**1. Introduction**

Late-stage functionalization of organic molecules has emerged as an important strategy in modern drug discovery programs as it allows for direct derivatization without the need for pre-functionalized synthetic handles [1]. This strategy, which involves the direct substitution of a C–H bond with a new functionality, is currently under-utilized in the field of natural products where the generation of analogues provides structure–activity relationships (SAR), a critical component often missing in current biodiscovery programs. The ability to make unusual derivatives of a bioactive scaffold via a simple one-step procedure is of particular relevance in the field of natural products where typically, only small amounts of material or analogues are available. Nitrogen-rich heterocyclic compounds sourced from nature have played a profound role in human health and these motifs are found in many of the current drugs that are used to treat various diseases [2]. Transition-metal-mediated cross-coupling reactions that require pre-functionalized starting materials have been used extensively in the synthesis of such molecules [3–5]; however, the direct C–H functionalization of biologically active heterocycles is still underdeveloped and worthy of further investigations [6–14].

Recent developments in radical-mediated C–H functionalization of heterocycles, including Minisci [15], borono-Minisci [7], and reactions with sulfinate reagents [2], have led to a resurgence in the use of radical-based methods, due primarily to improvements in substrate scope and mild reaction conditions.

In 1971, Minisci et al. reported the addition of carbon-centered radicals to heteroaromatic bases through the silver-mediated decarboxylation of carboxylic acids in the presence of persulfate [15]. One feature that makes this form of innate C–H functionalization appealing for pharmaceutical applications is that protection and deprotection protocols are rarely needed [6]. While these conditions are compatible with alkyl and acyl radicals (derived from alkyl halides, carboxylic acids, and related derivatives), limitations in functional group compatibility, high reaction temperatures (>70 ◦C), and the requirement of transition-metal additives and strongly oxidizing conditions [6] make them

unsuitable for more complex chemical structures. Since this initial publication, numerous methods of C–H functionalization have been reported by Baran [2,7,9,14], Molander [8], and others [16,17], which have significantly increased the scope and generality of this strategy. With the goal of the direct transformation of C–H bonds into C–C bonds in a more practical manner, Baran et al. have developed a radical-based functionalization strategy that involves the use of zinc bis(alkanesulfinate) reagents [12,14]. The most notable features of C–C bond formation using this sulfinate chemistry are that it involves a one-pot reaction, occurs under mild conditions with no need for pre-functionalized starting materials, and the reactions can be conducted in open flasks as they do not require the exclusion of air or moisture [2]. This approach has the advantage of rapidly accelerating drug discovery timelines regardless of whether the compounds of interest are natural or synthetic.

With the intent of identifying hit or lead compounds that are based on bioactive natural products, our research focuses on the semi-synthesis of biodiscovery libraries using unique natural product scaffolds that have been isolated from sources such as fungi, plants and marine invertebrates [18–24]. Herein, we report late-stage functionalization studies on papaverine, a nitrogen-containing heterocyclic natural product, using the commercially available sulfinate reagents known as DiversinatesTM. Some unexpected chemistry was identified during these studies.

#### **2. Results and Discussion**

Commercially available papaverine hydrochloride (**1a**) [25] was chosen as a model compound for our initial foray into C–H functionalization studies utilizing the sulfinate chemistry that has been described by Baran and other research groups [2,9,14]. Fluoroalkyl substituents have become increasingly valuable in modern drug discovery due to their resistance toward oxidation by cytochrome P450 oxidases. Also, incorporation of halogen atoms on hit/lead compounds has been performed in order to exploit their steric effects through the ability of these atoms to occupy the active site of molecular targets [26–28], and establish intermolecular bonds in a manner that resembles H-bonding [29–31].

Baran et al. have published a number of sulfinate reaction conditions, where the DiversinateTM reagents are compatible with different organic solvents (e.g., DMSO, CH2Cl2, ClCH2CH2Cl, perfluorotoluene, perfluorohexane, and anisole). Furthermore, it has been determined that fluoroalkyl zinc sulfinate reagents react best in halogenated solvents, such as CH2Cl2, alkyl zinc sulfinate salts react more favorably in DMSO [2], and stoichiometric conditions for the peroxide and DiversinateTM reagents, as well as reaction temperatures and times, vary greatly in the literature. Before synthesizing the targeted papaverine library, we initially conducted several experiments that tested the effect of solvents (e.g., DMSO/CH2Cl2 / H2O), reagent stoichiometry, and additives [e.g., trifluoroacetic acid (TFA)] with papaverine HCl (**1a**) and the commonly used DiversinateTM, zinc trifluoromethanesulfinate [(CF3SO2)2Zn] (Table S71). From this data, it was clear that the best yield for the major mono-CF3 analogue (**2**) was obtained using 6 mol eq. of (CF3SO2)2Zn and 6 mol eq. of *tert*-butyl hydroperoxide (TBHP) in CH2Cl2 for 16 h. In order to ascertain whether the presence of HCl was affecting the yields (the chloride ion could be competitively oxidized by TBHP), we generated the free base of papaverine (**1b**) and repeated the test reactions. Surprisingly, the reaction on the free base gave a lower yield of **2** (15%) compared to the papaverine HCl reaction (compound **2**, 24%). Subsequently, the effect of the addition of TFA was investigated using both the free base and HCl salt of papaverine and a mixture of products was produced, with the free base (**1b**) affording the best yield (10%) with the solvent conditions CH2Cl2/TFA/H2O (1 mol eq. TFA). Based on these data (Table S71), all subsequent reactions were performed using the HCl salt of papaverine and 6 mol eq. of the DiversinateTM reagent in CH2Cl2/H2O (2.5:1) for the fluorinated DiversinatesTM, while DMSO/H2O (2.5:1) was chosen for use with the non-fluorinated DiversinatesTM.

For example, scaffold **1a** (0.1 mmol of the HCl salt) was treated with (CF3SO2)2Zn (0.6 mmol) in a mixture of CH2Cl2 (100 μL) and H2O (40 μL) at 0 ◦C. The mixture was stirred at 0 ◦C and TBHP (0.6 mmol) was slowly added, followed by stirring for 20 min. The mixture was then allowed to warm to room temperature over 16 h [2], before evaporation under a stream of nitrogen and chromatography (HPLC on NH2-bonded silica) was undertaken in order to give the major products, substituted papaverines **2** (24%) and **3** (4%), and the recovered starting material (7%) (Scheme 1). HPLC showed a multitude of UV-active peaks. This, and the recovery of only 7% of the starting material, suggested that the reaction of trifluoromethyl radicals with papaverine was very unselective and/or the products were unstable under the reaction conditions. The preferred attack on the electron-rich 3,4-dimethoxybenzyl ring was not surprising as the trifluoromethyl radical is electrophilic.

**Scheme 1.** Reaction of papaverine HCl (**1a**) and free base (**1b**) with zinc trifluoromethanesulfinate [(CF3SO2)2Zn] and *tert*-butyl hydroperoxide (TBHP) in CH2Cl2/H2O.

Interestingly, when the reaction was repeated with the less electrophilic reagent sodium 1,1-difluoroethanesulfinate, the products obtained were those involving substitution on the isoquinoline ring, rather than substitution on the dimethoxybenzyl ring. The products obtained were **4** (7%) and **5** (3%), as well as the recovered starting material (8%) (Figure 1). In all of these reactions, a significant amount (generally between 7 and 14%) of starting material was recovered even though a six-fold excess of reagents was employed.

**Figure 1.** Chemical structures of the other synthesized papaverine analogues **4**–**11**.

The closely related zinc difluoromethanesulfinate gave a low yield of **6** (3%) analogous to **5**, but only when the reaction was carried out in the presence of TFA (0.1 mmol). No products were detected in the absence of TFA. Although the reaction employed the HCl salt of papaverine under the reaction conditions of excess (CF2HSO2)2Zn and *t*-BuOOH, the HCl would be neutralized by the zinc hydroxide formed. Presumably, the addition of TFA ensured at least partial protonation of the isoquinoline nitrogen, rendering the papaverine more electrophilic and therefore more reactive toward radicals that had some nucleophilic character. Baran et al. found that the difluoromethyl radical had nucleophilic properties, preferring to attack *N*-heterocyclic compounds at electron-deficient centers [12].

Use of the nucleophilic radical precursor zinc isopropylsulfinate gave the papaverine derivative **7** (4%), analogous to **5** and **6**. Once again, TFA was required for the formation of **7**. Surprisingly, the major product was **10** (19%), formed by an apparent radical substitution reaction. Analogous reactions were observed with the nucleophilic radical precursors zinc 4-methoxybenzylsulfinate, zinc benzylsulfinate, and 4,4-difluorocyclohexylsulfinate to give **8** (4%), **9** (14%), and **11** (4%), respectively. A suggested mechanism for these radical substitution reactions is outlined in Scheme 2.

**Scheme 2.** Proposed single-electron transfer (SET) mechanism for the formation of the side-products **8**–**11**.

The isolated yields in these reactions were generally low. This was partly due to the large number of minor products formed with this particular scaffold (as evidenced by the HPLC data) and partly due to losses during HPLC purification. Some of these minor products may be due to the attack by trifluoromethyl radicals on the dimethoxyisoquinoline ring and/or bis-trifluoromethylation (which could yield 21 different compounds). In addition, the papaverine scaffold contains four methoxyl groups and a benzylic methylene group, all of which could undergo hydrogen atom abstraction by *tert*-butoxyl radicals leading to a plethora of minor products. Interestingly, Kuttruff et al. [32] have recently reported rather low yields (3–30%) and sometimes extensive decomposition with these reactions. They employed a range of scaffolds incorporating benzene, pyridine, pyrimidine, imidazole, pyrazole and thiazole rings. They also explored different reaction conditions and were able to achieve modest increases in yields in some cases via the addition of Fe(acac)3. However, in other cases, the presence of Fe(acac)3 led to rapid decomposition. Baran et al. [9] have also noted that one of the limitations of the method is that some substrates deliver only moderate amounts of product.

We suggest that the main reason for the poor reactivity of **1a** toward difluoromethyl radicals is the presence of the dimethoxybenzyl group at position 1. Difluoromethyl radicals react readily with isoquinoline and with 3-methylisoquinoline at position 1 (the most electrophilic position), but they do not react with 1-methylisoquinoline [33]. It is interesting to note that the more nucleophilic isopropyl radicals attack at position 1 of the papaverine despite the significant steric hindrance to attack at this position. This is followed by loss of the (more stable) 3,4-dimethoxybenzyl radical to give **10** in moderate (19%) yield.

Compounds **2**, **8**, and **9** have previously been synthesized via multi-step syntheses; **2** was generated using the copper-mediated trifluoromethylation-allylation of arynes [34]; **8** was produced from a reaction of Raney nickel with N-benzylsulfonamide [35], and via the desulfonylation of N-sulfonyl tetrahydroisoquinolines with KF/Al2O3 [36]; and **9** was synthesized using a ruthenium-mediated dual catalytic reaction, and oxidative cross-dehydrogenative coupling with methyl arenes [37,38]. However, this is the first report of the synthesis of **2**, **8**, and **9** using sulfinate chemistry. Furthermore, these three compounds were only partially characterized and none of them had their 1H and 13C chemical shifts assigned to specific positions within the alkaloidal skeleton. We report here the first synthesis of several other papaverine analogues as their free base; full characterization using 1D/2D NMR, UV, IR, and MS data was performed during these studies.

The first products to be fully characterized were compounds **2** and **3**. The 1H NMR spectra (Table 1) of these two mono-CF3 derivatives enabled the definitive positioning of the CF3. Careful comparison of the natural product scaffold (**1b**) NMR data in CD3OD with the fluorinated analogue **2** indicated that this molecule had a CF3 group attached to C-15, since the 1H NMR chemical shifts and multiplicities associated with the pendant benzyl moiety of the starting material had changed from a classic 1,3,4-trisubstituted benzene system (δ<sup>H</sup> 7.03, d, *J* = 2.1 Hz, H-11, 6.92, d, *J* = 8.4 Hz, H-14 and 6.82, dd, *J* = 8.4, 2.1 Hz, H-15) to a 1,3,4,6-tetrasubstituted benzene system (δ<sup>H</sup> 6.41, s, H-11 and 7.26, s, H-14) (Table 1). Furthermore, a heteronuclear multiple bond correlation (HMBC) (Figure 2) from H-14 (δ<sup>H</sup> 7.26) to CF3 (δ<sup>C</sup> 126.4) provided further proof of the structural assignment. In a similar fashion to **2**, NMR data analysis of **3** also showed that the CF3 was attached to the pendant aromatic ring of papaverine; however, in this case, the fluorinated moiety was attached to C-14, as indicated once again by the 1H NMR chemical shifts and multiplicities associated with the pendant benzyl moiety of papaverine that had changed from the 1,3,4-trisubstituted benzene system to a 1,3,4,5-tetrasubstituted benzene system (δ<sup>H</sup> 7.22, d, *J* = 2.2 Hz, H-11 and 7.11, d, *J* = 2.2 Hz, H-15) (Table 1). The HMBC spectrum analysis of **3** was also critical in confirming the CF3 positioning; key HMBC correlations for **3** are shown in Figure 2.

Detailed 2D NMR data analyses were also performed on all other analogues generated during these studies (see Supplementary Materials).

Surprisingly, the reactions that used other commercially available DiversinatesTM, such as zinc chloromethanesulfinate, zinc chloroethanesulfinate, sodium (2,4-dichlorophenyl)methanesulfinate, sodium 1-(trifluoromethyl)cyclopropanesulfinate, and sodium *tert*-butylsulfinate were not successful using our methodology (with or without TFA), as determined by LCMS analysis of the crude reaction mixtures after 16 h. The lack of reaction with sodium *tert*-butylsulfinate and sodium 1-(trifluoromethyl)cyclopropanesulfinate was possibly due to steric hindrance, but it is unclear why the other DiversinateTM reactions were unsuccessful.

**Figure 2.** Key HMBC ( ) correlations for **2** and **3**.


**Table 1.** 1H (800 MHz) and 13C (200 MHz) NMR data for the mono-CF3 papaverine analogues **2** and **3** in CD3OD at 25 ◦C.

In Scheme 2, we propose that the zinc bis(alkanesulfinate) underwent oxidation by TBHP via a single-electron transfer (SET) process to give a *tert*-butoxyl radical, hydroxide ion, and the radical cation of the zinc sulfinate, which then underwent fragmentation to give the alkyl radical and sulfur dioxide. The nucleophilic alkyl radical then underwent addition at position 1 of papaverine followed by elimination of the dimethoxybenzyl radical. This addition–elimination mechanism is analogous to that proposed for the reaction of carbon-centered radicals with β-bromostyrene [39]. The *tert*-butoxyl radical generated in the first step could oxidize a second molecule of sulfinate to give the *tert*-butoxide anion and the radical cation of the zinc sulfinate. A similar mechanism has been proposed by Baran et al. [9] where the presence of a trace metal initiates *tert*-butoxyl radical formation, compared to our proposed mechanism that involves SET. The dimethoxybenzyl radical could also be oxidized by TBHP to give the corresponding dimethoxybenzyl alcohol and a *tert*-butoxyl radical.

Again, it is interesting that the more nucleophilic radicals attack position 1 of the isoquinoline ring, while the 1,1-difluoroethyl and difluoromethyl radicals, which appear to have both nucleophilic and electrophilic properties, prefer to attack at the 3- or 4-position.

In a final attempt to improve the yields and/or the selectivity of these reactions, we decided to modify the reactivity of the papaverine scaffold. It was envisaged that converting papaverine to its *N*-oxide might increase reactivity, particularly toward electrophilic trifluoromethyl radicals. The free base of papaverine was readily converted to the corresponding *N*-oxide derivative **12** in moderate yield (58%) using *meta*-chloroperbenzoic acid (MCPBA) in CHCl3 without the need for chromatography, using a method previously described by Bremner et al. [40] (Scheme 3). Unfortunately, treatment of **12** with (CF3SO2)2Zn and TBHP under the same conditions as that used for **1a** did not result in an improvement in the yield of the desired DiversinateTM product. After work-up and C18 HPLC purification of the reaction mixture, 1H NMR and LCMS analysis indicated that the major product formed was compound **13,** albeit with a low yield (<12%) and purity (<80%). Significant amounts of starting material (**12**, 26%) were isolated, indicating that papaverine *N*-oxide was less reactive than papaverine toward zinc trifluoromethanesulfinate. As the yield of the desired product had not improved, the reaction of papaverine *N*-oxide with other DiversinateTM reagents was not investigated.

**Scheme 3.** Conversion of papaverine free base (**1b**) to papaverine *N*-oxide (**12**) using *meta*-chloroperbenzoic acid (MCPBA) in CHCl3 and the subsequent reaction of **12** with zinc trifluoromethanesulfinate [(CF3SO2)2Zn] and *tert*-butyl hydroperoxide (TBHP) in CH2Cl2/H2O to form compound **13**.

#### **3. Materials and Methods**
