery systems with uptake/activation mechanisms based upon nanoparticle-enhanced permeability and retention (EPR) effects [5], hypoxia-specific net reduction [2,14], glutathione S-transferase-catalyzed thiolysis [50], acid-catalyzed activation by solid tumor extracellular acidity [47], and extracellular matrix protease activation [51], can then be used to more selectively deliver improved 90CE analogs and add a further degree of tumor selectivity. Thus, improvements in 90CE type moieties can bring about a wide range of benefits to the design of BSH prodrugs. When coupled with delivery strategies these moieties can display activity against MGMT-expressing tumors by overwhelming their MGMT levels [52,53]. In addition, tumor-targeted prodrugs of short-lived methylating BSH can be used to selectively sensitize tumor cells by ablating tumor MGMT [2,52]. These agents have a significant advantage over equivalently targeted O 6 -benzylguanine prodrugs in terms of the diminished escape of the activated agent from the target site due to their short half-lives [2,52]. Early decomposition studies on BSHs were conducted in weak Tris-HCl (pH 7.4) buffers where the newly identified Brønsted-Lowry base-catalyzed decomposition pathway was almost absent and therefore overlooked [54]. Recent studies in buffers with high Brønsted-Lowry base activity (such as phosphate) indicated that haloethyl BSHs (and some analogs thereof) decomposed almost entirely by an alternative pathway that is a relatively soft thiophilic, and aromatic nitrogen preferring electrophile (Figure 1 pathway B) in this environment [34]. This pathway does not generate the therapeutic oxophilic haloethylating species favoring the alkylation of DNA guanine O-6, but rather a reactive, softer electrophile that reacts via 'Michael type' addition reactions and most likely contributes to non-specific toxicities that show no MGMT activity dependence [34]. Even though the molar yield of guanine O-6 chloroethylations by 90CE and eventual yields of cross-links are ~ 2-fold greater than those obtained from BCNU they still represent a small proportion of the total overall alkylations [23,29]. However, the 90CE guanine O-6 alkylations account for a much greater proportion of its cytotoxicity as evidenced by the ~ 20-fold increase in resistance produced by MGMT expression, compared to only a 2- can all easily be independently assayed/determined, allowing the identification of analogs that strongly favor the desired A pathway over the B/C pathways. In addition, the relative rates of the different guanine O-6 adducts generated by pathway A to progress to crosslinks or be repaired by MGMT can also readily be evaluated [34,47].

The contribution of Different Decomposition Routes to Efficacy and Toxicity of 90CE
Guanine O-6 adducts with the greatest MGMT 'interceptability' are expected to exhibit a greater inverse MGMT level dependent cytotoxicity. Thus, by maximizing the flux down pathway A, and the resultant lesion's MGMT repairability, warheads/agents with an extreme inverse MGMT level dependent cytotoxicity will be produced, thereby minimizing toxicity towards normal MGMTexpressing cells and greatly enhancing toxicity to cancers selected for low or no MGMT expression.

Optimization of Pathway A
The therapeutic oxophilic electrophile generated by 90CE (CH 3 SO 2 N=NCH 2 CH 2 Cl) favors the chloroethylation of hard oxygenbased nucleophiles. Thus, water, in part because of its presence in a vast molar excess, intercepts the majority of these electrophiles, decreasing the yields of O 6 -(2-chloroethyl)guanine [23,34]. By tuning the electrophilic hardness of the generated electrophile to more strongly favor the alkylation of guanine O-6 relative to the alkylation of water, potency can be increased, and unwanted alkylations minimized.

Minimizing Unwanted Pathway B
We have compared the Brønsted-Lowry base (inorganic phosphate in these experiments) catalyzed flux down pathway B for 90CE and a number of analogs; 90BE, 90IE, and 90SE ( Figure   2). It can be seen that the acidity of the proton beta to the halide (or methylsulfonate) rather than the leaving group ability of the halide or methylsulfonate controls the propensity of the molecule to react via the Brønsted-Lowry base catalyzed pathway B. In the case of 90IE (iodide is an exceptionally good leaving group), and an additional mechanism independent of that catalyzed by Brønsted-Lowry base(s) appears to be operative as evidenced by the observation that  The effects of Phosphate acting as a Bronsted-Lowry base on the mole fraction of 90CE analogs that decomposes via reaction pathway B to generate a thiophilic electrophile. The inset table indicates the concentration pf phosphate that results in a half maximal phosphate-induce pathway switch. Note 90IE is somewhat unusual in that a significant proportion generates a thiophilic electrophile independent of the phosphate concentration. This likely represents a mechanism switch when "X" is a superb leaving group.
A. Decrease the acidity of the beta hydrogen in the primary chloroethylating species by utilizing a pseudohalide, or other leaving group, comparable in leaving group ability to Cl but less electron-withdrawing.

B.
Decrease the acidity of the beta hydrogen by replacing the N-2 methylsulfonyl moiety with a less electron-withdrawing moiety, e.g., a dimethylsulfamoyl moiety.

C.
Replace the beta hydrogens (or all hydrogens on the chloroethyl moiety) with deuteriums. Since the C-D bond is significantly stronger than the C-H bond (~ 12 KJ/mol), the activation energy required to break the former is larger, resulting in a smaller rate constant [54,55].
Thus, a pronounced kinetic isotope effect is observed on deuterium/hydrogen substitution when the breaking of this bond is involved in the rate-determining step. The extent of reaction slowing is quite variable and is likely dependent upon the degree of breakage of the C-D bond in the transition state. Although reaction rate differentials of < ~ 7-fold are usually observed, values as high as 70-fold have been documented [56]. It is expected that by using these methods, or a combination thereof, toxicities arising from, and potency losses due to, flux down pathway B can be rendered inconsequential.

Minimizing Unwanted Flux Down the C Pathways
Further fragmentation of the therapeutic alkylating species, CH 3 SO 2 N=NCH 2 CH 2 Cl, via the C pathways to generate very hard,

Optimization of BSH pKa Values
Weak acids like BSHs have advantages in treating solid tumors since these cells possess atypically large and inverted transplasmalemma pH gradients [58]. Normal tissues have an intracellular pH value of ~ 7.2 and an extracellular pH of ~ 7.5, while tumor tissues have intracellular pH values ~ 7.5 and extracellular pH values around 6.9. Weakly acidic agents become more protonated in the acidic extracellular environment, and thus more permeable, but upon entering the cell become more ionized and more strongly retained. These beneficial pH gradient effects are expected to be further exaggerated by the effects of pH on BSH half-life [47]. In the acidic extracellular tumor environment BSHs will be stabilized but upon cancer cell entry, their half-life will shorten. This will prevent the attainment of a pH biased concentration equilibrium and will maintain a concentration gradient to drive BSH entry [47]. The c. To maximize differential toxicity solely based on the differentials in intracellular half-life (max differential ~ 1.8fold) between normal and tumor cells the best pKa value is 7.35.
The relative importance and interplay of these three major factors is complex. Therefore, experimentation would be required determine the optimal BSH pKa range using analogs in in vitro clonal survival assays conducted over relevant extracellular pH values. Preliminary experiments utilizing 90CE (pKa ~ 6.7) revealed extracellular pH dependent changes in cytotoxicity consistent with the expected pH biased transplasmalemma equilibria [47]. It should be noted that BSH half-life has a greater dependency upon the leaving group ability than the electron withdrawing effect of the N-1 substituent, while the pKa shows the converse relationship.
Therefore, these two important characteristics do not have to move in parallel. Weakly basic compounds exhibit the opposite accumulation effect to those of weak acids like the BSHs, in response to pH gradients, and this attenuates their cancer cell selectivity.
Thus, anticancer drugs such as doxorubicin and vinblastine, are negatively impacted, reducing their desirability as tumor targeted warheads [58].While the deuteration of the chloroethyl group is expected to primarily influence only decomposition pathway B directly, other changes will more strongly influence multiple facets of the decomposition pathways to various extents [59]. The use of computational chemistry is expected to aid not only in the selection of compounds for synthesis by predicting effects that a purely intuitive approach to compound selection may miss, but also in the interpretation of structure-activity relationships [60]. It is likely that the reaction energy barrier to cyclization following the

Experimental Approaches
The experimental approaches to investigate the decomposition pathway fluxes and alkylation preferences of the newly synthesized 90CE analogs are listed below.

Synthesis of deuterated 90CE and initial agents of interest
arising from computational studies.

F.
Testing of simple laromustine-like prodrugs of the most promising 90CE analogs in murine models with tumors expressing and not expressing MGMT.

G.
Measurement of solubility and octanol/water partition coefficient of simple prodrug forms of novel BSHs.
H. Measurement of BSH half-life and pKa values (determined from the pH dependency of decay rate).

Chemical Synthesis
No complex chemistry is expected to be required for the synthesis of the described BSHs, and a minimally equipped chemical synthesis laboratory should suffice [35,47,48,54,60,61]. A new synthetic route to 90CE using readily available di-tert-butylhydrazine-1,2dicarboxylate as the starting material is illustrated in Figure 3, this should allow the facile synthesis of the deuterated analogs, using readily available deuterated 2-bromoethanol. This method is based upon a modification of a published high yield method [62,63] ( Figure 3).