Multimodal Nonlinear Optical Imaging for Sensitive ... - ACS Publications

Subscriber access provided by UNIV OF ESSEX

Article

Multimodal Non-linear Optical Imaging for Sensitive Detection of Multiple Pharmaceutical Solid-State Forms and Surface Transformations Dunja Novakovic, Jukka Kalle Samuel Saarinen, Tatu Petteri Rojalin, Osmo Antikainen, Sara Jane Fraser-Miller, Timo Laaksonen, Leena Peltonen, Antti Isomäki, and Clare J. Strachan Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b02639 • Publication Date (Web): 26 Sep 2017 Downloaded from http://pubs.acs.org on October 4, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 13

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Multimodal Non-linear Optical Imaging for Sensitive Detection of Multiple Pharmaceutical Solid-State Forms and Surface Transformations Dunja Novakovic†,‡ Jukka Saarinen†,‡,*, Tatu Rojalin#, Osmo Antikainen†, Sara J. Fraser-Miller†,⊥, Timo Laaksonen#,∥, Leena Peltonen†, Antti Isomäki§, Clare J. Strachan† †Division of Pharmaceutical Chemistry and Technology, Viikinkaari 5E, 00014 University of Helsinki, Finland #Division of Pharmaceutical Biosciences, Viikinkaari 5E, 00014 University of Helsinki, Finland ⊥Dodd-Walls Centre for Photonic and Quantum Technologies, Department of Chemistry, University of Otago, Dunedin, New Zealand §Biomedicum Imaging Unit, Haartmaninkatu 8, 00014 University of Helsinki, Finland ∥Laboratory of Chemistry and Bioengineering, Tampere University of Technology, Korkeakoulunkatu 8, 33720 Tampere, Finland ‡Dunja Novakovic and Jukka Saarinen contributed equally to this work. *E-mail: [email protected] ABSTRACT: Two non-linear imaging modalities, coherent anti-Stokes Raman scattering (CARS) and sum-frequency generation (SFG), were successfully combined for sensitive multimodal imaging of multiple solid-state forms and their changes on drug tablet surfaces. Two imaging approaches were used and compared: (i) hyperspectral CARS combined with principal component analysis (PCA) and SFG imaging, and (ii) simultaneous narrowband CARS and SFG imaging. Three different solid-state forms of indomethacin — the crystalline gamma and alpha forms, as well as the amorphous form — were clearly distinguished using both approaches. Simultaneous narrowband CARS and SFG imaging was faster, but hyperspectral CARS and SFG imaging has the potential to be applied to a wider variety of more complex samples. These methodologies were further used to follow crystallization of indomethacin on tablet surfaces under two storage conditions: 30°C/23% RH and 30°C/75% RH. Imaging with (sub)micron resolution showed that the approach allowed detection of very early stage surface crystallization. The surfaces progressively crystallized to predominantly (but not exclusively) the gamma form at lower humidity and the alpha form at higher humidity. Overall, this study suggests that multimodal non-linear imaging is a highly sensitive, solid-state (and chemically) specific, rapid and versatile imaging technique for understanding and hence controlling (surface) solid-state forms and their complex changes in pharmaceuticals.

Solid-state selection of an active pharmaceutical ingredient (API) and associated excipients is a key component in solid dosage form (e.g. tablet) development. Selecting the solidstate form (e.g. crystalline polymorph, salt, cocrystal, or the amorphous form) with the most desirable physicochemical properties is important because it can greatly affect the critical quality attributes of the product. For example, if crystalline forms of an API are poorly water soluble, use of the disordered amorphous form of the API can substantially increase apparent solubility, dissolution rate and bioavailability, and thus allow a therapeutic effect. However, the inherent physical instability of the amorphous form means that it tends to crystallize during processing, storage or even administration to the patient. Analysis and control of solid-state transformations is thus crucial. It is increasingly required by pharmaceutical regulatory authorities, and also has commercial and intellectual property implications. One of the more challenging tasks in solid-state analysis is the detection of subtle but important transformations (e.g. crystallization) occurring on surfaces. Although the surface makes up

a tiny fraction of the total mass of most solid drug particles and dosage forms (e.g. tablets), solid-state specific surface analysis is important due to i) the different solid-state behaviour of surfaces compared to the bulk, and ii) the crucial influence of the surface solid-state form on many critical quality attributes, such as processability, stability and dissolution. Crystallization is often faster at the surface than within the interior of drug particles and dosage forms and has been attributed to higher molecular mobility at the surface (below the glass transition temperature (Tg)).1–4 For example, free surface crystallization of indomethacin films at 40°C (2°C below Tg) has been found to be two orders of magnitude faster than that of the bulk.1 Existing studies into surface crystallization of drugs have mostly involved individual drug particles or glass films.1,5,6,7 However, solid-state transformations on solid dosage form surfaces such as tablets are also important.8,9 For solid-state analysis, well-established techniques in pharmaceutics like X-ray powder diffraction (XRPD), differential scanning calorimetry (DSC), Raman and infrared (IR) spectroscopies have little or no surface sensitivity in the commonly

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

used configurations for solid-state analysis (with the exception of attenuated total reflection (ATR) IR spectroscopy). As a result, surface-specific and solid-state sensitive analysis has received comparatively little attention in pharmaceutical academia or industry. Laboratory-based and surface-sensitive solid-state imaging techniques are particularly desirable. Scanning electron microscopy (SEM) offers very high resolution but lacks chemical and solid-state specificity. Raman and FTIR imaging provide chemical and solid-state resolution and can be used to probe the surfaces of solids (more or less selectively). Non-linear optical imaging is a method based on phenomena involving multiple photons. These photons interact with the sample to create a photon with a different wavelength that is then detected.10 Examples include coherent-anti Stokes Raman scattering (CARS) and sum-frequency generation (SFG), including second-harmonic generation (SHG). These processes can be used for chemical and solid-state specific imaging (with no requirement for labels). To generate non-linear signals, there are some instrument requirements. These include high intensity pulsed lasers (picosecond or femtosecond lasers), high numerical aperture objectives (NA) and suitable detectors such as photomultiplier tubes (PMTs).11,12 The laser pulses are tightly focused and the signal is generated from the small focal point which makes these techniques inherently confocal (unlike spontaneous Raman imaging/mapping). This focal point can be scanned across the sample in horizontal planes over multiple z-levels, thus making rapid 3-dimensional imaging possible. Coherent anti-Stokes Raman scattering (CARS) imaging is a third-order non-linear method based on stimulated Raman scattering associated with molecular vibrations. CARS imaging is much faster than confocal Raman mapping based on spontaneous Raman scattering (at least 100 times faster).13,14 The CARS process is a four-wave mixing process where three photons are combined and interact in a small focal point to create a fourth photon, which is then detected.11,12,15 In addition to speed, chemical-specificity and 3D imaging capability, CARS can also have submicron axial spatial resolution.16 In biomedical applications CARS has been used extensively for label-free imaging of e.g. lipids in cells and tissues.12,17–20 In pharmaceutical applications CARS microscopy is gaining interest.10 Applications include visualization of chemical component distribution in dosage forms and drug carriers13,21,22, dissolution23 and release16, solid-state transformations during dissolution 24,25, and drug delivery into cells and tissues.26 As an example of pharmaceutical solid-state analysis, Fussell et al. used CARS microscopy to visualize theophylline anhydrate dissolution and associated solution-mediated conversion to theophylline monohydrate in a flow through cell.25 This transformation was correlated with a decreased dissolution rate, confirmed by in-line UV analysis. SFG is another non-linear optical process and its signal can be generated from samples in which inversion symmetry is broken, such as crystals with a non-centrosymmetric structure.27– 29 SFG is a second order process and thus requires two input photons that interact with the sample and each other, resulting in frequency-summing of the two photons. SHG is a special case of SFG in which the two input photons have the same energy (and wavelength). In this case, the emitted photon has exactly double the energy (and half the wavelength) of the incident photons.30 SFG (and SHG) is especially suitable for

Page 2 of 13

visualizing low levels of crystallinity (as long as the crystals are non-centrosymmetric) in an otherwise amorphous sample (amorphous materials are not SFG active) with high sensitivity.8,28,31,32 Simpson et al. have used SHG to image pharmaceutical powders in particular. For example, they visualized the crystallization of cryo-milled and melt-quenched griseofulvin with the crystallinity limit of detection estimated to be 4 ppm.7 They have also used SHG to quantify crystallinity within lyophilized Abraxane® powder28, a formulation that forms an amorphous nanosuspension of albumin-bound paclitaxel.33 Crystals were detected with SHG and subsequently identified as paclitaxel crystals using confocal (spontaneous) Raman analysis. The aim of the present study was to investigate the potential of multimodal non-linear optical imaging for spatially-resolved and solid-state specific analysis of pharmaceutical surfaces. This multimodality (in a single instrument) is beneficial, since simultaneous usage of two techniques relying on different physical phenomena can either be used as cross-validation and result in more reliable data interpretation or alternatively provide information not accessible with either technique alone. We combined two non-linear imaging techniques (CARS and SFG/SHG imaging) to image amorphous indomethacin and the crystalline gamma and alpha forms on tablet surfaces. The feasibility of both hyperspectral and narrowband multimodal imaging approaches was investigated. Further, the imaging was utilized to gain new insights into the crystallization behavior of amorphous indomethacin tablets stored at low and high humidities, and the results were compared with those from Raman and FTIR spectroscopies. MATERIALS AND METHODS Preparation of different solid states of indomethacin. The gamma form of indomethacin (Orion, Finland) was used as received. Amorphous, alpha and delta forms were prepared as described elsewhere.34,35 The gamma form has a triclinic P1 crystal structure, with 1 annotating the existence of a center of symmetry (CSD code INDMET03).36 The alpha form has a noncentrosymmetricmonoclinic P21 crystal structure (CSD code INDMET02).37 Preparation of indomethacin tablets. Tablets (300 mg, 13 mm diameter) were compressed with a hydraulic press (Specac, Kent, UK) using a weight equal to 1 ton for 30 seconds. Mixture tablets used in the first part of the study consisted of equal amounts (1:1:1 w/w) of gamma, alpha and amorphous indomethacin. Tablets used for monitoring crystallization were prepared from pure amorphous indomethacin. The bottom side of each tablet was imaged. Non-linear microscopy. A fully-integrated Leica TCS SP8 CARS microscope was used for non-linear and bright-field imaging. The setup consisted of an inverted microscope equipped with a laser-scanning confocal scan-head and photomultiplier tube (PMT) and GaAsP hybrid (HyD) photodetectors. The CARS excitation source was a Nd:YVO4 solid-statelaser integrated with an optical parametric oscillator (OPO) (APE GmbH, Germany). Typically, in CARS setups one of the wavelengths is provided by the fundamental laser with a fixed wavelength (Stokes beam (ωs)). Part of the Stokes beam is also combined with an OPO for tuning to the second required wavelength (pump


Page 3 of 13

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


beam (ωp)). The pump beam wavelength can be tuned so that the energy difference between these beams corresponds to some molecular vibration resonance. The vibration is then probed with a probe photon (ωpr), which can originate from the same beam as the pump photon. These beams are coherently driven into the sample and wave mixing results in generation of the fourth, blue-shifted, anti-Stokes photon (ωas), which is then detected. SFG signals can be simultaneously generated in noncentrosymmetric crystals using the same lasers. The SFG wavelengths can be calculated as follows: =

,

(1)

where λ1 and λ2 are the irradiation wavelengths. In this case, three SFG signals are possible: two SHG signals (one each corresponding to frequency doubling of each laser) and an SFG signal corresponding to the sum frequency of the two different lasers. In the present setup, ωs had a fixed wavelength of 1064.5 nm and the pulse duration was 7 ps. The bandwidth was about 2–3 cm-1 and the repetition rate 80 MHz. The ωp and ωpr beams were generated from the OPO with a pulse duration of 5–6 ps. The excitation light beams were linearly polarized (extinction ratio of 100:1) and co-polarized with respect to each other. CARS and SFG/SHG signals were detected in the backward (epi) direction. CARS signals were detected using nondescanned PMT detectors. SFG/SHG signals were collected using PMT detectors and HyD detectors. Bright field (BF) imaging was additionally used to visualize the particle and tablet surface morphology. Samples were illuminated with a 633 nm He/Ne laser and transmitted light (with powders) or reflected light (with tablets) was detected with a PMT. The powders or tablets were placed on glass coverslips and images (512 × 512 or 1024 × 1024 pixels) were acquired with a pixel dwell time of 1.2 µs (line average 4). Measurements were performed from the same fields of view so that the CARS, SFG and BF images could be overlaid and directly compared. A water-immersion objective (25×) with an NA of 0.95 (Leica HCX IR APO L 25× / 0.95 W) was used in every imaging modality. The Leica Application Suite Advanced Fluorescence (LASAF) was used for image acquisition and processing together with Fiji Image J (open-source distribution), GNU Image Manipulation Program v2 (open-source distribution), OriginPro 8.6 (OriginLab, Northampton, Massachusetts, USA) and MATLAB R2016a (MathWork, MA, USA). CARS and SFG/SHG spectra acquisition. CARS spectra were measured between 1413 cm-1 and 1800 cm-1 by systematically tuning the wavelength of the pump laser 33 times (893.3 nm – 925.3 nm). The second-order non-linear spectra containing the SFG and SHG signals were recorded by exciting the sample with the Stokes wavelength of 1064.5 nm and a pump wavelength of 901.3 nm while collecting the emitted photons with a HyD detector from 400 nm to 700 nm by changing the detection range in 10 nm steps. In this way a spectrum containing the three SFG/SHG signals could be obtained, with the predicted SHG signals at 450.65 nm and 532.25 nm and the SFG signal at 488.063 nm (Equation 1). The positions of the peaks and absence of a broader background signal were used to confirm that the detected signal originates from SFG/SHG

(and therefore crystalline material) rather than the less solidstate specific two-photon excited fluorescence (TPEF) (the intensities of the second-order signals were not calibrated for the detector´s spectral response). Multimodal hyperspectral CARS and SFG imaging. CARS and second-order non-linear spectra were recorded as described above. MATLAB was used to process the hyperspectral CARS data. Images (33 images with different wavenumbers, each with 512 × 512 pixels) obtained from each CARS spectral scan were converted into ascii- files so that each pixel had a value corresponding to the intensity of the pixel. The data was treated with standard normal variate (SNV) transformation. These matrices were then overlaid to achieve a 3D data stack (x,y,ω) and the data was mean centered. Thus, each pixel represented a single CARS spectrum resulting in a total of 262 144 spectra. Principal component analysis (PCA) was then performed on the data matrices. The first three principal components (PCs) were used to create RGB color images based on PC score values at each pixel. The PC score values were normalized so that the minimum PC score value was set to 0 and the maximum score value to 1 and all values in between scaled linearly. PC1, PC2 and PC3 scores are represented by red, green and blue coloring, respectively. The relative intensity of these colors at each pixel depended on the PC score values and can vary case by case. It was also possible to extract the original spectrum from each pixel. The SFG peak signal intensities from 480 to 490 nm were extracted from the second-order non-linear spectra and subsequently overlaid with the PCA-based CARS images. Multimodal narrowband (single-shift) CARS and SFG/SHG imaging. The CARS signal at a specific CARS shift unique for one of the solid-state forms (identified from the CARS spectra) and the SFG/SHG signals (with the bandpass filter 465 nm ± 85 nm) were simultaneously detected using two separate channels. Images obtained from these two channels were then overlaid. To remove any TPEF signal interference from CARS signal, the signal detected when the sample was excited with one laser only (pump beam) was subtracted from the overall signal detected when the excitation was carried out with the both Stokes and pump/probe beams. Crystallization. The crystallization behavior of amorphous indomethacin tablets was studied at 30°C and two different humidities: 23% RH and 75% RH. The tablets were placed in open glass vials in desiccators with the bottom surface exposed to the air. Desiccators with saturated salt solutions — potassium acetate and sodium chloride — were used to create 23% RH and 75% RH, respectively. Eleven tablets were prepared in total. A freshly made tablet was used as the day 0 sample. For the other time points — days 1, 2, 5, 7 and 22 — a separate tablet from each of the two desiccators was used for each time point. The stored tablets were analyzed by visual inspection, SEM, FTIR-ATR spectroscopy, Raman microscopy (method descriptions in Supporting Information), and CARS and SFG/SHG imaging. Each of the tablets were analyzed in triplicate from three different surface sites. Simca software (v.14.1, Umetrics, Umeå, Sweden) was used to perform preprocessing (SNV transformation and mean centering) and PCA on the FTIR and Raman spectra.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

RESULTS AND DISCUSSION Characterization of indomethacin solid-state forms. The FTIR and Raman spectra of the gamma, alpha and amorphous forms of indomethacin are distinct in the carbonyl (C=O) stretching region between 1600 and 1750 cm-1 (Figure 1). The indomethacin molecule contains two carbonyl groups; these are contained in the benzoyl and carboxylic acid moieties. The molecular vibrations involving these groups have previously been described and are listed in Table S1.38,39 SEM and BF microscopy revealed that gamma indomethacin exhibited prismatic morphology with a particle size of approximately 1 µm or less, whereas the alpha crystals were needle-like with crystals up to 100 µm in length. The amorphous particles were prismatic, similar to that of gamma indomethacin, and sized between those of the gamma and alpha crystals (Figure S1). Based on their CARS spectra (Figure 2a), which show similar features to the Raman spectra (Figure 1b), the following peaks were selected for narrowband CARS imaging: 1701 cm1 for gamma indomethacin and 1676 cm-1 for amorphous indomethacin. The strongest CARS signal intensity at the highest peak was produced by the gamma form, followed by the amorphous and alpha forms respectively (Figure S2). The characteristic spectral peaks due to the second-order nonlinear response are seen in Figure 2b. The solid-state specific CARS images of indomethacin powders are shown in Figures 2c-d. No observable TPEF signal was detected when samples were illuminated with only one laser wavelength. Strong SFG/SHG signals were observed for the non-centrosymmetric alpha indomethacin crystals (Figures 2e-f). No such signals were observed for the centrosymmetric gamma and amorphous forms. An image based on the combination of SHG/SFG signals specific for the non-centrosymmetric alpha form is shown in Figure 2e. The yellow color assigned to the SFG signal from the alpha crystals, as shown in Figure 2f, is used in all multimodal hyperspectral CARS and SFG images. The colors assigned to the solid-state forms are blue, red and green for amorphous, gamma and alpha indomethacin, respectively (used for all narrowband CARS and SHG/SFG images). Figure 1. FTIR (a) and Raman (b) spectra of the different solidstate forms of indomethacin. Line styles — dotted, dashed and solid for amorphous, gamma and alpha indomethacin, respectively — are used throughout the manuscript when reference to the spectra of the pure forms is made. Figure 2. CARS spectra of different polymorphs (a) and secondorder non-linear emission spectrum of alpha indomethacin (b). CARS images (c-d) of gamma and amorphous indomethacin, an image of alpha indomethacin based on the combination signal of SHG/SFG detected at (465 nm ± 85 nm) (green bar in spectrum b) (e) and SFG image of alpha indomethacin (f) detected between 480 and 490 nm from the spectral scan (yellow bar in spectrum b). Gamma indomethacin (c) was imaged using the CARS shift at 1701 cm-1 and amorphous indomethacin (d) at 1676 cm-1.

Hyperspectral imaging of different solid-state forms of indomethacin on tablet surfaces. Surfaces of indomethacin tablets (1:1:1 w/w, amorphous, alpha and gamma) were visualized using hyperspectral CARS and SFG imaging. The resulting PCA based CARS images alone and combined with the SFG signal, are shown in Figures 3 and S3. CARS spectra generated from differently colored regions are also depicted (Figures 3d-f) along with reference spectra of the pure materials.

Page 4 of 13

The different solid-state forms of indomethacin are clearly resolved (Figure 3) based on CARS microscopy. Red indicates gamma indomethacin, green is associated with alpha indomethacin, while blue-green corresponds to amorphous indomethacin. These assignments were based on the features observed in the PC loadings (Figure S4). Good correlation between SFG and CARS was observed for regions showing alpha crystals. Therefore, the two non-linear microscopy methods, relying on different mechanisms (detection of molecular vibrations (CARS) and SFG signal produced by noncentrosymmetric crystals) were synergistically combined to improve image interpretation confidence. Figure 3. PCA based CARS image (a), overlaid RGB (CARS) image and SFG image (b) of the surface of a tablet containing equal ratios of gamma, alpha and amorphous indomethacin and CARS spectra extracted from the red, green and blue regions shown as arrows (c-e). The reference spectra of the different solid-state forms of indomethacin in powdered forms are shown for comparison (c-e). The RGB image is generated from a CARS spectral scan in the region 1413 cm-1 to 1800 cm-1, using the score values of the first three PCs. The SFG image is obtained from pixels showing signal detected at 480 – 490 nm obtained from second-order non-linear spectrum (yellow, b). PCA loadings are shown in Figure S4.

Hyperspectral CARS imaging has previously been used with PCA in biomedical applications.40,41 As shown previously40,41 PCA can be used with CARS even though CARS is not a linear imaging method. However, one has to be aware of the challenges in processing of CARS spectra. Usually CARS spectral peaks are broader compared to Raman spectra and there is always a non-resonant background. Therefore, especially in the cases where substantial mixing of signals could occur in much of the sample, it would be beneficial to use some other (supervised) multivariate data-analysis method or transform CARS spectra to linear Raman-like spectra by extracting the imaginary part.42 This would, however, typically require a broadband CARS setup and controlling of the phase term. One potential benefit of PCA, in addition to it being relatively simple and requiring no a priori spectral input (i.e. it is an unsupervised method), is that it is perhaps the best known and understood multivariate spectral data analysis method in the pharmaceutical setting. The present study represents the first of its kind where PCA is combined with CARS (and SFG/SHG) imaging for pharmaceutical analysis. Comparison of hyperspectral and narrowband (single shift) CARS imaging. We compared the narrowband CARS and SFG/SHG image (Figure 4a) to the hyperspectral PCA based CARS/SFG image (Figure 4b). Colors were assigned as described in the Materials and Methods section. In a narrowband approach, red and blue were used to depict the resulting CARS signals at 1701 cm-1 (gamma form) and 1676 cm-1 (amorphous form), respectively. The CARS images were then overlaid with the SFG/SHG image from the alpha form (shown in green, Figure 4a). The corresponding overlaid hyperspectral CARS image (PCA) and SFG image is shown in Figure 4b. The different solid-state forms of indomethacin are clearly distinguishable in both the hyperspectral and narrowband CARS and SFG/SHG images. Figure 4. An overlaid narrowband CARS and SFG/SHG image of the surface of a tablet containing gamma, alpha and amorphous


Page 5 of 13

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


indomethacin in equal amounts (a), and the corresponding PCA based CARS image overlaid with the SFG image of alpha crystals detected at 480 – 490 nm (yellow) (b). Gamma indomethacin was imaged using the CARS signal at 1701 cm-1 (red), amorphous indomethacin with CARS signal at 1676 cm-1 (blue) and alpha indomethacin with the SFG/SHG signal (green) (a). The PCA based image was created from the CARS spectrum from 1413 to 1800 cm-1 (b). The associated PCA loadings are shown in the Supporting Information (Figure S5).

Both imaging approaches have inherent benefits. The narrowband imaging involving simultaneous CARS and SFG detection is faster and 3D imaging is more feasible. In contrast, the more data-rich hyperspectral approach holds potential for more complex samples with overlapping (unresolved) spectral features, especially when combined with multivariate analysis. Crystallization of amorphous indomethacin. Solid-state transformations occurring on amorphous tablet surfaces were investigated. The storage conditions were selected based on a study by Andronis et al.,43 which provided evidence that amorphous indomethacin crystalizes to the gamma form at 30 °C when the humidity is below 43% RH and to the alpha form when the humidity is higher. Their study, however, involved amorphous glass films rather than tablets. The tablets in the present study stored at 30°C/23%RH changed color from the typical yellow of amorphous indomethacin to white, suggesting surface crystallization. The tablets stored at 30°C/75%RH did not change their color so drastically, indicating a different crystallization rate or mechanism (Figure S6). SEM also indicated different solid-state behavior (Figure S7). The overall trend at 30°C/23%RH was that small prismatic particles (typical to gamma form) appeared and increased on the tablet surfaces (Figure S1b). These particles were observed within 2 days of storage, and after 22 days large regions of such crystals were observed. On the tablets stored at 30°C/75%RH, the visible surface structure changes were slower. However after 22 days, regions of the tablet surface with needle-like particles exhibiting the typical morphology of alpha crystals were observed (Figure S1c). The color and morphological changes indicated different surface crystallization behaviour for the two storage conditions, but do not give direct evidence of crystallinity or the resulting solid-state forms. To achieve this, CARS and SFG/SHG imaging was performed. Raman microscopy and FTIR-ATR spectroscopy were used for comparison. There was no evidence of immediate compression-induced crystallization on the tablet surfaces prior to storage based on any of the analysis techniques (Figures S8 and S9). The general crystallization behavior observed with CARS and SFG/SHG imaging was in general agreement with that previously reported, with crystallization to the alpha form at the higher relative humidity and crystallization to the gamma form at the lower relative humidity (Figure 5).6,43 Crystals were detected within the first one to two days of storage. The transformation was, however, not exclusive. For instance, some small SFG/SHG-active regions were detected amongst the inactive amorphous and gamma regions at 30°C/23%RH (Figure 5, spot 9). These regions were still visible on day 22. As the gamma polymorph is SFG/SHG inactive, this activity was most likely to be due to either (i) small amounts of the alpha form or (ii) small amounts of some other SFG/SHG

active form. As unambiguous alpha indomethacin CARS spectra could not be extracted from all the small green (SFG/SHG active) regions observed in tablets stored at 30°C/75%RH, the existence of other indomethacin polymorphs was also considered. The prepared delta form was also SFG/SHG active and the crystal morphology was needle-like, similar to alpha indomethacin. However, the presence of the delta form was ruled out by comparing its CARS spectrum with those extracted from the tablet surface images (data not shown). Similarly, with the tablets stored at 30°C/75%RH, several small regions generated a CARS signal at 1701 cm-1 (attributed to the gamma form) (Figure 5). Other reported indomethacin polymorphs were not analyzed by non-linear optical methods in this study. However, by comparing their reported FTIR and Raman38 spectra with our experimental data, we did not find any evidence that would indicate their presence on the tablet surfaces. Figure 5. Multimodal narrowband (single shift) CARS and SFG/SHG images of amorphous indomethacin tablets during storage at 30°C/23%RH (top) and 30°C/75%RH (bottom) (a). Images are overlaid CARS signal detected at 1701 cm-1 for gamma indomethacin (red), 1676 cm-1 for amorphous indomethacin (blue) and SFG/SHG signals for alpha indomethacin (green). CARS spectra extracted from the red, green and blue regions are indicated by arrows and numbers 1-9 (b-d). Reference spectra of the different solid-state forms of indomethacin are shown for comparison in black (b-d). The spectra are offset for clarity. Images of a freshly prepared tablet can be seen in the Supporting Information (Figure S9).

FTIR-ATR and Raman microscopy were also performed on the same tablets. Gamma indomethacin was confirmed by FTIR-ATR spectroscopy after 5 days’ storage at 30°C/23%RH. This method also confirmed alpha indomethacin at 30°C/75%RH after 22 days of storage with a peak associated with the alpha form (1650 cm-1) appearing in the IR spectra (Figure S8). Gamma indomethacin was also observed on the surfaces of tablets stored at 30°C/23%RH using Raman microscopy with the gamma peak at 1701 cm-1 visible after 5 days (Figure S8). However, alpha indomethacin was not clearly observed with Raman microscopy on the tablets stored at 30°C/75%RH. Instead, the Raman spectra, supported by their PCA scores plot, resembled that of amorphous indomethacin. This suggests that SFG/SHG is more sensitive than Raman microscopy in detecting small amounts of (alpha form) crystallinity on tablet surfaces. This is likely to be due to the different spatial (axial and lateral) resolutions (SFG/SHG having higher resolution) of the two techniques and high sensitivity of SFG/SHG imaging to non-centrosymmetric crystals7, combined with the small size (nanometer scale width), needle-like morphology, growth patterns and limited tablet surface coverage of the alpha indomethacin crystals, as also evidenced by SEM analysis. Typically, even in the case of confocal Raman microscopy, the axial resolution is several (up to tens of) micrometers in practice.44–46 In comparison, CARS microscopes can have submicron axial spatial resolution.47 In our CARS setup, the axial spatial resolution was approximately 2.5 µm. While the FTIR-ATR setup used is not an imaging setup, its sampling depth was limited to approximately 2 µm. This makes the technique more surface-specific in practice than the Raman microscopy. However, the compression of the tablet surface



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

onto the ATR crystal will have affected the surface structure of the tablets. Overall, it appears that the crystallization at 23%RH (to the gamma form) was more rapid than the crystallization at 75%RH (to the alpha form). The non-linear optical imaging detected changes earlier and more sensitively than FTIR-ATR and Raman microscopy methods. Further improvement in the SFG/SHG sensitivity and in the speed of CARS spectral imaging could be obtained by using ultrashort, broadband excitation pulses. This comes, however, at a cost of added complexity in the imaging setup. CONCLUSIONS This work represents the first simultaneous application of two complementary non-linear imaging techniques (CARS and SFG/SHG imaging) to visualize crystallinity and its changes in pharmaceutical samples. This study demonstrates the feasibility and potential benefits of such multimodal non-linear imaging method in visualizing multiple solid state forms of an API on surfaces and monitoring subtle but important solid-state changes during storage. The combination of these methodologies is especially beneficial, since SFG/SHG imaging was very sensitive to non-centrosymmetric crystals on tablet surfaces and CARS imaging, with detection based on molecular vibrations, was successfully used to distinguish solid states of API that could not be resolved using SFG/SHG imaging. With the increasing availability of (fully-integrated) non-linear optical imaging setups, these methodologies have potential for widespread characterization and optimization of solid-state forms of drugs and dosage forms in pharmaceutical industry. Since the imaging methods are also non-destructive, the same imaged samples may then be subjected to subsequent pharmaceutical testing. This would allow the relationship between surface characteristics and critical quality attributes (e.g. dissolution) to be better understood, optimized and controlled.

ASSOCIATED CONTENT Supporting Information Description of complementary methods, table summarizing Raman and FTIR peak assignments, SEM and BF images of pure indomethacin powders; CARS spectra of indomethacin powders showing relative intensities; RGB color images (PCA) of tablet surfaces; PCA loadings for Figures 3 and 4b; photographs of tablets during crystallization; SEM images of tablet surfaces during crystallization; PCA score plots, loadings and average FTIR and Raman spectra of tablet surfaces during crystallization; BF and CARS image of freshly prepared amorphous tablet. The Supporting Information is available free of charge on the ACS Publications website.

AUTHOR INFORMATION Corresponding Author *E-mail: [email protected] Author Contributions ‡Dunja Novakovic and Jukka Saarinen contributed equally to this work.

ACKNOWLEDGMENTS

Page 6 of 13

JS and CS acknowledge the University of Helsinki for a threeyear research project grant (2014-2016) and the Academy of Finland (grant no. 289398). DN acknowledges Doctoral Programme in Drug Research (DPDR) funding. TL acknowledges the Academy of Finland (grant no. 258114).

REFERENCES (1) Wu, T.; Yu, L. Pharm. Res. 2006, 23 (10), 2350–2355. (2) Yoshioka, M.; Hancock, B. C.; Zografi, G. J. Pharm. Sci. 1994, 83 (12), 1700–1705. (3) Zhu, L.; Jona, J.; Nagapudi, K.; Wu, T. Pharm. Res. 2010, 27 (8), 1558–1567. (4) Zhu, L.; Wong, L.; Yu, L. Mol. Pharm. 2008, 5 (6), 921–926. (5) Priemel, P. A.; Laitinen, R.; Barthold, S.; Grohganz, H.; Lehto, V. P.; Rades, T.; Strachan, C. J. Int. J. Pharm. 2013, 456 (2), 301–306. (6) Priemel, P. A.; Grohganz, H.; Gordon, K. C.; Rades, T.; Strachan, C. J. Eur. J. Pharm. Biopharm. 2012, 82 (1), 187–193. (7) Wanapun, D.; Kestur, U. S.; Taylor, L. S.; Simpson, G. J. Anal. Chem. 2011, 83 (12), 4745–4751. (8) Mah, P. T.; Novakovic, D.; Saarinen, J.; Van Landeghem, S.; Peltonen, L.; Laaksonen, T.; Isomäki, A.; Strachan, C. J. Pharm. Res. 2016, 34 (5), 957–970. (9) Thakral, N. K.; Mohapatra, S.; Stephenson, G. A.; Suryanarayanan, R. Mol. Pharm. 2015, 12 (1), 253–263. (10) Strachan, C. J.; Windbergs, M.; Offerhaus, H. L. Int. J. Pharm. 2011, 417 (1–2), 163–172. (11) Rodriguez, L. G.; Lockett, S. J.; Holtom, G. R. Cytometry. A 2006, 69 (8), 779–791. (12) Evans, C. L.; Xie, X. S. Annu. Rev. Anal. Chem. 2008, 1, 883–909. (13) Hartshorn, C. M.; Lee, Y. J.; Camp, C. H.; Liu, Z.; Heddleston, J.; Canfield, N.; Rhodes, T. A; Hight Walker, A. R.; Marsac, P. J.; Cicerone, M. T. Anal. Chem. 2013, 85 (17), 8102–8111. (14) Slipchenko, M. N.; Chen, H.; Ely, D. R.; Jung, Y.; Carvajal, M. T.; Cheng, J.-X. Analyst 2010, 135 (10), 2613–2619. (15) Cheng, J.; Xie, X. S. J. Phys. Chem. B 2004, 108 (3), 827–840. (16) Kang, E.; Wang, H.; Kwon, K.; Robinson, J.; Park K.; Cheng J. Anal. Chem. 2006, 78 (23), 8036–8043. (17) Wong, C. S. Y.; Robinson, I.; Ochsenkühn, M. A.; Arlt, J.; Hossack, W. J.; Crain, J. Biomed. Opt. Express 2011, 2 (1), 2504–2516. (18) Saarinen, J.; Sözeri, E.; Fraser-Miller, S. J.; Peltonen, L.; Santos, H. A.; Isomäki, A.; Strachan, C. J. Int. J. Pharm. 2017, 523 (1), 270–280. (19) Nan, X.; Potma, E. O.; Xie, X. S. Biophys. J. 2006, 91 (2), 728–735. (20) Pirhonen, J.; Arola, J.; Sädevirta, S.; Luukkonen, P.; Karppinen, S. M.; Pihlajaniemi, T.; Isomäki, A.; Hukkanen, M.; Yki-Järvinen, H.; Ikonen, E. PLoS One 2016, 11 (1), 1–14. (21) Fussell, A. L.; Grasmeijer, F.; Frijlink, H. W.; de Boer, A. H.; Offerhaus, H. L. J. Raman Spectrosc. 2014, 45 (7), 495–500. (22) Fussell, A. L.; Mah, P. T.; Offerhaus, H.; Niemi, S.-M.; Salonen, J.; Santos, H. A; Strachan, C. Acta Biomater. 2014, 10 (11), 4870–4877. (23) Windbergs, M.; Jurna, M.; Offerhaus, H. L.; Herek, J. L.; Kleinebudde, P.; Strachan, C. J. Anal. Chem. 2009, 81 (6), 2085–2091. (24) Jurna, M.; Windbergs, M.; Strachan, C. J.; Hartsuiker, L.; Otto, C.; Kleinebudde, P.; Herek, J. L.; Offerhaus, H. L. J. Innov. Opt. Health Sci. 2009, 2 (1), 37–43. (25) Fussell, A.; Garbacik, E.; Offerhaus, H.; Kleinebudde, P.; Strachan, C. Eur. J. Pharm. Biopharm. 2013, 85 (3), 1141–1147. (26) Darville, N.; Saarinen, J.; Isomäki, A.; Khriachtchev, L.; Cleeren, D.; Sterkens, P.; van Heerden, M.; Annaert, P.; Peltonen, L.; Santos, H. A; Strachan, C. J.; Van den Mooter, G. Eur. J. Pharm. Biopharm. 2015, 96, 338–348. (27) Campagnola, P. J.; Loew, L. M. Nat. Biotechnol. 2003, 21 (11), 1356–1360. (28) Schmitt, P. D.; Trasi, N. S.; Taylor, L. S.; Simpson, G. J. Mol. Pharm. 2015, 12 (7), 2378–2383. (29) Franken, P. A.; Hill, A. E.; Peters, C. W.; Weinreich, G.. Phys. Rev. Lett. 1961, 7 (4), 118-120. (30) Mouras, R.; Rischitor, G.; Downes, A.; Salter, D.; Elfick, A. J. Raman Spectrosc. 2010, 41 (8), 848–852. (31) Strachan, C. J.; Lee, C. J.; Rades, T. J. Pharm. Sci. 2004, 93 (3), 733– 742. (32) Kestur, U. S.; Wanapun, D.; Toth, S. J.; Wegiel, L. A; Simpson, G. J.; Taylor, L. S. J. Pharm. Sci. 2012, 101 (11), 4201–4213.


Page 7 of 13

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(33) Miele, E.; Spinelli, G. P.; Miele, E.; Tomao, F.; Tomao, S. Int. J. Nanomedicine 2009, 4, 99–105. (34) Crowley, K. J.; Zografi, G. J. Pharm. Sci. 2002, 91 (2), 492–507. (35) Kaneniwa, N.; Otsuka, M.; Hayashi, T. Chem. Pharm. Bull. 1985, 33 (8), 3447–3455. (36) Cox, P. J.; Manson, P. L. Acta Crystallogr. 2003, E59 (7), 986–988. (37) Chen, X.; Morris, K. R.; Griesser, U. J.; Byrn, S. R.; Stowell, J. G. J. Am. Chem. Soc. 2002, 124 (50), 15012–15019. (38) Surwase, S. A.; Boetker, J. P.; Saville, D.; Boyd, B. J.; Gordon, K. C.; Peltonen, L.; Strachan, C. J. Mol. Pharm. 2013, 10 (12), 4472–4480. (39) Taylor, L. S.; Zografi, G. Pharmaceutical Research. 1997, 14 (12), 1691–1 (40) Lin, C.-Y.; Suhalim, J. L.; Nien, C. L.; Miljković, M. D.; Diem, M.; Jester, J. V; Potma, E. O. J. Biomed. Opt. 2011, 16 (2), 21104. (41) Lim, R. S.; Suhalim, J. L.; Miyazaki-Anzai, S.; Miyazaki, M.; Levi,

M.; Potma, E. O.; Tromberg, B. J. J. Lipid Res. 2011, 52 (12), 2177–2186. (42) Pohling, C.; Buckup, T.; Motzkus, M. J Biomed Opt 2011, 16 (2), 21105. (43) Andronis, V.; Yoshioka, M.; Zografi, G. J. Pharm. Sci. 1997, 86 (3), 346–351. (44) Everall, N.; Lapham, J.; Adar, F.; Whitley, A.; Lee, E.; Mamedov, S. Appl. Spectrosc. 2007, 61 (3), 251–259. (45) Everall, N. Spectroscopy 2004, 19 (10), 22–28. (46) Haaser, M.; Windbergs, M.; McGoverin, C. M.; Kleinebudde, P.; Rades, T.; Gordon, K. C.; Strachan, C. J. J. Pharm. Sci. 2011, 100 (10), 4452–4459. (47) Wang, H.; Fu, Y.; Zickmund, P.; Shi, R.; Cheng, J.-X. Biophys. J. 2005, 89 (1), 581–591.

For TOC only



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 1. FTIR (a) and Raman (b) spectra of the different solid-state forms of indomethacin. Line styles — dotted, dashed and solid for amorphous, gamma and alpha indomethacin, respectively — are used throughout the manuscript when reference to the spectra of the pure forms is made. 82x119mm (300 x 300 DPI)


Page 8 of 13

Page 9 of 13

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 2. CARS spectra of different polymorphs (a) and second-order non-linear emission spectrum of alpha indomethacin (b). CARS images (c-d) of gamma and amorphous indomethacin, an image of alpha indomethacin based on the combination signal of SHG/SFG detected at (465 nm ± 85 nm) (green bar in spectrum b) (e) and SFG image of alpha indomethacin (f) detected between 480 and 490 nm from the spectral scan (yellow bar in spectrum b). Gamma indomethacin (c) was imaged using the CARS shift at 1701 cm-1 and amorphous indomethacin (d) at 1676 cm-1. 84x120mm (300 x 300 DPI)



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 3. PCA based CARS image (a), overlaid RGB (CARS) image and SFG image (b) of the surface of a tablet containing equal ratios of gamma, alpha and amorphous indomethacin and CARS spectra extracted from the red, green and blue regions shown as arrows (c-e). The reference spectra of the different solidstate forms of indomethacin in powdered forms are shown for comparison (c-e). The RGB image is generated from a CARS spectral scan in the region 1413 cm-1 to 1800 cm-1, using the score values of the first three PCs. The SFG image is obtained from pixels showing signal detected at 480 – 490 nm obtained from second-order non-linear spectrum (yellow, b). PCA loadings are shown in Figure S4. 177x131mm (300 x 300 DPI)


Page 10 of 13

Page 11 of 13

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 4. An overlaid narrowband CARS and SFG/SHG image of the surface of a tablet containing gamma, alpha and amorphous indomethacin in equal amounts (a), and the corresponding PCA based CARS image overlaid with the SFG image of alpha crystals detected at 480 – 490 nm (yellow) (b). Gamma indomethacin was imaged using the CARS signal at 1701 cm-1 (red), amorphous indomethacin with CARS signal at 1676 cm-1 (blue) and alpha indomethacin with the SFG/SHG signal (green) (a). The PCA based image was created from the CARS spectrum from 1413 to 1800 cm-1(b). The associated PCA loadings are shown in the Supporting Information (Figure S5). 83x41mm (300 x 300 DPI)



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 5. Multimodal narrowband (single shift) CARS and SFG/SHG images of amorphous indomethacin tablets during storage at 30°C/23%RH (top) and 30°C/75%RH (bottom) (a). Images are overlaid CARS signal detected at 1701 cm-1 for gamma indomethacin (red), 1676 cm-1 for amorphous indomethacin (blue) and SFG/SHG signals for alpha indomethacin (green). CARS spectra extracted from the red, green and blue regions are indicated by arrows and numbers 1-9 (b-d). Reference spectra of the different solid-state forms of indomethacin are shown for comparison in black (b-d). The spectra are offset for clarity. Images of a freshly prepared tablet can be seen in the Supporting Information (Figure S9). 177x123mm (300 x 300 DPI)


Page 12 of 13

Page 13 of 13

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


For TOC only 84x47mm (300 x 300 DPI)


Multimodal Nonlinear Optical Imaging for Sensitive ... - ACS Publications

Recommend Documents