Early pediatric formulation development with new chemical entities: Opportunities of e-tongue besides human taste assessment

Accepted Manuscript Title: Early pediatric formulation development with new chemical entities: opportunities of e-tongue besides human taste assessment Authors: Laura Isabell Immohr, Angela Dischinger, Peter Kuhl, ¨ Heidemarie Kletzl, Stefan Sturm, Andreas Guenther, Miriam Pein-Hackelbusch PII: DOI: Reference:

S0378-5173(17)30677-4 http://dx.doi.org/doi:10.1016/j.ijpharm.2017.07.069 IJP 16884

To appear in:

International Journal of Pharmaceutics

Received date: Revised date: Accepted date:

28-3-2017 19-7-2017 21-7-2017

Please cite this article as: Immohr, Laura Isabell, Dischinger, Angela, Kuhl, ¨ Peter, Kletzl, Heidemarie, Sturm, Stefan, Guenther, Andreas, Pein-Hackelbusch, Miriam, Early pediatric formulation development with new chemical entities: opportunities of e-tongue besides human taste assessment.International Journal of Pharmaceutics http://dx.doi.org/10.1016/j.ijpharm.2017.07.069 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Early pediatric formulation development with new chemical entities: opportunities of e-tongue besides human taste assessment

Laura Isabell Immohr1, Angela Dischinger2, Peter Kühl2, Heidemarie Kletzl3, Stefan Sturm3, Andreas Guenther3, Miriam Pein-Hackelbusch4* 1 Institute of Pharmaceutics and Biopharmaceutics, Heinrich-Heine-University Duesseldorf, Universitaetsstr. 1, 40225 Duesseldorf, Germany 2 Formulation Research and Development, F. Hoffmann-La Roche Ltd, Basel, Switzerland 3 Clinical Pharmacology, F. Hoffmann-La Roche Ltd, Basel, Switzerland 4 Life Science Technologies, University of Applied Sciences Ostwestfalen-Lippe, Georg-WeerthStrasse 20, 32756 Detmold, Germany

*

Corresponding author

Graphical Abstract

Abstract The palatability of a pediatric drug formulation is one of the key prerequisites for therapeutic success. Liquid formulations are often chosen for pediatric drug products, and they require special attention regarding their taste, as they have direct contact to the taste buds and a relatively long residence time in the oral cavity. For ethical reasons, the role of electronic tongues in the development of oral drug formulations with new chemical entities (NCEs) for pediatric use is growing, however, little is known about the strategies how this instrumental taste assessment can be performed. The present study illustrates two possibilities to combine in-vitro and in-vivo data for the characterization of the palatability of the new drug candidates CSE3104 and CSE3165. As a first step, the implementation and suitability of electronic tongue measurements has been demonstrated by comparison of in-vivo and in-vitro data. In alignment with the taste assessment results during a singlecenter, double-blinded, randomized, placebo-controlled, single ascending dose (SAD) study in healthy subjects, the bitter taste perception of CSE3104 was assessed with e-tongue measurements. Moreover, the sensor response pattern showed comparable results of the e-tongue measurements to the human taste study of CSE3165: With increasing concentration, the bitterness values were increased. In addition, the human taste pattern showed increasing values for sourness due to higher volumes of the citric acid buffer. Results of the hedonic descriptor “unpleasant” within the human taste assessments could be related to bitterness in the instrumental taste assessment. For the second step in electronic tongue guided formulation development two possibilities are depicted in the article focusing on the effect of different excipients on the formulation on the one hand and on the assessment and comparison of two drug formulations on the other hand. Based on these results, the low number of healthy volunteers for the taste assessment in a Phase 1 study led to a meaningful interpretation, by applying in addition the electronic tongue. Using this instrumental approach led to reproducible data versus the human taste assessment, without ethical concerns, and with a reduction in time and costs. Keywords: NCE; market formulation; human taste assessment; Phase I dose escalation study

2

1. Introduction In the pharmaceutical development of new chemical entities an early planning for a pediatric development is necessary. A pediatric investigation plan (PIP) has to be submitted to the EMA after the Phase I results have been issued according to Regulation (EC) No 1901/2006 and Regulation (EC) No 1902/2006 [1, 2]. A strategy for selecting a suitable pediatric formulation needs to be set up at a very early time point, at which only predictive dose information are available and the toxicological assessment is limited. Ideally, the characterization of the new compound in support of the pediatric formulation strategy can be even covered already in the preformulation phase. The selection of an age appropriate formulation depends strongly on the targeted age groups in children. Beside the palatability and swallowability of the formulation the dosing flexibility has high importance since in the early stages of clinical development the predicted final therapeutic doses underlie high uncertainties. Highest possible dosing flexibility can be provided with liquid dosage forms [3]. On the other hand, liquid forms involve an increased risk of lacking acceptance potentially caused by taste issues that cannot be fully overcome by use of flavors and sweeteners [4, 5]. Increasingly, highly active compounds are developed nowadays and with the lack of knowledge about the toxicological profile in the early stages of drug development, preclinical tools for the assessment of the taste have become more relevant during pharmaceutical development [6]. The electronic tongue represents a tool which allows an early taste assessment in Phase 0, prior to elaborating the pediatric formulation strategy [7]. Once the Phase I trials in healthy volunteers start, the opportunity is provided to integrate taste assessment questionnaires into the studies. The obtained early taste information can be used to confirm or adjust the formulation strategy [8]. During the safety evaluation of a new compound in Phase 1, human studies with the solely purpose of a taste evaluation by a (trained) panel are not justified, particularly for highly potent compounds. Different oral liquid formulations can be tested in-vitro by using an electronic tongue with respect to their comparability and similarity in taste, particularly if for one of the formulations the taste in humans has been already assessed. However, for a new compound, for which the concentration dependent taste perception in humans is not known, the sensor responses of the electronic tongue can only give a certain idea of the taste pattern when analyzed with sensors of global selectivity [9, 10]. The electronic tongue provides the opportunity to obtain early signs for taste issues with clinical formulations [7, 11-13]. For pharmaceutical small molecules the most frequent taste issues relate to bitterness. In this paper, assessments for two new drug candidates with the electronic tongue are presented using simple oral solutions, which are typically used in Phase I dose escalation studies at different doses and drug concentrations. The drug candidates, CSE3104 and CSE3165, have been selected for the clinical development of the treatment of a pediatric indication, for which drug solutions were required as age appropriate formulation providing high dosing flexibility. Based on obtained results for the two drug candidates from a human taste assessment within Phase I trials, the in-vitro data from the electronic 3

tongue are evaluated in comparison to the outcomes of the taste questionnaires. The retrospective review of the electronic tongue results was performed in the light of in-vivo data in order to discuss the possibilities to predict in future the taste of new entities and drug formulations with electronic tongue studies and take them as guide for formulation development strategies. 2. Materials and Methods 2.1 Materials Samples provided by the company F. Hoffmann-La Roche have been prepared based on the following excipients: the inactive pharmaceutical ingredient (API) CSE3104 (free base, Roche), lactose anhydrous (Supertab 21AN, DFE Pharma, Germany), mannitol (Pearlitol 160C, maltodextrin (Glucidex 17), maltodextrin (Kleptose Linecaps 17) and betadextrin (Kleptose) from Roquette Freres (France), hydroxypropylcellulose LF (Ashland, US), tribasic sodium citrate dihydrate (Jungbunzlauer, Austria), citric acid anhydrous (Merck KGaA, Germany). The standard solution for the e-tongue measurements was prepared based on potassium chloride (Grüssing, Filsum, Germany) and tartaric acid (AppliChem, Darmstadt, Germany), the external standard solution was prepared using quinine hydrochloride dihydrate (Buchler GmbH, Germany) and quinine sulfate () was used as model substance; the washing solutions based on ethanol (100 %, VWR international, Darmstadt, Germany) and either potassium hydroxide (0.1 M, Grüssing, Filsum) and potassium chloride or aqueous hydrochloric acid (1 M, Merck, Darmstadt, Germany). API CSE3165 (free base, Roche), tribasic sodium citrate dihydrate and citric acid anhydrous (as used for CSE3104), sucralose (Merck KGaA, Germany), sodium benzoate from Emerald Kalama Chemical BV (NL), mannitol (Parteck M100, Roquettes Freres, France), ascorbic acid (DSM Nutritional Products, UK), disodium edetate dihydrate (Titriplex III, Merck Farma, Spain), polyethylenglycol 6000 (Polygykol 6000, Clariant, Germany), tartaric acid (UD Chemie, Germany), strawberry flavor (Givaudan, Switzerland). As summarized in Table 1, properties of the NCEs CSE3104 and CSE3165 are similar to those of quinine.

CSE3165 has a higher potency and requires for this reason significantly lower drug concentrations in solution. 2.2 Electronic taste sensing system For electronic tongue measurements, the taste sensing system Insent SA402B (Intelligent Sensor Technology, Inc., Atsugi-chi, Japan) was used. The electronic tongue sensors were purchased from TecLabS (Essen, Germany). They fully met the requirements of the routinely performed sensor checks done by the electronic tongue [16, 17]. 2.3 Measurement procedure of the Insent taste sensing system 4

The taste sensing systems SA402B is a sensor array system equipped with an Ag/AgCl-reference electrode and a couple of different sensor types. The following sensors were used for the evaluation: SB2AAE: umami taste, SB2CT0: saltiness, SB2CA0: sourness, SB2AE1:

astringency,

SB2AC0:

bitterness (cationic substances), SB2AN0: bitterness (cationic substances), SB2BT0: bitterness (cationic substances), SB2C00: bitterness (anionic substances). Each sample is measured with the electronic tongue four times in a row. The first measurement is excluded from further calculations to avoid any unstable data. Per measurement run 10 samples can be measured, since the Insent taste sensing system provides an autosampler with a sample tray for 10 samples. If more samples need to be evaluated, it is useful to use an external standard, to minimize drift effects. In this study, a solution of quinine hydrochloride (0.5 mM) was used as external standard and the corresponding sensor responses were subtracted from the sensor responses of the samples. Hence, the presented values are the arithmetic mean of three subsequent measurements and represent relative sensor responses. The sensor signals are given as a voltage in mV. Using the recommended measurement setup ABCABC (A, B, and C are representatives of sample beakers) the measuring routine is the same for every measurement: Dipping for 90 s in cleaning solution 1, then 120 s in cleaning solution 2, and subsequently 120 s in cleaning solution 3. The cleaning solutions have been prepared according to the supplier`s recommendations for the sensor types. A sensor check is performed in conditioning solution for 30 s prior to sample measurement (30 s). After 3 s in cleaning solution 4 and 3 s in cleaning solution 5, the CPA (change in membrane potential due to adsorption) values for the aftertaste are measured in the CPA solution for 30 s [17]. 2.4 Sample preparation for electronic tongue measurements 2.4.1 CSE3104 A calibration series of CSE3104 was performed by diluting 1000 mg of API in 100.0 mL citric buffer pH 4.4 (50 mM, tribasic sodium citrate dihydrate: 0.739 g, citric acid anhydrous: 0.478 g per 100.0 mL), which served as stock solution. Concentrations of 0.1, 0.5, 1 and 2.5 and 5.0 mg/mL were prepared by diluting the stock solution with the according citric buffer. Drug solutions containing various combinations of excipients DF_01-DF_06 (Table 2) and the according placebo solutions P_01-P_06 (Table 2) were diluted with 100.0 mL of citric buffer pH 4.4 each. The placebo samples resulted in (viscous) solutions. DF_01-DF_04 and DF_06 were yellow clear solutions after preparation, DF_05 resulted in a slight suspension.

2.4.2 CSE3165 The calibration samples in support of the assessment of the API taste pattern were prepared by diluting the highest concentrated drug solution of 1.0 mg/mL CSE3165 in 50 mM citric buffer pH 3.0 to 0.5 and 5

0.1 mg/ml API. The solutions contained further 2.0 mM ascorbic acid and 0.5 mM disodium edetate. The oral solutions which were used in the SAD study had a comparable composition. The citric buffer used in the clinic had a lower concentration of 30 mM instead of 50 mM and the pH was slightly lower at 2.7 to 2.9. Besides, a calibration was also performed in aqueous tartaric acid (see Table 3 for more details). A second type of calibration series was performed for the in-vitro evaluation of CSE3165 oral solutions by using a formulated vehicle. The vehicle formulations and calibration solution according to Table 3 were used in in-vitro assessment with the e-tongue. 2.5 Evaluation of Results All figures were created based on the raw data of the sensor signals using Microsoft Excel 2013 or SIMCA-P 12.01 (Umetrics AB, Umea, Sweden). 2.6 in-vivo evaluation During the safety evaluation of a new compound in Phase 1, human studies with the solely purpose of a taste evaluation by a (trained) panel are not justified, particularly for highly potent compounds. The assessment of the taste was therefore in this study integrated into the first human dose escalation study in healthy volunteers with the primary purpose of pharmacokinetic and toxicological evaluation. A taste questionnaire with a scoring of 1 to 5 was used for this purpose. The study design for compound CSE3104 and CSE3165 as outlined in Table 4 allowed for the assessment of the taste on each level of the ascending single doses, which were administered as drug solutions. The 4 and 6 dose levels, respectively, were dosed in subsequently scheduled cohorts, which consist of different subjects, who received verum or placebo solutions for a single dose only. The healthy volunteers were asked to evaluate the taste 1 minute after intake of the drug solution, as a worst case scenario to include e.g. lingering taste perception. No additional cohorts were setup for this in-vivo assessment, instead the already planned cohorts for pharmacokinetic and safety evaluation were used. The volunteers were not trained on taste assessment.

Study medication was for CSE3104 a solution in 50 mM citric buffer solution pH 4.5 and for CSE3165 a solution in 30 mM citric buffer pH 2.7 with ascorbic acid and disodium edetate. The following doses of CSE3104 were given to the volunteers: 0.5 mg (1 mL of 0.5 mg/mL solution), 4 mg (8 mL of a 0.5 mg/mL solution), 14 mg (2.8 mL of a 5 mg/mL solution), 40 mg (8 mL of a 5 mg/mL solution), and 80 mg (16 mL of 5 mg/mL). For CSE3165 the dose escalation was as follows: 0.6 mg (6 mL of 0.1 mg/mL), 2 mg (20 mL of 0.1 mg/mL), 6 mg (60 mL of 0.1 mg/mL), and 18 mg (18 mL of 1.0 mg/mL). The study subjects were asked to assess the taste of the assigned solution of study medication and to rank their taste perception according to the following: overall taste intensity (1 = no taste, 5 = very intense taste). If the subject´s answer was 2 or higher the questionnaire was continued with the scoring 6

for taste characteristics: unpleasant taste (1 = very pleasant, 5 = very unpleasant), bitterness (1 = no bitter taste, 5 = very bitter), metallic (1 = no metallic taste, 5 = very metallic), sourness (1 = no sour taste, 5 = very sour), saltiness (1 = no salty taste, 5 = very salty), sweetness (1 = no sweet taste, 5 = very sweet). Taste intensity and unpleasant taste are clearly hedonic ratings of the overall taste perception, which can be only received by human taste panel data and not by in-vitro assessment via e-tongue. However, the ratings of taste intensity and pleasantness can be used to crosscheck against the individual ratings for different taste characteristics, since the study subjects are not trained in taste assessments. 3. Results and discussion The two APIs CSE3104 and CSE3165 were aimed to be formulated as oral solution formulations for pediatric use. As taste is known to be a significant determinant of therapeutic success [7], the taste patterns of the developed drug substances were of particular interest in order to develop liquid drug formulations with a good acceptability. The taste patterns of both APIs were therefore evaluated firstly in a Phase 0 evaluation by e-tongue measurements and afterwards within a Phase I SAD study by healthy volunteers in an early formulation development stage. Since the number of subjects is very limited in Phase 1 studies, subjects are not trained to assess taste and the study designed is not optimal for taste panels, the obtained results are subjected to high variability and limited statistical power. In order to facilitate a reliable interpretation of the data, the use of electronic tongue measurements were considered as preferred approach in conjunction with first human data to define a pediatric formulation strategy. Moreover, prior to Phase 1, results from e-tongue measurements could indicate risks of taste issues. However, the suitability of electronic tongue sensors to assess the taste pattern of the particular API with high correlation to human taste impression is not proven at the time point at which an electronic tongue guided formulation development could provide strong benefit. Beside the assessment of drug concentration dependent taste patterns, the formulation development can also assess the impact of various excipients in the stage of preformulation development. This approach was taken for CSE3104 exemplarily to support the selection of an appropriate formulation. The second alternative development approach was taken with CSE3165 for which prototype formulations of oral solutions were selected for the assessment instead of solutions with different excipient combinations at standardized concentration levels. 3.1. CSE3104 3.1.1. Evaluation of the taste pattern of CSE3104 In order to assess the criticality of taste issues in preclinical phase prior to a human taste assessment, electronic tongue measurements were used as an instrumental analytical tool to guide a formulation development of the API CSE3104. The taste of CSE3104 was determined analyzing increasing concentrations of the API dissolved in citric buffer (50 mM, pH 4.5). In this manner, the concentration dependency of the applied sensors could be determined as an indicator for the sensitivity towards the 7

API. To perform measurements below the saturation concentration, API concentration had to be 5 mg/mL at maximum. Sensor SB2BT0 showed the most pronounced concentration dependent sensor signals for the pure API (Figure 1). The other two bitter sensors (SB2AC0 and SB2AN0) showed minor concentration dependency in the concentration range of 0.1-1 mg/mL. In contrast, sensors SB2AAE, SB2CA0 and SB2AE1 (Figure 1) showed slightly increasing signals in the concentration range of 1-5 mg/mL. CSE3104 is a basic drug and for reasons of solubility it was administered in a citric buffer solution. However, most sensor signals of the API containing solutions are very similar compared to the sensor signals of the pure citric buffer.

The taste pattern of CSE3104 was afterwards evaluated within the planned Phase I SAD study cohorts using oral solutions in a citric buffer (50 mM, pH 4.5) with two different drug concentrations, i.e. 0.5 and 5.0 mg/mL. In order to limit the dose volume, the concentration of the solution was changed from 0.5 mg/mL to 5.0 mg/mL between the dose of 4 mg and 14 mg. The study subjects (n = 6 per dose, male, aged 23-45 years) were asked to taste the assigned solution of study medication and to rank their taste perception as described in paragraph 2.6. The corresponding results are summarized in Table 5. With increasing concentration of API and dose volume an increase in the perception of bitterness, unpleasant taste, and in the taste intensity was denoted on average, particularly between 4 mg and 14 mg, where the increase in the drug concentration in the applied solution occurred. The perception of sourness was not clearly linked to API concentration, volume or rating of unpleasantness. The increase in dose between 14 and 80 mg administered with the high concentration of API was not correlated anymore with an increase in unpleasant taste, bitterness and taste intensity, but clearly stayed at a high level. At the high drug concentration of 5.0 mg/mL the taste perception was not dependent on the dose volume. The overall scattering in the taste rating was high (due to inter-individual differences = each participating healthy volunteer received different verum or placebo solutions for a single dose only, see section 2.6 for further details) and therefore confirmed the need for additional quantifiable data, e.g. those of e-tongue measurements. No significant change in saltiness, sourness, sweetness or metallic taste was reported. Most pronounced, an unpleasant perception was perceived by the healthy volunteers. A verification of the electronic tongue results concerning the taste pattern of CSE3104 was performed by comparison of the mean values of the results from the human taste assessment in the SAD study (Figure 2a) with the e-tongue results (Figure 2b). The concentrations tested with the e-tongue include also the concentration of 0.5 and 5.0 mg/mL as used in the human study. Similar results were observed, in particular regarding the attributes “unpleasant taste”, “bitterness” and “taste intensity” from the invivo study compared to sensor signals of SB2BT0 and SB2AC0, which are dedicated to cationic bitterness. Even though both sensors are dedicated to cationic bitterness, the sensor responses differed. This is due to the different composition of the sensor membranes in accordance to the variety of bitter 8

receptors on the human tongue [18]. Therefore differences in the interactions between the sample molecules and the sensor are detected. Signals and panel scores show increasing distances from the placebo solution with increasing concentration of CSE3104.

3.1.2. E-tongue guided formulation development for CSE3104 3.1.2.1. Formulation development Due to the potential risk of unpleasant taste for an oral solution formulation of CSE3104, the addition of excipients which may have a positive effect on taste was seen as beneficial for the initial drug formulation development. Thus, taste improving strategy by assessing different excipients added to the citric buffer went along with the formulation of a stable solution. To reach both aims, mannitol, anhydrous lactose, maltodextrin Glucidex 17, maltodextrin Linecaps 17, betadextrin (Kleptose) and hydroxypropylcellulose LF were added in different ratios and amounts (Table 2). Mannitol and anhydrous lactose are common soluble pharmaceutical filler excipients with a slightly sweet taste [19]. Maltodextrin and betadextrin have recently been used to successfully mask the taste of cetirizine hydrochloride, loperamide hydrochloride and dextromethorphan hydrobromide [20]. The addition of hydroxypropylcellulose LF increased the viscosity, which could also lead to a physical taste barrier [19, 21]. However, the aforementioned excipients are not particularly known as best tastemasking agents – but allowed for a stable, possibly taste improved formulation. 3.1.2.2. Multivariate evaluation of the e-tongue sensor responses The sensor responses of the whole sensor array resulted in a specific pattern for each analyzed sample based on the information of all applied sensors. This sensor response pattern can be interpreted as the according taste pattern, which can be displayed by multivariate data analysis. A principal component analysis (PCA) provides the possibility to reduce the multidimensional data in a two dimensional plot. This plot is called Score Scatter Plot and represents the samples as scores in a new coordinative system. The according graph is based on the information of all applied sensors and was used to compare the taste patterns of the different samples. The axes of the new coordinative system are called principal components (PC). Data points distributed along the x-axis (first and main component) represent the majority of the total sensor information, data points distributed along the y-axis the second less information. The according Score Scatter Plot is displayed in Figure 3, where 86.8 % of the information is represented along the x-axis and 7.3 % along the y-axis.

9

The locations of the API solutions of different concentrations within the Score Scatter Plot map (Figure 3, orange color) concluded that information along the x-axis was mainly based on the concentration of detected CSE3104. This assumption is based on the location of the calibration samples along the first principal component from right to the left and proved the sensitivity of the sensors towards CSE3104. The calibration samples were used to compare the drug formulations and their corresponding placebos functioning as a reference. The sensor response pattern of the pure drug solutions changed by addition of the different excipients. With decreasing interaction of the drug substance with the sensor membranes, either due to decreasing concentrations or decreasing contact as a result of taste masking, the samples were arranged more to the left quadrants of the Score Scatter Plot. This explains why the placebo formulations (P_01 – P_06) and the buffer solution were found in the left quadrants of the Score Scatter Plot and the drug containing formulations (DF_01 – DF_06) on the right. The Loading Scatter Plot (Figure 4) displays moreover that samples detected with increasing signals of bitter sensors SB2AN0, SB2AC0 and SB2BT0 were located more to the right. The bitter sensor SB2BT0 showed the most distinctive results and samples located in the right quadrants were supposed to be detected with an increasing bitterness according to the Loading Scatter Plot (Figure 4). Differences along the y-axis could mainly be explained by two different runs in which either samples 01 and 06 (run 1) or 02-05 (run 2) and the calibration samples (run 2) were analyzed. This was necessary due to the implementation of the Insent e-tongue, which is only able to analyze 10 solutions (nine samples and one external standard, see section 2.3) at once. Therefore, the measurements were conducted in two subsequent and randomized runs resulting in differences due to environmental influences on the measurements. As seen in the Loading Scatter Plot (Figure 4), the component p[2] in the y-axis of the Scores Scatter Plot was mainly directed by sensor SB2AE1 . Regarding the taste-masking efficacy of the added excipients only minor differences were seen. If the two different runs were evaluated individually, mannitol and anhydrous lactose provided fewest tastemasking, resulting in data points located the furthest to the right and thus, comparable to the highest concentration of CSE3104. However, neither differences between those formulation containing hydroxypropylcellulose LF (DF_06), nor between those containing maltodextrin Glucidex 17 (DF_03), maltodextrin linecaps 17 (DF_04) and betadextrin Kleptose (DF_05) (besides CSE3104 and mannitol) were detected. The excipients, tested for a potential formulation strategy, did not show an explicit change in the taste pattern of the NCE CSE3104 in this stage of clinical trials. However, the results indicated a certain repression in taste of this drug, assuming further improvement in taste perception, e.g. by use of flavors and sweeteners. 3.2 CSE3165 3.2.1 Evaluation of the taste pattern of CSE3165 Similar to what is described in section 3.1.1, the taste pattern of CSE3165 was evaluated by electronic tongue measurements prior to taste assessment within a Phase I SAD study by healthy volunteers. The samples for electronic tongue measurements were prepared in 50 mM citric buffer pH 3.0 to be in 10

alignment with the aqueous medium used in the human study. Sensors SB2AAE, SB2CT0, SB2CA0, SB2C00, SB2AE1, SB2AC0 and SB2AN0 were able to detect CSE3165 but with low sensitivity (Figure 5). A sufficient concentration dependency was only observed for sensor SB2BT0 (dedicated to cationic bitterness) with regard to sensitivity [22].

Supply material for the SAD study was prepared in 30 mM citric buffer pH 2.7 to 2.9 containing in addition 2 mM ascorbic acid and 0.5 mM disodium edetate. The solutions used for the in-vitro evaluation with e-tongue were similarly prepared in citric buffer 30 mM at pH 3.0. The dose was increased from 0.6 mg to 18 mg and the results are summarized in Table 6. For the dose of 18 mg a higher drug concentration of 1.0 mg/mL was used in contrast to 0.1 mg/mL for lower doses. With increasing dose volumes, the bitterness was increased between 0.6 mg and 2 mg dose whereas the taste intensity was not affected significantly across all doses. The perception for bitterness appeared to be mainly impacted by the volume rather than by concentration of the administered CSE3165 doses. Interestingly, taste intensity and unpleasantness were rated similar for placebo and highest concentration applied. In contrast to CSE3104, no clear link between unpleasantness and the different applied concentrations and volumes could be drawn. The difference in the mean score values between the 2 drug concentration at comparable dose volumes, e.g. 20 mL of 0.1 mg/mL solution (2 mg) and 18 mL of 1.0 mg/mL (18 mg) is minimal. An increased value for sourness for the highest dose of 18 mg cannot be explained, as all solutions have the same citric buffer solution in common and the dose volume was lower for 18 mg than for 6 mg. In this context it is noteworthy that also the placebo was rated quite high with regards to sourness, taste intensity and unpleasantness. This result indicates a significant likelihood to obtain remarkable variation in the rating if the number of subjects in the study is relatively small.

The sensor response pattern (Figure 6) showed comparable results of the e-tongue measurements (Figure 6B) to the human taste assessment (Figure 6A): With increasing concentrations of the API CSE3165, only the bitterness values with cationic sensor SB2BT0 were increased. In comparison, the human taste pattern showed just moderately increasing values for bitterness, but no clear increase in the perception of “unpleasant taste”. In addition, an increase in perception of sourness was denoted due to higher volumes of the sample buffer with citric acid. These results demonstrated the suitability of the electronic tongue to assess the taste of the drug formulation containing CSE3165. According to the whole patterns that are presented in Figure 6, the taste of CSE3165 in citric buffer can be described as bitter and sour, which can be seen as equivalent with an unpleasant taste. This was true in particular for the higher doses and volumes, which therefore would need improvement by an appropriate taste-masking.

11

3.2.2. In-vitro evaluation of the final CSE3165 formulation The evaluation of the taste patterns of CSE3165 demonstrated a huge influence of the buffer system on the sensor responses. High concentrations of ions in the buffer can interfere with the interactions of the sample molecules and the sensor membranes which might lead to a decreased sensitivity of the sensors towards the sample molecules. This superimposing effect of the buffer system was verified by analysis of two concentrations of the bitter model quinine sulfate (0.001 and 0.01 mM) in tartrate buffer of 0.2 mM, 2 mM and 20 mM pH 3.5: the most sensitive sensor SB2AC0 significantly discriminated the two concentrations (0.001 mM: -62.3 mV ± 7.0 mV; 0.01 mM: -51.02 mV ± 4.5 mV) in tartrate buffer of 0.2 mM and considerably detected the two concentrations in 2 mM tartrate buffer (0.001 mM: -61.2 mV ± 6.1 mV; 0.01 mM: -51.3 mV ± 4.6 mV), while same quinine sulfate concentrations showed weak and overlapping sensor signals in tartrate buffer of 20 mM (0.001 mM: -48.0 mV ± 5.2 mV; 0.01 mM: -43.1 mV ± 4.4 mV). These findings were used for the following investigations on CSE3165 formulation development. The second approach to guide the development of a palatable liquid drug formulation included the taste assessment of final Phase II formulation. The development of a formulation for compound CSE3165 resulted in a complex system at pH 3.5 using tartaric acid at concentrations of 7.7 mM for 0.1 and 0.25 mg/mL CSE3165 as vehicle 1, and 13.6 mM for 1.5 mg/mL as vehicle 2, retrospectively (composition shown in Table 3). Both systems contained furthermore sodium benzoate as preservative. These systems still did not meet the sensitivity requirements for a robust analysis by electronic tongue measurements. However, this formulation was developed to gain sufficient in-use stability of the drug solution and was therefore chosen for in-vitro analysis as well. Again, the sensor system was calibrated with the pure API CSE3165 dissolved in the buffer system containing 11.2 mM tartaric acid, 2 mM ascorbic acid and 0.5 mM disodium edetate (0.05 mg/mL – 1.5 mg/mL) and afterwards three final formulations and their corresponding placebo solutions were analyzed. Univariate evaluation of the sensor responses towards the CSE3165 calibration samples showed a high sensitivity dependent on the concentration at sensors SB2AC0, SB2AN0 and SB2AAE and lower sensitivity at the other sensors. However, a log-linear correlation between the API concentration and the sensor signals were obtained only for concentrations higher 0.1 mg/mL. These results show that even at a higher buffer concentration the API could be concentration dependently detected. Sensors SB2AC0, SB2AN0 and SB2AAE (cationic bitterness and umami) that were most sensitive towards CSE3165 (Figure 5), showed sensor responses for the drug formulations containing either 0.1 mg/mL or 0.25 mg/mL CSE3165 guiding in the direction of the corresponding API sample in tartrate buffer. These drug formulations were prepared in the same vehicle (vehicle 1). The drug formulation containing 1.5 mg/mL CSE3165 was prepared using a vehicle (vehicle 2), with higher amount of tartaric 12

acid in order to dissolve the amount of API with the same target pH of 3.5 in the final drug solution leading to contrary sensor responses. The complex taste pattern of the final drug formulation was evaluated by performing a principal component analysis with two principal components based on all 8 applied electronic tongue sensors. The PCA map is shown in Figure 8 where 71.7 % of the information is displayed on the x-axis and 26.9 % on the y axis. The calibration samples containing CSE3165 in different concentrations were clustered in the lower and upper right quadrants and could be differentiated by the API concentration. In the lower left quadrant, drug formulations prepared within the vehicle 1 containing either 0.1 or 0.25 mg API/mL. This result complied with the univariate evaluation (Figure 7), showing considerable differences in the sensor responses for drug formulations and calibration samples, especially for sensors SB2C00 and SB2AE1. These two sensors were determined to have a noticeable impact on the scores of the analyzed samples, as they were located in the upper right quadrant of the loadings plot farthermost from the origin. These two sensors have already shown the most prominent differences in the sensor responses for the CSE3165 calibration samples and the drug formulations. As displayed in Figure 3 for CSE3104, these sensors showed a high sensitivity to the samples containing sodium benzoate within the drug formulation. The drug formulations containing CSE3165 did also contain sodium benzoate as preservative additive. The drug formulation prepared in vehicle 2 was arranged in the upper left quadrant of the PCA map (Figure 8). Due to the different vehicle of this drug formulation, it caused contrary sensor responses for some of the applied sensors (compare Figure 7). Due to the higher amount of tartaric acid in vehicle 2, which was necessary for the dissolution of a higher API content, the protonated proportion of the incorporated sodium benzoate was also higher than for vehicle 1, since the pKa values of CSE3165 and benzoic acid are very close to each other, i.e. 4.5 versus 4.2. This higher amount of protonated sodium benzoate has an impact on the e-tongue sensors working according to a potentiometric measurement principle. However, these differences in the absolute sensor responses did not affect the relative comparison of the sensor responses among samples using the same vehicle, but did also not permit a direct comparison between samples prepared with either vehicle 1 or 2.

The influence of the buffer system, pH value and incorporated excipients within the formulation demonstrated the high sensitivity of the e-tongue sensors to critical properties of the samples. A univariate evaluation of the results showed furthermore the capability of some sensors to resist against these influences. Therefore, a multivariate analysis of the whole sensor response pattern might lead to wrong interpretation of the taste pattern. In this case a univariate evaluation of the data seems to be preferable since it allows for isolated interpretation of the intensity of the sensor signal of a particular sensor that is dedicated to a particular taste. 13

Euclidean distances (Figure 10) can be used as an indicator for the taste-masking capability of the vehicle [23-25]. Thereby, the differences in the taste patterns of two samples are expressed by the sensor responses of the whole sensor set. A concentration dependent, bitter and unpleasant taste is generally assumed for the drug substance (pure drug solution and formulations) whereas the formulated vehicle (placebo) is assumed to have good taste. The less the taste pattern of the pleasant tasting vehicle is changed by the drug concentration (formulation), the better the vehicle is suitable to mask the bitter and unpleasant taste of the drug substance. Concurrently, large changes in the taste pattern between the pure drug solutions and drug formulations of the same drug concentration indicate a significant difference in the taste pattern of those samples. However, such interpretation of the data as described is only suitable if the pure drug solutions are prepared with the same solvent as the placebo and drug formulation e. g. water. In this case no impact on the taste of the drug solution is assumed for the sample solvent. In this study, the pure drug solutions were prepared in a buffer solution as described in Table 3. The components of the chosen vehicles 1 and 2 (see Table 3) did not exactly match the buffer solution. This is why they are located at three different corners in the PCA map (Figure 8) and authors decided to additionally calculate and compare Euclidean distances between the pure drug solutions and its pure solvent (sample 0.0 mg/mL in Figure 8) and between the corresponding drug formulation and the according vehicles (Figure 10). With this approach the influence of CSE3165 on the taste pattern of either the pure solvent or the vehicle was evaluated.

Regardless of the different influences of the two vehicles on the final drug formulation, the comparative Euclidean distances did not show a significant difference. The distances between the placebo samples (vehicle) and their corresponding drug formulations were slightly lower than the distances between the pure drug solutions and their solvent but only for the low concentrated drug containing samples . CSE3156 1.5 mg/mL showed larger Euclidean distances for the drug solution and its corresponding solvent indicating a concentration dependency of the sensor response pattern. These results supported the assumption that a higher drug amount caused larger impacts on the taste pattern. Based on the obtained data, no taste masking effect could be determined neither for vehicle 1 nor vehicle 2 using electronic tongue measurements and the described methods for data evaluation and interpretation. The sensor response pattern and therefore the assumed taste pattern of the analyzed samples were comparably changed by the addition of CSE3165, regardless from the formulation or pure drug solutions. The lacking proof of a taste masking effect could further be caused by low interactions between the sensor array and compound of the vehicles such as volatile flavors or sugar alcohols. 4. Conclusions Taste sensing systems cannot predict the human taste perception of an API or a drug formulation, but they can be useful tools in the early stage of drug development of taste-masked formulations. Especially 14

for NCEs in pediatric drug development, electronic tongue measurements are capable to address both: the ethical concerns of taste assessment by human taste panels and the official requirements for an appropriate taste of the new pediatric drug product. Two different strategies for the application of electronic tongues for the development of taste-masked pharmaceuticals were presented in this study. Both were based on a prior e-tongue study, which results were afterwards verified by in-vivo taste data. Subsequently, either an electronic tongue guided formulation development or evaluation of the taste pattern of drug formulation candidates were conducted. The applied approach of a retrospective comparison of in-vivo taste assessments with the results obtained from the taste sensing system has confirmed the suitability of the electronic tongue for early preclinical assessment of potential taste issues, in particular for APIs with at least pH dependent solubility and ionic character in solution The dose dependent increase in perception of unpleasant, bitter taste was found for CSE3104 in alignment with the in-vitro results with sensors for bitter taste. With CSE3165 a significantly less pronounced API concentration dependent response of the bitter sensors was found, which was also in alignment with weak to moderate increase in the scoring for bitter taste in the in-vivo taste assessment. Linking the e-tongue results with the received data from human taste assessment, the suggested bitter taste by e-tongue measurements could be confirmed despite the high overall scattering of the human data. As a consequence, the taste assessment results from a very limited number of subjects in a Phase I study could be effectively supported by applying the electronic tongue. Using this instrumental approach for the taste assessment led to improved data without ethical concerns and a reduction of time and costs: By combining the two data sets, which on their own might not have given a clear answer, more conclusive results were obtained.

15

5. References 1 Regulation (EC);1; n. 1902/2006 of the European Parliament and of the Council of 20 December 2006 amending Regulation 1901/2006 on medicinal products for paediatric use. Off J Eur Union, 2006. L378: p. 20-21.

2 Regulation (EC);1; no 1901/2006 of the European parliament and of the Council of 12 December 2006 on medicinal products for paediatric use and amending Regulation (EEC) No 1768/92, Directive 2001/20/EC, Directive 2001/83/EC and Regulation (EC) No 726/2004. Official Journal of the European Union, 2006. L378: p. 1-19.

3 Liu, F., et al.,;1; Formulation factors affecting acceptability of oral medicines in children. International journal of pharmaceutics, 2015. 492(1): p. 341-343.

4 Sohi, H., Y. Sultana, and R.K. Khar,;1; Taste masking technologies in oral pharmaceuticals: recent developments and approaches. Drug development and industrial pharmacy, 2004. 30(5): p. 429-448.

5 Wagh, V.D. and S.V. Ghadlinge,;1; Taste masking methods and techniques in oral pharmaceuticals: current perspectives. Journal of Pharmacy Research Vol, 2009. 2(6).

6 Mohamed-Ahmed, A.H.A., et al.,;1; Non-human tools for the evaluation of bitter taste in the design and development of medicines: a systematic review. Drug Discovery Today, 2016. 21(7): p. 11701180.

7 Legin, A., et al.,;1; Electronic tongue for pharmaceutical analytics: quantification of tastes and masking effects. Analytical and Bioanalytical Chemistry, 2004. 380(1): p. 36-45.

8 Cram, A., et al.,;1; Challenges of developing palatable oral paediatric formulations. International Journal of Pharmaceutics, 2009. 365(1): p. 1-3.

9 Kobayashi, Y. and H. Ikezaki,;1; Advanced Taste Sensors Based on Artificial Lipid Membrane. Biochemical Sensors: Mimicking Gustatory and Olfactory Senses, 2013: p. 5.

16

10 Toko, K., Taste sensor with global selectivity. Materials Science and Engineering C,;1; 1996. 4(2): p. 69-82.

11 Zheng, J.Y. and M.P. Keeney,;1; Taste masking analysis in pharmaceutical formulation development using an electronic tongue. International journal of pharmaceutics, 2006. 310(1): p. 118-124.

12 Woertz, K., et al.,;1; Taste sensing systems (electronic tongues) for pharmaceutical applications. International journal of pharmaceutics, 2011. 417(1): p. 256-271.

13 Anand, V., et al.,;1; The latest trends in the taste assessment of pharmaceuticals. Drug Discovery Today, 2007. 12(5–6): p. 257-265.

14 Takács-Novák, K., K.J. Box, and A. Avdeef,;1; Potentiometric pK a determination of water-insoluble compounds: validation study in methanol/water mixtures. International journal of pharmaceutics, 1997. 151(2): p. 235-248.

15 Lewis, R.J.S.,;1; Hawley's Condensed Chemical Dictionary 15th Edition. Inc. New York, NY 2007: p. 1068.

16 Pein, M., et al.,;1; Interlaboratory testing of insent e-tongues. International journal of pharmaceutics, 2014. 469(2): p. 228-237.

17 Woertz, K., et al.,;1; Performance qualification of an electronic tongue based on ICH guideline Q2. Journal of pharmaceutical and biomedical analysis, 2010. 51(3): p. 497-506.

18 Kobayashi, Y., et al.,;1; Advanced taste sensors based on artificial lipids with global selectivity to basic taste qualities and high correlation to sensory scores. Sensors, 2010. 10(4): p. 3411-3443.

19 Walsh, J., et al.,;1; Playing hide and seek with poorly tasting paediatric medicines: do not forget the excipients. Advanced drug delivery reviews, 2014. 73: p. 14-33.

17

20 Preis, M., et al.,;1; A comparative study on solubilizing and taste-masking capacities of hydroxypropyl-β-cyclodextrin and maltodextrins with high amylose content. Sensors and Actuators B: Chemical, 2014. 193: p. 442-450.

21 Momin, M., S. Rathod, and S. Kar,;1; K. Taste masking techniques for bitter drugs-an overview. Int J of Pharmacy and Technology, 2012. 4(2): p. 2100-18.

22 Immohr, L.I. and Pein-Hackelbusch, M.,;1; Development of stereoselective e-tongue sensors considering the sensor performance using specific quality attributes—A bottom up approach. Sensors and Actuators B: Chemical, 2017. 253: p. 868-878.

23 Immohr, L.I., et al.,;1; Suitability of E-tongue Sensors to Assess Taste-Masking of Pediatric Liquids by Different Beverages Considering Their Physico-chemical Properties. AAPS PharmSciTech, 2016: p. 1-11.

24 Petrovick, G.F., J. Breitkreutz, and M. Pein-Hackelbusch,;1; Taste-masking properties of solid lipid based micropellets obtained by cold extrusion-spheronization. International Journal of Pharmaceutics, 2016. 506(1–2): p. 361-370.

25 Tokuyama, E., et al.,;1; Famotidine orally disintegrating tablets: bitterness comparison of original and generic products. Chemical and Pharmaceutical Bulletin, 2009. 57(4): p. 382-387.

18

Figures Legend Figure 1: Calibration samples of the API CSE3104 in citric buffer (50 mM, pH 4.5); measured with eight different sensors and the Insent SA402B: 3 bitter sensors in top graph; pure buffer samples are plotted on the y-axes (blank), (n = 3, mean ± sd).

19

Figure 2: Results of the human taste assessment of CSE3104 (A) and e-tongue measurements (B); scaling of the axes belong to (A) the scoring of the panelists (0-5) (mean ± sd, n =6, placebo: n=2) or to (B) sensor responses in mV (mean ± sd, n = 3).

Figure 3: Score Scatter Plot (PCA map) for CSE3104 containing the information of all samples based on the detection by all 8 sensors (n= 3, ctr scaled). (R2 ~ 0.941, Q2 ~ 0.570).

20

Figure 4: Loading Scatter Plot of the Score Scatter Plot displayed in Figure 3.

Figure 5: Calibration samples of the API CSE3165 in citric buffer 50 mM pH 3; measured with eight different sensors and the Insent SA402B, (n = 3, mean ± sd).

21

Figure 6: Results of the human taste assessment (A) and e-tongue measurements (B) of CSE3165 in citric buffer formulation; scaling of the axes belong to (A) the scoring of the panelists (1-5) (mean ± sd, n = 3 for 0.6 and 2mg, n = 6 for placebo, 6 and 18 mg,) or to (B) sensor responses in mV (mean ± sd, n = 3).

Figure 7: Univariate evaluation of the calibration series of the API CSE3165 and the formulation samples (mean ± sd, n = 3).

22

Figure 8: Score scatter plot of the principal component analysis of the CSE3165 calibration samples and the final drug formulations together with their corresponding placebos (mean, n = 3, ctr scaled, R² ~ 0.986, Q² ~ 0.881).

23

Figure 9: Loadings plot of the Score Scatter plot in Figure 8.

Figure 10: Euclidean distances for drug formulations and either their corresponding vehicles or drug solutions (mean ± sd, n= 3).

24

Tables

Table 1: Characteristics of CSE molecules compared to those of quinine

molecule

pKs

solubility

MR [g/mol] CSE3104 3.1 and 10.9 very good pH dependent solubility in aqueous buffers < 500 CSE3165 4.5 and 6.8 only sufficient and strongly pH dependent solubility at a pH < 500 <4 quinine 4.2 and 8.6 [14] soluble in alkalis and acids (with formation of salts) [15] 324

Maltodextrin Linecaps 17

Betadextrin (Kleptose)

Hydroxypropylce llulose LF

Maltodextrin DE15-18

Lactose anhydrous

Mannitol

CSE3104

labeling

Table 2: Labeling and content of the investigated formulations (DF = drug formulation, P = placebo)

1.00

1.00

0.25

% (m/V) DF_01 DF_02 DF_03 DF_04 DF_05 DF_06 P_01 P_02 P_03 P_04 P_05 P_06

0.50 x x x x x x

2.50 x

2.50

1.00

x x x x x x x x x Stock solution (1000 mg dose) x x x x x x x x x x

x

x

25

Table 3: Composition of the calibrations samples and the vehicles 1 and 2 used for the CSE3165 Phase 2 formulations

solution strength: Mannitol [mg] Tartaric acid [mg] Sodium benzoate [mg] ascorbic acid [mg] Polyethylene glycol 6000 [mg] disodium edetate [mg] Sucralose [mg] Strawberry flavor [mg] Water for injection

Solutions for calibration (in buffer) 0.05-1.5 mg/ml

Formulations for Phase 2 clinical study in vehicle 1 in vehicle 2

-168.5 -35.2 -18.6 --Ad 100.0 ml

0.1 mg/ml and 0.25 mg/ml

1.5 mg/ml

2524.9 115.0 80.0 35.2 31.3 18.6 20.0 300.0 Ad 100.0 ml

2435.8 204.1 80.0 35.2 31.3 18.6 20.0 300.0 Ad 100.0 ml

Table 4: Study design of CSE3104 and CSE3165 Phase 1 dose escalation study Study design

Single-center, double-blind, randomized, placebo-controlled, single ascending dose (SAD) study with a parallel group design

Number of

CSE3104: 6 cohorts with 8 subjects each (6 on active drug, 2 on placebo)

subjects

CSE3165: 4 cohorts, first cohort with 5 subjects each (3 on active, 2 on placebo), second cohort with 4 subjects (3 on active, 1 on placebo) last 2 cohorts with 8 subjects (6 on active, 2 on placebo)

Target

Healthy male subjects aged 23 (CSE3104)/18 (CSE3165) - 45 years, inclusive.

population

26

Table 5: Summarized results of the human taste assessment of CSE3104 (n = 6, mean ± sd); placebo and active solutions; “solvent”: citric buffer (50 mM, pH 4.5)

placebo

0.5 mg

4 mg

14 mg

40 mg

80 mg

taste intensity

2.8±0.6

3.3±0.5

3.8±0.8

4.2±0.8

3.7±1.0

4.0±0.6

unpleasant

2.9±0.5

3.5±0.8

3.2±0.4

4.0±0.6

4.0±0.6

4.0±0.6

saltiness

1.9±1.1

1.7±1.0

2.0±1.1

2.7±0.5

1.8±1.0

1.7±1.2

sourness

2.6±1.4

3.7±0.5

2.0±1.2

2.7±0.8

2.2±1.3

2.5±1.2

sweetness

1.5±0.5

1.0±0.0

1.5±0.8

1.0±0.0

1.0±0.0

1.2±0.4

metallic

2.2±1.3

2.0±1.3

2.2 ±1.2

2.0±0.7

2.0±1.1

2.3±1.0

bitterness

1.9±1.0

2.8±1.5

3.2±0.8

4.2±0.8

3.7±0.8

4.0±0.6

27

Table 6: Summarized results of the human taste assessment of CSE3165 (n = 3 for 0.6 and 2mg, n = 6 for Placebo, 6 and 18 mg, mean ± sd); placebo and active solutions; “solvent”: 30 mM citric buffer pH

2.7 to 2.9 containing in addition 2 mM ascorbic acid and 0.5 mM disodium edetate

taste intensity unpleasant bitterness saltiness sweetness sourness metallic

Placebo 3.1±1.1 3.3±1.1 1.4±0.8 1.3±0.5 1.9±1.1 2.3±1.1 1.6±1.1

0.6 mg 2.3±0.6 2.7±0.6 1.3±0.6 1.0±0.0 1.3±0.6 2.7±1.5 1.7±0.6

2 mg 3.3±0.6 3.3±0.6 2.3±1.5 1.0±0.0 1.0±0.0 2.7±1.2 3.3±1.2

6 mg 2.7±0.5 3.0±0.6 2.3±1.0 2.2±1.0 1.5±0.8 2.2±1.0 1.8±1.0

18 mg 3.3±0.8 3.3±0.8 2.5±1.1 1.7±0.5 1.5±1.2 3.7±1.0 1.8±1.3

28