Structure Elucidation of Poly-Faldaprevir: Polymer Backbone Solved Using Solid-State and Solution Nuclear Magnetic Resonance Spectroscopy

Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

Contents lists available at ScienceDirect

Journal of Pharmaceutical Sciences journal homepage: www.jpharmsci.org

Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Structure Elucidation of Poly-Faldaprevir: Polymer Backbone Solved Using Solid-State and Solution Nuclear Magnetic Resonance Spectroscopy Nina C. Gonnella 1, *, Carl A. Busacca 2, Li Zhang 2, Anjan Saha 2, Jiang-Ping Wu 2, Guisheng Li 2, Mark Davis 1, Thomas Offerdahl 1, Paul-James Jones 3, Lars Herfurth 4, Tim Reddig 4, Klaus Wagner 4, Michael Niemann 4, Ulrike Werthmann 4, Julia Grupe 4, Helmut Roos 4, Gaby Reckzügel 4, Andreas Ding 4 1

Material and Analytical Sciences, Boehringer Ingelheim Pharmaceuticals, Inc., Ridgefield, Connecticut 06877 Chemical Development, Boehringer Ingelheim Pharmaceuticals, Inc., Ridgefield, Connecticut 06877 3 Information Technology, Boehringer Ingelheim Pharmaceuticals, Inc., Ridgefield, Connecticut 06877 4 Boehringer Ingelheim Pharma GmbH & Co. KG, Germany 2

a r t i c l e i n f o

a b s t r a c t

Article history: Received 2 October 2015 Revised 6 March 2016 Accepted 15 March 2016

A large-scale synthesis of the hepatitis C virus drug Faldaprevir revealed precipitation of an unknown insoluble solid from methanol solutions of the drug substance. The unknown impurity was determined to be a polymer of Faldaprevir based on analytical methods that included size exclusion chromatography in combination with electrospray ionization mass spectrometry, solution nuclear magnetic resonance (NMR), matrix-assisted laser desorption ionizationetime of flight, ultracentrifugation, elemental analysis, and sodium quantitation by atom absorption spectroscopy. Structure elucidation of the polymeric backbone was achieved using solid-state NMR cross-polarization/magic angle spinning (CP/MAS), cross polarization-polarization inversion, and heteronuclear correlation (HETCOR) experiments. The polymerization was found to occur at the vinyl cyclopropane via a likely free radical initiation mechanism. Full proton and carbon chemical shift assignments of the polymer were obtained using solution NMR spectroscopy. The polymer structure was corroborated with chemical synthesis of the polymer and solution NMR analysis. © 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.

Keywords: NMR spectroscopy solid-state NMR mass spectrometry polymers structure

Introduction The hepatitis C virus (HCV) is a serious cause of chronic hepatitis, cirrhosis of the liver, and hepatocellular carcinoma. The HCVencoded NS3 protease is known to be essential for the replication of the HCV; hence, the NS3 enzyme has been considered an important pharmaceutical target for the development of small molecule inhibitors for therapeutic intervention.1-4 Faldaprevir is a potent inhibitor of the NS3 enzyme that was under development for the treatment of the HCV infection. This active pharmaceutical ingredient (API) is a complex molecule with 5 stereocenters and a molecular weight of 869.83 Da (Fig. 1).5,6

This article contains supplementary material available from the authors by request or via the Internet at http://dx.doi.org/10.1016/j.xphs.2016.03.017. * Correspondence to: Nina C. Gonnella (Telephone: 203-798-5518). E-mail address: [email protected] (N.C. Gonnella).

To enable large-scale production, a practical synthesis was developed for this complex molecule where the preparation was free of protecting groups, cryogenic conditions, and chromatography.7 During the large-scale production of Faldaprevir, however, a precipitate was found to form in the reaction vessel. Subsequent isolation of the unknown impurity from methanol yielded significant amounts of impurity with approximately 7%-15% of residual API present. Initial examination of the unknown impurity showed the material to be a polymer of Faldaprevir based on a variety of analytical methods. Techniques that included size exclusion chromatography (SEC) in combination with electrospray ionization (ESI) mass spectrometry, matrix-assisted laser desorption ionizationetime of flight (data not shown), and ultracentrifugation (data not shown) showed the impurity to be composed of polymerized Faldaprevir sodium salt having a molecular weight distribution of approximately 30 to 200 kDa. This distribution was attributed to chain lengths of repeating units of Faldaprevir with n ¼ 35-225.

http://dx.doi.org/10.1016/j.xphs.2016.03.017 0022-3549/© 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.

1882

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

O

3 Br

39

N

9

MeO

10 11

N

8

4

6

7

HN

S (Z)

42B

5

(R)

19

30A

17 N

31

30C

(S)

NH

35

O

34

20

H N

40

22

41 ONa

(R)

23

O

33

32 (S)

O

18

O

21

16

37A

42A

1

14 O

30B

2

15

13

12

(Z)

of the unknown material and preventing impurity formation in the production process. Here we report the full structure elucidation of the Faldaprevir polymer. The structure of the backbone was solved from purified material using ssNMR experiments that included cross-polarization/ magic angle spinning (CP/MAS), cross polarization-polarization inversion (CPPI) spectral editing experiments, and 13C, 1H HETCOR heteronuclear correlated spectra. Further confirmation was obtained from ssNMR and solution NMR spectral analysis of synthesized polyFaldaprevir. This study established both the backbone structure of the polymer and a likely mechanism of formation.

(S)

24 25 26

O

36

38A 38B

37B

Figure 1. Numbered chemical structure of the Faldaprevir sodium salt.

Although the repeating units in the polymer consisted of API, the structure of the polymeric backbone was not known. Experimental evidence supported involvement of the exocyclic double bond; however, the chemical connectivity and mechanism of formation could not be determined. Identifying the structure of this impurity and a mechanism of formation was essential to the development process in the evaluation of potential toxic properties

Experimental Sample Preparation Faldaprevir sodium salt (Boehringer Ingelheim) was dissolved in approximately 600 mL dimethyl sulfoxide-d6 and used without further manipulation. The isolated impurity, poly-Faldaprevir sodium salt, was purified using SEC (private communication, Mark Davis). Approximately 1 mg of isolated material was charged to a vial, and 30 mL dimethyl sulfoxide-d6 containing a trace amount of deuterated acetic acid-d4 was added, giving clear solution after brief exposure to a vibrating platform shaker. The solution was added to a 1.7-mm capillary probe for proton NMR investigation. 13 C NMR samples of poly-Faldaprevir sodium salt were prepared by dissolving the material in approximately 600 mL dimethyl sulfoxide containing 20 mg of Chromium (III) acetylacetonate (Cr(acac)3). Solid-State

13

C NMR

Solid-state 13C NMR spectra were acquired using a Bruker Avance III 400 MHz wide bore NMR spectrometer (Bruker-Biospin) operating at 100.71 MHz for 13C. A Bruker DVT (a variable temperature gas channel separate from the bearing and drive channel) CP-MAS probe equipped with a 4-mm rotor was used to acquire all spectra. Samples were packed into 4-mm Zirconia rotors and sealed with Kel-F caps. Spectra were acquired using variable amplitude

Figure 2. SECeUVeESIeMS chromatogram of poly-Faldaprevir (~4-6 min) and Faldaprevir (~7.3 min): upper trace: UV absorbance at 265 nm; lower trace: total ion current (ESIpositive mode). Slight differences in retention times reflect in-line arrangement of the mass spectrometer and UV detector. Different peak shapes for the earlier eluting polymeric species are consistent with a nonlinear electrospray response in the ion source for different chain lengths versus in the UV detector.

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

1883

Table 1 Elemental Analysis of the Polymer Impurity (poly-Faldaprevir_Na) and Faldaprevir_Na

Polymer impuritya,b (%) Faldaprevirc (%) a b c

Carbon

Hydrogen

Nitrogen

Oxygen

Bromine

Sulfur

Sodium

49.2

5.5

8.4

20.1

8.0

3.2

2.5

53.9

5.4

9.4

16.2

9.0

3.6

2.6

Solvias AG, Switzerland (elemental analysis). The small discrepancy in experimental elemental analysis values is likely due to the presence of residual solvent. Theoretical elemental analysis calculation for Faldaprevir.

cross-polarization,8-11 magic angle spinning (MAS),12,13 and highpower proton decoupling with 2-phase pulse modulation.14-17 Contact times of 2 ms were used to acquire all spectra. The MAS spinning yr ¼ 12.0 kHz and the proton decoupling field was approximately 50 kHz. Recycle delays, based on saturation of 13C signal intensity, were sufficiently long to allow full relaxation of 13C signal. All data were collected at ambient temperature. Carbon-13 chemical shifts were quoted with respect to tetramethylsilane at 0 ppm relative to the external reference of adamantane (29.5 ppm). The CPPI spectral editing techniques were used to identify resonances corresponding to C, CH, CH2, and CH3 groups.18-20 These editing techniques rely on dipolar coupling of nuclei to differentiate between the multiplicity of the carbon atoms. Two-dimensional (2D) CP heteronuclear correlation (CPHETCOR) spectra between 1H and 13C nuclei were obtained using 4-mm DVT CP/MAS probe, MAS spinning yr ¼ 12.0 kHz with frequency-switched LeeeGoldburg homonuclear decoupling at 50 kHz.21 Ramp CP transfers were used for HETCOR experiments with durations ranging from 100 ms to 1 ms. Solution NMR Experiments The 1H NMR experiments were performed on a Bruker-Biospin AV 600 spectrometer operating at 600.2 MHz and using a 1.7-mm CP TCI Z gradient probe. Indirect detection 1H13C NMR experiments were performed with 13C operating at 150.92 MHz. 1H and 13C chemical shift assignments were referenced to dimethyl sulfoxided6 at 2.5 and 39.5 ppm, respectively; 1D 1H data were obtained using a spectral width of 12,376 Hz, a 10 pulse, and a 2 s relaxation delay. All 1D data were processed using exponential multiplication and

applying line broadening of 0.05 Hz. All experiments were performed nonspinning at 333K unless otherwise noted. 13 C data were obtained using a 2D 1H-13C heteronuclear single quantum correlation (24) with 1H/X correlation via double inept transfer. The spectra were collected in a phase-sensitive mode using Echo/Antiecho-TPPI gradient selection with a sine.1 shape pulse and decoupling during acquisition. Spectra were obtained (1K complex points in f2, 256 real points along f1) using globally optimized alternating-phase retangular pulses decoupling, a relaxation time of 1.5 s per transient, and a spectral width of 8012 Hz in f2 and 30,187 Hz in f1. The carrier frequency for 13C was 15,092 Hz. Delays were used for a 1J (13C-1H) coupling constant of 145 Hz. These data were processed by zero filling to 1K complex points in f2 and 1K real points in f1, followed by QSINE apodization in both dimensions. Fourier transformation and phasing were applied to achieve positive absorption in both f1 and f2 dimensions. 2-D 1H-1H gradient nuclear Overhauser effect correlation spectroscopy22,23 data were obtained (2K complex points in f2, 512 real points along f1) using a relaxation time of 1.5 s per transient and a spectral width of 8389 Hz in both dimensions. A 30 ms mixing time was used. Time proportional phase incrementation was used to achieve quadrature detection in f1. These data were processed by zero filling to 1K real points in f1, followed by cosine apodization in both dimensions, Fourier transformation, and phasing to achieve pure absorption in both f1 and f2 dimensions. 2-D 1H-1H cleanetotal correlation spectroscopy24 data were obtained (2K complex points in f2, 256 real points along f1) using a relaxation time of 1.0 s per transient and a spectral width of 8680 Hz in both dimensions. A 50.5 ms mixing time was used. Time proportional phase incrementation was used to achieve quadrature

Figure 3. SECeESI mass spectrum of the SEC peak at 4.9 min obtained from the polymeric impurity.

1884

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

Figure 4. Low mass region of the SECeESI mass spectrum of the SEC peak at 4.9 min (see Fig. 3). The chemical structure of Faldaprevir provides rationalization of the observed mass spectrometry fragments of the polymeric impurity that are related to the API.

Figure 5. Solution and ssNMR spectra of Faldaprevir. (a) 13C NMR of Faldaprevir, recorded in dimethyl sulfoxide-d6 at 300K. (b) 13C ssNMR of crystaline Faldaprevir, recorded at ambient temperature. Chemical shift assignments are made from comparison with solution NMR assignments, spectral editing experiments, and HETCOR experiments. (c) 13C ssNMR of amorphous Faldaprevir, recorded at ambient temperature.

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

1885

relaxation time of 0.5 s, and a spectral window 36,232 Hz. The 1D high-temperature 13C, 1H decoupled spectrum of the polymer was collected at 333K using a 13C pulse width of 90 degrees, a relaxation time of 1.0 s, and a spectral window 37,879 Hz. The data were processed by zero filling to 32K or 64K complex points, applying a line broadening of 1.0 Hz, Fourier transformed, phased for pure absorption, and baseline corrected using an automatically calculated fourth-order polynomial function.

SECeESI Mass Spectrometry The SECeESI mass spectrometry experiments were performed on a Waters quadrupole time of flight Synapt G2-S instrument, coupled to an Acquity UPLC system. For SEC analysis, the residue arising from the filtrate of the Faldaprevir drug substance in methanol was dissolved in dimethyl sulfoxide/acetonitrile/water/ trifluoroacetic acid (25:95:5:0.1) to yield a concentration of approximately 0.7 mg/mL. Chromatographic separation of the impurity was achieved on a Waters BEH-125SEC column (300  4.6 mm) at ambient temperature, mobile phase (isocratic):acetonitrile/ water/TFA (95:5:0.1); and flow rate: 0.3 mL/min. UV absorbance detection was performed using a photodiode array detector operated in the range of 220-300 nm. The electrospray source of the QTOF mass spectrometer was run in the positive ion mode with the acquisition mass range from 300 to 5000. The mass spectrometer was calibrated using 0.1% sodium formate in acetonitrile/water 1:1, and lockmass acquisition was

Figure 6. Comparison of 13C ssNMR spectra (A) amorphous Faldaprevir; spectrum (red) shows starred peaks from vinyl cyclopropane moiety which are missing in the polymer. (B) poly-Faldaprevir; spectrum (black) shows the presence of 3 new peaks indicated by arrows consistent with the structure of the polymer backbone.

detection in f1. These data were processed by zero filling to 512 real points in f1, followed by cosine apodization in both dimensions, Fourier transformation, and phasing to achieve pure absorption in both f1 and f2 dimensions. The 1D 13C, 1H decoupled spectrum of Faldaprevir sodium salt was collected at 300 K using a 13C pulse width of 90 degrees, a

a

b

c

Figure 7. (a) 13C CP/MAS ssNMR (MAS 12 kHz) spectrum of amorphous Faldaprevir; (b) ssNMR (MAS 12 kHz) spectrum of amorphous poly-Faldaprevir.

13

C CPPI ssNMR (MAS 12 kHz) spectrum of amorphous poly-Faldaprevir. (c)

13

C CP/MAS

1886

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

Figure 8. 13C 1H HETCOR ssNMR (MAS 12 kHz) spectrum of amorphous poly-Faldaprevir. Peak at 65 ppm has no attached protons consistent with carbon 220 . Peak at 129 ppm shows correlation to protons in the olefinic region consistent with carbons 240 and 250 .

OMe Br N N

O

NH

O O

S

O

N

N H O

O

O

25'

23'

26' R1

22'

24' NaO

42B

21

O

1 O 3

ONa

40

NH

HN 4

20

42A

2

S N

19 15

7 N 8 Br

O

18

5 6

n

NH

13 12

9 MeO 39

16 14 O

10

N

O 33

34 32 H N

35 O

31

17

36

37B

O 30

38B 37A

38A

11

Figure 9. Numbered chemical structure of poly-Faldaprevir. The polymer comprised repeating units of the API.

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

1887

Figure 10. (a) 13C NMR of Faldaprevir, recorded in dimethyl sulfoxide-d6 at 300K. (b) 13C NMR of purified poly-Faldaprevir (API content <0.3%), recorded in dimethyl sulfoxide-d6 at 333K after addition of approximately 20 mg of Cr(acac)3.

performed using 0.2 mg/mL leucine enkephalin in acetonitrile/ water 1:1 at m/z 556.2771.

Results and Discussion Because the impurity could not be examined by reversed-phase HPLC, likely due to its size and low solubility, SEC was used to detect and separate the polymer from the API. In this case, the Faldaprevir API was observed to elute as a sharp peak with retention time of approximately 7.3 minutes, reflecting its lower, monodisperse molecular weight versus that of the polymeric species eluting as a broad peak between 4 and 6 minutes and characteristic of a molecular weight distribution (Fig. 2). The material was found to be amorphous based on X-ray powder diffraction data. A mass distribution of approximately 30-200 kDa was observed by matrix-assisted laser desorption ionizationetime of flight mass spectrometry. The high molecular weight of the impurity was further corroborated by an ultracentrifugation technique showing peaks at approximately 90 kDa. The structural relationship between the polymer and the API became evident by comparison of the experimental elemental analysis of the impurity with the theoretical values of the API. The results are listed in Table 1. The values for the polymer largely agree with the atomic composition of the API, suggesting the material is a polymer of Faldaprevir with the molecular mass of Faldaprevir

conserved per repeating unit. The small discrepancy in the polymer analysis is likely due to the presence of residual solvent. Quantitation of sodium by atom absorption spectroscopy indicated that the sodium concentration in the impurity was very similar to that of the Faldaprevir sodium salt. IR spectra of the impurity showed good agreement to that of Faldaprevir, further supporting the structural similarity between the impurity and the API. Both 1H NMR and 13C NMR signals obtained from solution NMR spectra of the impurity in dimethyl sulfoxide-d6 revealed severe line broadening; hence, analysis of the data was highly challenging. Most of the sharp peaks in the spectra were caused by residual amounts of Faldaprevir in the polymer. The line broadening was consistent with the polymeric nature of the impurity. Electrospray mass spectrometry coupled to SEC (SECeESIeMS) enabled partial characterization of the SEC peak for the highemolecular weight impurity. The ESI spectra for the precipitate peak showed a broad distribution of unresolved masses in the m/z range between 2000 and 4000 (Fig. 3). Because electrospray ionization produces multiply charged ions for larger molecules, the data support these signals as being derived from compounds having much higher molecular weights than the detected ions. However, the unresolved mass peaks and, therefore, the inability to derive the charge states of the ions did not allow calculation of the molecular weight of any of these large molecules. The MS data did show the presence of large parts of the API chemical structure

1888

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

Figure 11. 1H NMR spectrum (600.2 MHz) of poly-Faldaprevir in dimethyl sulfoxide-d6 with trace amount of deuterated acetic acid at 333K.

within the polymeric compound where several characteristic fragments of Faldaprevir were detected (Fig. 4). With the nature of the process impurity established as a polymer, ssNMR studies were carried out to further characterize the impurity in solid form. To enable structural analysis of the ssNMR spectrum of the polymer, it was necessary to obtain ssNMR characterization of the crystalline form of Faldaprevir. Carbon 13 chemical shift assignments of crystalline Faldaprevir were made by comparing the 13 C CP/MAS resonances with unequivocal 13C assignments obtained from solution NMR. The 13C solution assignments of Faldaprevir (Fig. 5a) showed good agreement with the ssNMR 13C spectrum

of the crystalline solid. Assignments of carbon resonances in the CP/MAS spectrum of the crystalline material (Fig. 5b) were further refined using CPPI and 13C,1H HETCOR experiments. The crystalline API was subsequently converted to amorphous material, and a CP/MAS spectrum was acquired (Fig. 5c). As expected, the carbon resonances were severely broadened. Compared with the crystalline CP/MAS spectrum (Fig. 5b), some clear correlations could be made between the crystalline and amorphous spectra and specific assignments. In particular, the 13C chemical shifts for C25 and C22 could be clearly assigned in the amorphous state due to their distinct chemical shifts free of

Table 2 1 H and 13C NMR Chemical Shifts of Poly-Faldaprevir in Dimethyl Sulfoxide-d6 Atom#

1

H ppma

13

C ppmb,c

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

2.803 e e e 7.965 e e e e e 7.143 7.979 e e 7.389 5.295 3.484, 4.321 4.704 2.303-2.542

33.4 (33.6) 175.2 (176.1) 158.0 (158.6) 114.0 (113.3) 153.3 (154.4) 148.4 (149.2) 145.7 (146.7) 108.1 (108.8) 157.1(157.2) 112.2 (112.5) 122.1 (122.4) 115.8 (116.2) 160.3 (160.3) 97.6 (98.4) 76.6 (76.6) 52.9 (52.9) 58.3 (58.7) 34.0 (34.0)

Atom No.

1

H ppma

20 21 220 230 , 260 240 , 250 30 31 32 33 34 35 36 37 38 39 40 42

e e e 2.529-2.895 5.271 0.872 e 4.116 e e e 4.264-4.605 1.114-1.599 1.198-1.554 3.895 e 1.105

13

C ppmb,c

170.4 (169.9) e 62.4 n/o 127.3 25.8 (26.3) (38.5) 58.7 (59.1) 169.3 (170.0) 155.6 (156.2) 76.0 (76.4) 31.7, 31.4 (32.0, 32.2) 22.6 (23.2, 23.1) 56.4 (56.7) 172.8 (171.9) 18.6 (19.2)

n/o, not observed. a Referenced to dimethyl sulfoxide-d6 at 2.5 ppm with trace amount of acetic acid-d4, temperature 333K. b Referenced to dimethyl sulfoxide-d6 at 39.5 ppm after addition of approximately 20 mg of Cr(acac)3 temperature 333K. c Assignment of poly-Faldaprevir, based on comparison of 1D 13C spectra with the fully assigned13C spectrum of Faldaprevir Na. Values for Faldaprevir Na are reported in parenthesis.

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

1889

O O H N

R1

H N

R1

ONa

ONa O

O

O

X

O X R1

R1

O

R1 O

O

NH

NH

R1

NH

O

O

X

X O

ONa

O O

NH

NaO O

NaO

O NaO R1 O ONa

NH

Br

N N

MeO

O

HN

O

S

X

n

O NaO

O

NH

O

R1 =

N O

O

O R1

NH O

Figure 12. Free radical mechanism illustrating the polymerization of vinyl cyclopropane leading to the formation of poly-Faldaprevir.

resonance overlap. These resonances were important in the structure analysis because they comprised the segment of the API involved in the polymeric backbone formation. When the spectrum of amorphous Faldaprevir was compared with the spectrum of poly-Faldaprevir, key distinctions became immediately obvious. Resonances corresponding to C25 and C22 in the spectrum of amorphous Faldaprevir were missing from the CP/ MAS spectrum of the impurity. In addition, 3 new resonances were present in the polymer spectrum at 129, 65, and 38 ppm (Fig. 6). The multiplicities of these carbon resonances were obtained from

the CPPI spectral editing experiment where olefinic methine carbons at 129 ppm are nulled, quaternary carbons at 65 ppm are phased up, and methylene carbons at 38 ppm are phased down (Fig. 7). The 13C,1H HETCOR experiment established protonecarbon connectivity for each of these new resonances showing olefinic and aliphatic correlations for the carbons at 129 and 38 ppm, respectively, whereas quaternary carbons at 65 ppm showed no proton correlation (Fig. 8). These chemical shifts were consistent with resonances expected for the polymeric backbone structure shown in Figure 7c where a cyclopropyl ring opening occurs to produce an

1890

N.C. Gonnella et al. / Journal of Pharmaceutical Sciences 105 (2016) 1881e1890

olefin with 2 adjacent methylene moieties. The chemical shifts for olefin carbons C240 and C250 at 129 ppm, quaternary carbon C220 at 65 ppm, and methylene carbons C260 and C230 at 38 ppm and their respective spin multiplicities were consistent with the proposed backbone structure. The full structure of the polymer showing 2 repeating units is displayed in Figure 9. The structure of the polymeric backbone was subsequently corroborated by solution NMR. The 13C NMR solution spectrum of the polymer was achieved with the addition of relaxation agent Cr(acac)3 and heating to 333K to enable observation of the carbon resonances. The observed chemical shifts were consistent with the structure of Faldaprevir with the exception of the vinyl cyclopropane moiety. As in the solid-state NMR (ssNMR) spectra, the key resonance for C25 in Faldaprevir was noticeably missing in the 13C spectrum of the polymer. Instead a new olefinic peak appeared at approximately 128 ppm in agreement with the ssNMR results (Fig. 10). Solution 1H NMR spectra of the polymer were also obtained. To improve peak linewidth, the spectra were run in the presence of a trace amount of deuterated acetic acid with heating to 333K. The 1D 1 H NMR spectrum of the polymer is shown in Figure 11. Full proton NMR assignments were made from a combination of gradient nuclear Overhauser effect correlation spectroscopy experiments and gradient heteronuclear single quantum correlation experiments (Supplementary Data). The 1H and 13C chemical shifts for polyFaldaprevir in dimethyl sulfoxide-d6 are given in Table 2. For both solid-state and solution NMR spectra of poly-Faldaprevir, comparison with the chemical shifts of Faldaprevir itself helped to guide the assignment of the polymer. Final corroboration of the polymeric impurity structure was accomplished by comparison of ssNMR 13C CP/MAS spectrum of the isolated polymer with the synthesized polymer. The synthesized polymer was made from Faldaprevir in the presence of a strong free radical initiator azobisisobutyronitrile.25 Comparison of the ssNMR spectra of the isolated polymeric impurity and the synthesized polymer showed an identical match between the 2 substances further supporting the poly-Faldaprevir structure (see Supplementary Data). Mechanism of Polymer Formation Polymerization of vinyl cyclopropane is known to occur via a radical mechanism where polymers are produced through a single ring-opening process.25 Vinyl cyclopropanes having free radical stabilization groups on the cyclopropane ring have been reported to afford relatively highemolecular weight polymers through a radical ring-opening process.25,26 Faldaprevir contains a vinyl cyclopropane group with a carboxylic acid and amide nitrogen attached to the cyclopropane ring. This synergistic (captodative) effect of electron donating and electron withdrawing substituents provides a free radical stabilization system at C220 that helps to promote formation of highemolecular weight polymers. The ring opening and composition of the 1,5 adduct yields a polymeric backbone containing an olefin flanked by two methylene groups. Although distinction between cis and trans orientation of the methylene groups flanking the olefin could not be determined, studies of polymerization of other vinyl cyclopropanes have reported mixtures of cis/trans isomers.27 The mechanism for polymerization of Faldaprevir is shown in Figure 12. Conclusions All analytical data showed the unknown impurity to be a polymer of Faldaprevir. The data showed the repeating units of the polymer remained unchanged relative to the API with most of the

structural functionality preserved. Polymerization was found to occur at the vinyl cyclopropane substituent via a free radical initiation mechanism. SEC/MS and other supporting technologies were critical in establishing the impurity as a polymer with full characterization of the backbone structure accomplished using ssNMR. Comparison of the ssNMR spectra with solution NMR data of the API, both crystalline and amorphous, were essential in spectral analysis and confirmation of the structure.

References 1. Joyce MA, Tyrrell DLJ. The cell biology of hepatitis C virus. Microbes Infect. 2010;12:263-271. 2. Tang H, Grise H. Cellular and molecular biology of HCV infection and hepatitis. Clin Sci. 2009;117:49-65. 3. Chatel-Chaix L, Baril M, Lamarre D. Hepatitis C virus NS3/4A protease inhibitors: a light at the end of the tunnel. Viruses. 2010;2:1752-1765. 4. Raney KD, Sharma SD, Moustafa IM, Cameron CE. Hepatitis C virus nonstructural protein 3 (HCV NS3): a multifunctional antiviral target. J Biol Chem. 2010;285:22725-22731. 5. Lamarre D, Anderson PC, Bailey M, et al. An NS3 protease inhibitor with antiviral effects in humans infected with hepatitis C virus. Nature. 2003;426:186-189. 6. LaPlante SR, Llinas-Brunet M. Dynamics and structure-based design of drugs targeting the critical serine protease of the hepatitis C virus - from a peptidic substrate to BILN 2061. Curr Med Chem Anti-Infect Agents. 2005;4:111-132. 7. Busacca CA, Wei X, Haddad N, et al. Practical large-scale synthesis of the hepatitis C virus protease inhibitor BI 201335. Asian J Org Chem. 2012;1:80-89. 8. Pines A, Gibby MG, Waugh JS. Proton-enhanced NMR of dilute spins in solids. J Chem Phys. 1973;59:569-590. 9. Hartmann SR, Hahn EL. Nuclear double resonance in the rotating frame. Phys Rev. 1962;128:2042-2053. 10. Pines A, Gibby MG, Waugh JS. Proton-enhanced nuclear induction spectroscopy. Method for high-resolution NMR of dilute spins in solids. J Chem Phys. 1972;56:1776-1777. 11. Peersen OB, Wu X, Smith SO. Enhancement of CP-MAS signals by variableamplitude cross polarization. Compensation for inhomogeneous B1 fields. J Magn Reson A. 1994;106:127-131. 12. Andrew ER. Narrowing of NMR spectra of solids by high-speed specimen rotation and the resolution of chemical shift and spin multiplet structures for solids. Prog Nucl Magn Reson Spectrosc. 1971;8:1-39. 13. Andrew ER, Bradbury A, Eades RG. Removal of dipolar broadening of nuclear magnetic resonance spectra of solids by specimen rotation. Nature. 1959;183: 1802-1803. 14. Schaefer J, Stejskal EO. Carbon-13 nuclear magnetic resonance of polymers spinning at the magic angle. J Am Chem Soc. 1976;98:1031-1032. 15. Schaefer J, Stejskal EO, Buchdahl R. High-resolution carbon-13 nuclear magnetic resonance study of some solid, glassy polymers. Macromolecules. 1975;8: 291-296. 16. Stejskal EO, Schaefer J, Waugh JS. Magic-angle spinning and polarization transfer in proton-enhanced NMR. J Magn Reson. 1977;28:105-112. 17. Bennett AE, Rienstra CM, Auger M, Lakshmi KV, Griffin RG. Heteronuclear decoupling in rotating solids. J Chem Phys. 1995;103:6951-6958. 18. Wu X, Burns ST, Zilm KW. Spectral editing in CPMAS NMR. Generating subspectra based on proton multiplicities. J Magn Reson A. 1994;111:29-36. 19. Wu X, Zilm KW. Methylene-only subspectrum in CPMAS NMR. J Magn Reson A. 1998;104:119-122. 20. Wu X, Zilm KW. Complete spectral editing in CPMAS NMR. J Magn Reson A. 1993;102:205-213. €rster H, de Groot HJH. High-field and high-speed CP-MAS 13C 21. van Rossum BJ, Fo NMR heteronuclear dipolar-correlation spectroscopy of solids with frequencyswitched Lee-Goldburg homonuclear decoupling. J Magn Reson A. 1997;124: 516-519. 22. Jeener J, Meier BH, Bachmann P, Ernst RR. Investigation of exchange processes by two-dimensional NMR spectroscopy. J Chem Phys. 1979;71:4546-4553. 23. Wagner R, Berger S. Gradient-selected NOESY-a fourfold reduction of the measurement time for the NOESY experiment. J Magn Reson A. 1996;123 A: 119-121. 24. Griesinger C, Otting G, Wuethrich K, Ernst RR. Clean TOCSY for proton spin system identification in macromolecules. J Am Chem Soc. 1988;110:78707872. 25. Sanda F, Takata T, Endo T. Synthesis and radical ring-opening polymerization of a vinylcyclopropane bearing a cyclic carbonate moiety, l-vinyl-5,7-dioxaspiro [2.5l octan-6-one. Macromolecules. 1994;27:3986-3991. 26. Sugiyama J, Ohashi K, Ueda M. Free radical polymerization of volume expandable vinylcyclopropane. Macromolecules. 1994;27:5543-5546. 27. Okazaki T, Sanda F, Endo T. Synthesis and radical ring- opening polymerization behavior of vinylcyclopropane bearing six-membered cyclic acetal moiety. J Polym Sci Part A. 1996;34(10):2029-2035.