To understand the effect of pH, an initial series of simulated Victoza® (liraglutide) drug products were prepared following the formulation described in the package insert with the pH varying ± 0.4 from the prescribed formulation pH (8.15). (4) Proton NMR of these samples were visually assessed for spectral differences. As the pH was increased, a singlet around 7.8 ppm appeared to gradually shift upfield by 0.10—0.15 ppm as shown in Fig. 2. This resonance, an imidazole proton (Hε) of the N-terminal histidine (7), has been previously identified as an amino acid known to have proton chemical shift changes as a function of pH in the vicinity of the pKa of the sidechain. (8) Similarly, as pH is increased across the range, the multiplet at 3.0 ppm also undergoes change, and this resonance can be putatively assigned to represent the His Hβ protons. (7) Hence, as pH changes across this range, chemical shift and pattern for these protons reflect differing degrees of protonation of the histidine residue (pKa ~ 7). (8) It is important to note that these spectral changes related to pH are completely reversible and unlikely to be the result of structural changes as the N-terminus of peptides and proteins are usually not involved in any HOS.

Fig. 2

Overlaid proton NMR spectrum of a simulated Victoza® (liraglutide) formulation at various pH conditions over a ± 0.4 pH range of the average formulation pH (8.15): pH 7.80 (purple), pH 7.93 (blue), pH 8.09 (green), pH 8.11 (gold), pH 8.31 (orange) and pH 8.58 (pink). a expansion of the N-terminal histidine Hε proton chemical shift region. b expansion of the N-terminal histidine Hβ proton chemical shift region

To illustrate the influence of average pH and pH variability on PCA “sameness” evaluations, we prepared the following series of simulated Victoza® (liraglutide) sample cohorts (4 samples each) representing the pH control an ANDA applicant might reasonably target based on the package insert information:

pH 8.13 ± 0.02 (RLD – Lower pH Variability),

pH 8.16 ± 0.10 (Generic – Higher pH Variability),

pH 8.13 ± 0.03 (Generic – Lower pH Variability),

pH 7.95 ± 0.03 (Generic – Lower Average pH), and,

pH 8.33 ± 0.02 (Generic – Higher Average pH).

The overall average pH and pH variability of each Cohort is presented in graphic form in the supporting information.

Following the FDA’s published best practices recommendation for carrying out PCA, the proton NMR data obtained on drug product sample comparison sets were analyzed by segregating sample spectral data into uniform “bins” of no more than 0.02 ppm width excluding resonance regions involving excipient protons (see supporting information for details). (2) The RLD – Lower pH Variability and the Generic – Higher pH Variability cohorts were designed to exemplify the lot-to-lot variability seen among RLD samples (Fig. 3a) and representative of client samples (Fig. 3b), respectively. Although the Generic cohort met pH specifications which would reasonably be viewed as acceptable from a human safety perspective, the pH range of the RLD cohort was more tightly controlled. Figure 4 shows a comparison of the cohorts where the proton NMR spectra are qualitatively similar with the exception of the two regions that correspond to the N-terminal histidine resonances mentioned above. Although the Generic cohort had an average pH close to that of the RLD cohort, its higher variability (± 0.10 pH units compared to ± 0.02 pH units), was sufficient to cause the spectral differences derived solely from pH to yield a DM of 4.28 for the first three (3) principal components (PCs).

Fig. 3

a Overlaid proton NMR spectra of the histidine resonances for RLD – Lower pH Variability cohort: pH 8.15 (blue), pH 8.14 (green), pH 8.13 (gold) and pH 8.11 (pink). b Overlaid proton NMR spectra of the histidine resonances for Generic – Higher pH Variability cohort: pH 8.08 (blue), pH 8.11 (green), pH 8.13 (gold) and pH 8.31 (pink)

Fig. 4

Overlaid proton NMR spectra of the RLD – Lower pH Variability cohort (red) and Generic – Higher pH Variability cohort (purple) samples with lot-to-lot variability. (Upper: overlaid histidine resonances; Lower: full overlaid spectrum cut to remove excipient resonances.) The resulting DM (3 PCs) = 4.28

Further PCA was conducted comparing the cohorts of RLD – Lower pH Variability to Generic – Lower pH Variability, wherein the Generic – Lower pH Variability cohort was generated by adjusting the pH of four (4) of the samples prepared and presented in Fig. 2 (pH 7.93, 8.09, 8.31, and 8.58 adjusted to pH 8.11, 8.13, 8.16 and 8.12, respectively). Following pH adjustment, the final pH range for the Generic – Lower pH Variability cohort was similar to that of the RLD – Lower pH Variability cohort. The His Hε singlet and His Hβ multiplets discussed above appeared nearly superimposable to those of the RLD – Lower pH Variability cohort resulting in a DM of 1.02 for the first three (3) PCs (see Fig. 5).

Fig. 5

Overlaid proton NMR spectra of the RLD – Lower pH Variability cohort (red) and Generic – Lower pH Variability cohort (blue) samples with lot-to-lot variability. (Upper: overlaid histidine resonances; Lower: full overlaid spectrum cut to remove excipient resonances.) The resulting DM (3 PCs) = 1.02

Additionally, PCA was conducted comparing the RLD – Lower pH Variability to Generic – Lower Average pH cohorts and then RLD – Lower pH Variability to Generic – Higher Average pH cohorts to simulate tight control of an ANDA applicant’s samples at a pH 0.10 – 0.15 below or above the average pH of the RLD formulation, respectively. In both cases the DM values of 11.04 and 5.31, respectively, far exceeded the recommended threshold for the two drug products to be considered the same (see Figs. 6a and 6b, respectively).

Fig. 6

a Overlaid proton NMR spectra of the RLD – Lower pH Variability cohort (red) and Generic – Lower Average pH cohort (teal) samples with lot-to-lot variability. The resulting DM (3 PCs) = 11.04. b Overlaid 1D proton NMR spectra of the RLD – Lower pH Variability cohort (red) and Generic – Higher Average pH cohort (light blue) samples with lot-to-lot variability. The resulting DM (3 PCs) = 5.31. In both a and b, insets correspond to overlaid histidine resonances and lower spectra display the full overlaid data, cut to remove excipient resonances

Based on the above findings, PCA of the proton NMR spectra of simulated Victoza® (liraglutide) drug product formulations were found to be quantitatively and reversibly influenced by both the overall average pH and the variability of pH from lot-to-lot. Table I summarizes the DM, average pH, and pH variability of this study.

Table I Average Liraglutide Formulation pH with Wariability and Resulting DM of the First Three (3) PCs

These results show that both the average pH and pH variability are relevant factors when evaluating “sameness” using PCA for Victoza® (liraglutide) and if not tightly controlled, then a failure to meet the FDA de facto DM threshold of < 3.3 is more likely to be observed.

Although the study described above is relevant to any formulation involving liraglutide, it should also be relevant to other peptide therapeutics containing terminal histidine residues, such as semaglutide (the active component of Ozempic®, Wegovy® and Rybelsus®) where the sequence of semaglutide differs from liraglutide by two amino acids and differences in the “fatty acid” residue. We therefore extended the study to Wegovy® (semaglutide) and found analogous behavior derived from shifting of assignable terminal histidine proton resonances with minor changes of pH (see Fig. 7). The shifting resonances were not limited to just the His Hβ and Hε protons, but also the His Hδ proton which was now spectrally visible within this formulation as no excipient resonances obscured the His Hδ singlet at 6.95 – 7.10 ppm. Moreover, other peptide therapeutics that contain other ionizable amino acid sidechains and are formulated at a pH close to the pKa of those amino acid sidechains (e.g. aspartic acid, glutamic acid, with pKa values of 3.8 and 4.3, respectively) should also be expected to display this behavior (8), and we confirmed this experimentally. Figure 8 illustrates this for Forteo® (teriparatide) where a series of proton NMR data show similar, if not more complex, differences as a function of pH above and below the drug product target pH of 4.

Fig. 7

Expansion of the N-terminal histidine resonances in the overlaid proton NMR spectra of the simulated Wegovy® (semaglutide) formulation at various pH conditions over a ± 0.4 pH range of the formulation pH (7.4): pH 6.95 (pink), pH 7.18 (orange), pH 7.36 (green, pH 7.58 (teal) and pH 7.83 (purple). Identification of the shifting N-terminal histidine resonances designated by the teal (Hε), blue (Hδ) and red (Hβ) arrows

Fig. 8

Expansion of the amide (N–H) and aromatic side chain resonances in the overlaid proton NMR spectra of the simulated Forteo® (teriparatide) formulation at various pH conditions over a ± 0.4 pH range of the formulation pH (4); pH 3.64 (pink), pH 3.78 (orange), pH 3.97 (green), pH 4.21 (teal) and pH 4.37 (purple)