A Workflow Approach

The ability to understand the levels of pharmaceutical impurities in drug manufacturing is not only a regulatory necessity, but a business imperative. However, the analytical determination of impurities is often timeconstraining and resource-consuming. It requires a range of mass spectrometry capabilities as well as software to facilitate the data processing of these complex impurity data sets.

A multidisciplinary approach to im-purity analysis was explored using a systematic workflow that is capable of the specific and sensitive detection and determination of impurities that are present in the quetiapine hemifumarate Active Pharmaceutical Ingredient (API) drug substance.

The designed approach incorporated chromatographic resolution, impurity identification, and rapid structural elucidation facilitated by intelligent and user-friendly software. This work flow-based methodology improved the ability to evaluate known and unknown impurities in a pharmaceutical drug substance.



Experimental Conditions
The instrument system that was used for this analysis was an Ultra Performance Liquid Chromatography (UPLC) system that was fitted with an a UPLC BEH C18 100 x 2.1 mm, 1.7 μm column. The mass spectrometer was fitted with an Electro-Spray (ES) positive source.

The software processed chromatographic and exact mass data to report retention times, peak area, mass accuracy, and isotope distribution values for m/z found. Elemental compositions were confirmed for known impurities and proposed for unknown impurities. The software also performed a fragment analysis, correlating the precursor ion information of the low-energycollision Mass Spectrometry (MS) scan to that of the product ion information of the high-energy MS scan.

The high-collision-energy MS scan data was imported into mass fragmenting software, a chemically intelligent structure elucidation tool for the structural assignment of product ions. This is where the structural fragmentation pathways of the impurity compounds were proposed, based on the likelihood of breaking certain bonds.

The workflow approach that is shown in Figure 1 may require several iterations to determine the accurate result for the unknown peak of interest. Evaluation of the data can be more involved depending on the complexity of the compound; however, the general workflow remains constant. The benefit of this approach is that it provides a systematic data-driven association to correlate the variety of data that is acquired by the two scan functions that are generated by MSE experiments.

The software enables the application of userdefined filters to configure how the reported data is viewed in the browser window. Some useful techniques to apply meaningful data filters were identified by investigating proper integration parameters. Mass defect filters, the dealkylation tool, spectrum intensity thresholding, and selection of components relative to the compound in the elemental composition tab, all proved useful in displaying more confident data.

For example, to obtain elemental composition for every peak that is found in a chromatogram, the analyst would typically need to combine MS scans and perform background subtraction for each peak of interest and then generate individual elemental composition reports.

To streamline this process, the browser populated all impurity peaks integrated in the Time-of-flight (Tof)-MS ES+ chromatographic trace with associated elemental compositions, mass accuracy, and isotope pattern scoring - and displayed the results in a single window (Figure 2).



Evaluating Impurities
Evaluation of the unknown impurity peaks by exact mass and elemental composition of quetiapine hemifumarate using data acquisition, processing, and interpretation software indicated that the mass accuracy of the API quetiapine was reported to be 0.4 ppm. A total of 80 impurity peaks were listed. Upon adjustments to integration and data filtering, 44 peaks were found to be relevant. Non-relevant peaks were observed to be anomalies of initial integration of noise and peaks with extremely low-level response in UV and MS detection.

Ten known impurities were observed with an average mass accuracy of 1.3 ppm. Two known masses, 398.19xx and 412.20xx, had three and four separate retention times listed respectively. The masses with multiple chromatographic retention times, which indicated possible structural isomers, were [M+H] = 398.19xx and observed four peaks, three of which met the reporting threshold.

The observed [M+H] = 398.1900, 398.1896, 398.1913 at Retention Times (RT) of 10.75 min, 11.08 min, and 11.58 min, with measured mass accuracies of 0.5 ppm, 1.5 ppm, and 2.8 ppm, respectively, resulted in an identified elemental composition of C22H28N3 O2S. Also [M+H] = 412.20xx observed five peaks, four of which met the reporting threshold.

The observed [M+H] = 412.2066, 412.2048, 412.2065, and 412.2059 at RT of 12.50 min, 12.76 min, 13.06 min, and 13.97 min, with measured mass accuracies of 1.7 ppm, 2.7 ppm, 1.5 ppm, and 4.1 ppm, respectively, resulted in an identified elemental composition of C22H28N3 O2S.

In terms of the unknowns that were identified, of 21 entries for 15 chromatographic peaks, peaks identified as doubly charged species were [M+2H]2+ = 353.1512, [M+H]+ 705.3013 at RT = 17.20 min; [M+2H]2+ = 309.1256, [M+H]+ 617.2514 at RT = 17.36 min and; [M+2H]2+ = 684.2089 with a large fragment at [M+H] = 382.3485.

Peaks with multiple m/z ions; which could be possible coelutions, included; Peak RT = 15.96 min observed [M+H] = 510.2073, 299.1627, 399.2523 (three intense m/z values) and Peak RT = 17.42 min observed [M+H] = 653.3301, 592.1955 (two intense m/z values).

From these, it is possible to generate and assess the data in the fragment analysis function of the data acquisition, processing and interpretation software by determining the relationship to the API based on the MSE precursor/product ion information.

Fragment Analysis
The fragment analysis tool aligned the high and low collision energy data that were simultaneously collected during the MSE acquisition. The resulting information was displayed in a collective window where the precursor and the collisioninduced product ions were evaluated spectrally and presented chromatographically.

The fragment analysis window allowed for numerous iterations by the analyst to assess common fragment ions between peaks of interest (Figure 3). Commonalities were observed between known impurity structures and fragmentation patterns that aided in proposing the structures of other unknown impurity entities.



An assessment of the common fragment ions of quetiapine identified the major fragment ions to be m/z 279, 253, 221, and 158. Of these, Extracted Ion Current (XIC) of precursor 279 was identified in 22 impurity peaks; XIC of precursor 253 was identified in 25 impurity peaks; XIC of precursor 221 was identified in 23 impurity peaks and; 14 impurity peaks were deemed not to be directly related to the parent.

Structural elucidation was enabled by chemically-intelligent software that combined the aligned high- and low-collisionenergy data in the fragment analysis window with the user's input about a hypothesized structure. Before performing the elucidation procedure, a proposed parent structure (or structures) was saved as a "*.mol" file.

Upon opening the software, a dialog window prompted the selection of the *.mol file. The fragment ion information from the fragment analysis product ion's high-collision-energy scan window of the selected observed impurity mass automatically exported to the software tool along with the *.mol file. Potential structures were assigned and scored for the precursor ions in the isotopically-filtered spectrum.



Figure 4 shows an example of the report that was generated by the software for the unknown impurity [M+H] 456.2305. Another conclusion that was determined by the software data was that many of the impurities that were observed in the API quetiapine have the common fragment ions m/z 279, 253, 221, and 158.

Also, the software confirmed similar fragmentation patterns of the imported structures with mass accuracy of generally less than 2.0 mDa. From these data it was also hypothesized that the structure undergoes a structural rearrangement after the cleavage of the piperazine ring. However, this did not seem to affect the mass accuracy of many of the proposed fragmentation pathways of the assumed parent structure of the unknown impurity.

Information Study
Data collection provided high chromatographic resolution, ample sensitivity, and mass accuracy to identify many of the impurities that were in the quetiapine hemifumarate drug substance. MSE provided the simultaneous acquisition of both high- and low-collision-energy, maximizing the information that was gathered from a single injection.

This analytical workflow was followed by a deliberate data processing workflow that streamlined the fragment analysis and structural elucidation process and provided greater confidence in the end results.

The browser provided a comprehensive list of elemental compositions for the known and unknown peaks; 10 known impurities were rapidly identified with an average mass accuracy < 3.0 ppm; and [M+H] = 398 and 412 were observed to have a series of structural isomers.

When using mass fragment analysis, a minimum of 25 impurity peaks were identified as being related to quetiapine, utilizing the common fragment ions m/z 279, 253, 221, and 158. And 14 integrated impurity peaks were identified with no common fragment ions.

Finally, the chemically-intelligent software tool enabled the structures of the 10 known impurities to be confirmed. Information of the possible structural isomers for [M+H] = 398 and 412 were easily compared to various proposed structural isomers for best-fit correlation to the high collision energy data.

In some cases where the peak identification was more challenging, the data acquisition, processing and interpretation software was able to help formulate decisions about compound determination.

The combination of these three software tools, along with the optimized instrument configurations for impurity analysis and efficient MSE acquisition, provided a systematic workflow approach that can readily be applied to identify and confirm known and unknown peaks in an impurity profile.

This workflow-based approach delivered the rapid and systematic set of comprehensive results that were needed to identify and confirm impurities in an API impurity profile.