A simple, reliable method for high-throughput screening for diabetes

Journal of Pharmacological and Toxicological Methods 82 (2016) 83–89

Contents lists available at ScienceDirect

Journal of Pharmacological and Toxicological Methods journal homepage: www.elsevier.com/locate/jpharmtox

A simple, reliable method for high-throughput screening for diabetes drugs using 3D β-cell spheroids Jesal Amin a, Karthik Ramachandran c, S. Janette Williams b,c, Annie Lee d, Lesya Novikova b, Lisa Stehno-Bittel b,c,⁎ a

University of Missouri-Kansas City School of Medicine, 2464 Charlotte St., Kansas City, MO 64108, United States University of Kansas Medical Center, Department of Physical Therapy and Rehabilitation Science, 3901 Rainbow Blvd., Kansas City, KS 66160, United States c Likarda, LLC, 2002 W 39th Ave, Kansas City, KS 66103, United States d Rockhurst University, Department of Chemistry, 1100 Rockhurst Rd, Kansas City, MO 64110, United States b

a r t i c l e

i n f o

Article history: Received 27 May 2016 Received in revised form 8 August 2016 Accepted 11 August 2016 Available online 20 August 2016 Keywords: Diabetes High-throughput screen Glybenclamide Caffeine β-cell Glucose Diazoxide Nifedipine

a b s t r a c t Early screens for new diabetes drugs rely on monolayers of β-cells, which are known to be poor predictors of the in vivo response. Previously, we developed a method to create uniform islet spheroids from freshly-dispersed human donor tissue for drug screening. While the human engineered islets worked well to reduce donor-todonor variability, it is difficult and expensive to obtain sufficient high-quality human islets for drug testing. Thus, this study utilized a genetically-modified β-cell culture line (INS-1832/13) in 2D and as 3D spheroids and compared the results to human islet tissue formed into spheroids using a high-throughput 384-well format. In response to increasing concentrations of glucose, all 3 groups increased insulin release, but the cultured β-cells (2D and 3D) were more sensitive to glucose (EC50 5.85 mM for 2D β-cells, 16.24 mM for 3D β-cell spheroids) than the human islet spheroids (EC50 53.69 mM). The order of responses to glybenclamide was human spheroids N 3D β-cell culture N 2D β-cell culture. In response to caffeine, the β-cells in 2D or 3D were more responsive compared to the human islet spheroids (EC50 0.39 and 0.31 mM for 2D and 3D β-cells respectively). When exposed to inhibitors of insulin secretion (nifedipine and diazoxide), the responses were more similar between groups. Z' calculations, indicative of assay quality, determined that the 3D β-cell spheroids reached the criteria of an excellent to ideal drug screen assay more consistently than the other test models. In conclusion, 3D β-cell spheroids from a cultured cell line can be used in HTS assays that, according to reference drugs tested here, are sensitive and predictive of the in vivo response. © 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

1. Introduction The United States, along with other growing countries such as India and China, have been affected by a dramatically increasing prevalence of diabetes. Today, the etiology of both type 1 and type 2 diabetes is thought to revolve around the dysfunction of β-cells, the insulin producing cells of the body (Law, Ashcroft, Zheng, & Sexton, 2014) (Lupi & Del Prato, 2008). When β-cells are attacked either by chronic inflammation or autoimmunity, the loss of insulin production leads to increased blood glucose levels and eventually a diagnosis of diabetes. Thus, even in cases of undiagnosed diabetes, reduction in the β-cell count or impaired β-cell function are present (Ferrannini & Mari, 2014; Watkins, Evans-Molina, Blum, & DiMeglio, 2014). Within the pharmaceutical industry, secondary testing of new potential compounds for diabetes often relies on assays aimed at ⁎ Corresponding author at: University of Kansas Medical Center, Department of Physical Therapy and Rehabilitation Science, 3901 Rainbow Blvd., Kansas City, KS 66160, United States E-mail address: [email protected] (L. Stehno-Bittel).

measuring an impact on insulin secretion and often uses of β-cell insulinoma cell lines and rodent islets (Walpita & Wagner, 2014). Rodent islets are not physiologically or morphologically similar to human islets (Ramachandran, Huang, & Stehno-Bittel, 2014), thus they are poor predictors of medicinal outcomes in humans. This lack of predictability has been cited as one of the reasons for the excessive expense associated with bringing a new drug to market (DiMasi, Grabowski, & Hansen, 2014). For the testing of new diabetes drugs, β-cell culture lines are created from transformed human or rat pancreatic β-cells. Grown in standard monolayers, these cell lines secrete insulin in response to some compounds (Burns et al., 2015), but in general are poor predictors of how the human body will respond to the compound in vivo (Rupnik, 2009; Wikstrom et al., 2012). The exception may be the adapted insulinoma cell line, INS-1 832/13, a β-cell line that has been shown to respond well to increasing glucose concentrations (Hohmeier & Newgard, 2004), with a rapid increase in ATP/ADP and malonyl CoA activity (Lorenz, El Azzouny, Kennedy, & Burant, 2013). In addition, INS-1 832/ 13 cells show signs of forming aggregates of ubiquitinated proteins in response to high glucose – something noted in the intact pancreas

http://dx.doi.org/10.1016/j.vascn.2016.08.005 1056-8719/© 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

84

J. Amin et al. / Journal of Pharmacological and Toxicological Methods 82 (2016) 83–89

(Kaniuk et al., 2007). INS-1 832/13 is currently considered one of the most physiologically relevant β-cell models available (Hectors, Vanparys, Pereira-Fernandes, Martens, & Blust, 2013). Importantly, these cells can maintain a differentiated phenotype for 6 months in culture (Hohmeier et al., 2000), and the cells have been used in screening environmental pollutants that could impact β-cell function (Hectors et al., 2013; Hohmeier et al., 1997). To date, there have been no published studies using the INS-1 832/13 β-cell culture line to create 3D spheroids for drug screening in the high-throughput format. Previously, we developed and tested a simple method for the creation of 3D spheroids that did not require scaffolds, gels or 3D printers. Rather, the procedure builds on the inherent self-aggregating properties of most cells (Ramachandran, Williams, Huang, Novikova, & StehnoBittel, 2013). Created in micromolds, the process can create 200,000 uniform spheroids in a standard cell culture plate. Later, we illustrated the ability of 3D reaggregated human islet spheroids, which we call Kanslets, to respond to known insulin secretagogues and insulin secretion inhibitors in a manner that was similar to the in vivo response (Ramachandran et al., 2014). We showed that they were superior to intact human islets in drug screens using reference compounds. However, creating the human islet spheroids still required donor islet tissue, which is expensive to obtain and there is often not enough healthy tissue present to run large-scale compound screens. The purpose of this study was to determine whether the same reaggregation process could be used to form 3D β-cell spheroids from a cultured cell line and to compare the results with the same cells in 2D and with reaggregated human islets. In addition, the procedure was modified for the high-throughput 384-well format. Reference drugs with known in vivo insulin secretion or insulin inhibition responses were tested on the 3 cell models.

2.3. Human Kanslet formation Human islets from 9 adult donors were obtained from the Integrated Islet Distribution Program (IIDP). Isolated islets were maintained in CMRL 1066 medium with 2 mM L-glutamine, 10% FBS and 1% antibiotic/antimycotic at 37 °C in a culture chamber containing 5% CO2. Islets were dispersed as described previously by digestion in cmf-HBSS supplemented with papain (10 units/ml; Worthington, Lakewood, NJ) following washing in media free of calcium and magnesium (Ramachandran et al., 2014). Suspensions were incubated on a rotator at 37 °C for 10–15 min, and manually dispersed with a pipette until the cell suspension primarily contained single cells. The cells were then washed to remove residual papain and transferred to an aggregate culture medium, consisting of CMRL 1066 supplemented with 2 mM Lglutamine, 10% FBS and 100 U penicillin, 100 μg streptomycin, and 25 μg amphotericin. The single cell suspension was then plated onto sterile glass micromolds using the same protocol described above for the cultured β-cells. Cells in the micromolds were incubated for 3–5 days at 37 °C and 5% CO2. Cells reaggregated while limiting the size to no bigger than the recess allowed (100 μm). Aggregate culture medium was changed every 24 to 48 h until reaggregated islets were formed. Reaggregated islet clusters (Kanslets) were removed by washing the micromold several times with culture medium until they were dislodged and were aspirated with a pipette to be plated in CMRLbased medium for long-term culture. Brightfield images of the spheroids forming in the micromold and after removal were captured on an inverted Nikon microscope, Olympus Fluoview Confocal Microscope, using a 20× objective.

2.4. Compound testing 2. Methods and materials 2.1. Cell culture The INS-1 832/13 cell line is a derivation of the INS-1 cell line and was graciously provided by the Sarah W. Stedman Nutrition and Metabolism Center, Duke University, Durham, NC. Cells were cultured in a growth medium consisting of RPMI-1640, along with 1 M HEPES buffer solution, 50× INS supplement (glutamine, Na-pyruvate, and B-mercaptoethanol), Antibiotic-Antimycotic (Gibco by LifeTechnologies), and Fetal Bovine Serum (HyClone). Cells were grown and sustained in a T75 cell culture flask. The cells were placed in a humidified incubator at 37 °C and 5% CO2. Due to the rapid growth of these cells, media was changed on every second day and passaged on every fourth day via trypsinization. At passage, one-fourth of the cells was moved to a T75 flask with the remainder of the cells placed in a micro-mold to allow the creation of uniform, islet spheroids. For this study, cells with the population doubling (PD) of 53-93 were used.

2.2. Formation of β-cell aggregates Custom micromolds were used to create uniform, β-cell spheroids with a maximum diameter of b100 μm (Ramachandran et al., 2014). Cells were loaded into the micromold at a density of approximately 1 million/ml medium. Medium was change daily by partial replacement. Cell aggregates were allowed to form in the micromold for 72 h and then were transferred to a 100 mm non-treated tissue culture dish for an additional 48 h prior to exposure to the test drug. Spheroids were passed through 70 μm and 100 μm cell strainers (#3431751 and #43152, Corning, Tewksbury, MA) and maintained in HEPES Balanced Salt Solution comprised of: 20 mM HEPES, 114 mM NaCl, 4.7 mM KCl,1.2 mM KH2PO4,1.16 mM MgSO4, 2.5 mM CaCl2, 25.5 mM NaHCO3 and 0.2% bovine serum albumin (fatty acid free), pH 7.2.

To test the glucose- or compound-stimulated insulin secretion or inhibition, cells in the 3 test conditions (2D β-cell culture, 3D β-cell spheroids or human Kanslets) were reconstituted in 1× EBSS supplemented with 0.1% BSA and low (2.8 mM) glucose and dispersed into 384-well plates. The average loading cell density for the 2D screens was 5 × 103 cells/well in the 382-well plates and approximately 2 × 104 cells/well in the pilot studies conducted in 96-well plates. To verify loading density, images of each well were captured. Experiments were run in replicates of quadruplicates, with each dose response relationship tested independently 3 times. Each well was subjected to a 60-minute static incubation at 37 °C and 5% CO2 in EBSS containing varying doses of glucose and/or test compounds. Compounds, prepared in EBSS, were added to each well to achieve desired compound dose and a final concentration of 11.2 mM glucose (unless otherwise noted) using procedures we have published previously (Ramachandran et al., 2014). Optimization of each test condition (2D, 3D or human Kanslets) led to the determination that ideal results were obtained with approximately 30 spheroids/well. After the 60 min incubation with test compounds, the secreted insulin was tested via Alpha-LISA test system (Perkin Elmer). Briefly, a twostage approach was utilized, which involved adding acceptor solution (acceptor beads, antibody and buffer) to the spheroids initially and 15 min later adding the donor solution (donor beads and buffer). The signal was read on an EnSpire Mulitmode Plate Reader (Perkin Elmer) with excitation 680 nm and emission measured at 615 nm. Values were converted to insulin concentration, and dose/response curves were determined for each single assay and single-run assay quality was assigned based on the signal window. Values of blank wells (media, with vehicle where appropriate) were subtracted from each well, and insulin concentration values were normalized by dividing by the 0 drug wells (media + cells). Further quality assurance was determined by calculating the signal-to-noise ratio (S/N) and Z' values using equations provided in the results section.


In order to validate the approach with 3D spheroids, separate studies were conducted using standard ELISA procedures. After incubation with the test compound, the conditioned medium was collected and kept frozen at − 80 °C for later determination of insulin content by ELISA. Human insulin was detected using the human ELISA kit (Mercodia Uppsala, Sweden) and rat insulin was detected with the rat version (APLCO, #80-INSRTH-E01, Salem, NH).

2.5. Assay quality Assay quality was assigned based on the signal window as defined in NIH's Assay Guidance Manual (Iversen et al., 2004). The signal-to-noise (S/N) calculation takes into consideration the amplitude of the test response over the background. Each plate used in an assay included a positive control, designed to elicit a maximal insulin secretion from the sample. In this study, the positive control consisted of a depolarizing solution in high glucose (35 mM K+ in 22 mM glucose). The S/N was calculated as: S=N ¼ ðmean signal−mean backgroundÞ =standard deviation of background

85

2.6. Statistics One-way ANOVA was used to compare different doses within the same experiment with post-hoc Tukey test when necessary. P b 0.05 was defined as statistically significant. 3. Results 3.1. Formation of 3D β-cell spheroids INS-1 832/13 cells were loaded into the micromold at a density of approximately 1 million cells/ml. Fig. 1A shows a representative picture of the divots within the mold containing 15–30 cells/divot only 20 min after seeding. 3 days later, the same mold shows clearly delineated spheroids in every divot (Fig. 1B). At that time, spheroids were removed from the micromold as shown in Fig. 1C and washed. The β-cell spheroids had the same general appearance and size as the spheroids produced from human islet cells, Kanslets (Fig. 1D). The average diameter of the β-cell spheroids at the time of removal from the micromold was 70 μm, but cells continued to replicate after removal from the mold. At the time of drug testing, the β-cell spheroids were between 70 and 150 μm in diameter. Kanslet spheroids were under 80 μm in diameter. 3.2. Insulin-secretion stimulation: glucose stimulation

Z' is another quality assurance measure, which has been shown to lower the influence of outliers (Gubler, 2006). The robustness of each response was determined based on the Z' following the procedures outlined by Zhang, Chung, and Oldenburg (1999). Z0 ¼ 1−½ð3 control standard deviation þ 3 sample standard deviationÞ =ðcontrol mean þ sample meanÞ

The response of isolated islets to increasing doses of glucose is a hallmark of healthy β-cell function, and is commonly used as a pre-screening test on freshly isolated tissues. β-cells cultured into 3D spheroids or in a 2D monolayer and human Kanslets were each exposed to increasing doses of glucose from 1.4 to 33.6 mM for 60 min. This range of glucose concentrations is the equivalent of 25 to 660 mg/dl, thus covering an equivalent physiological range of severe hypoglycemia to extreme

Fig. 1. Formation of 3D spheroids. A) A representative picture of the divots within the mold containing 15–30 cells/divot 20 min after seeding. Scale bar = 50 μm. B) 72 h later, the same mold shows mature spheroids in every divot. Scale bar = 50 μm. C) β-cell spheroids were removed from the micromold had the same general appearance and size as the spheroids produced from D) human islet cells, Kanslets. Scale bar for both = 100 μm.

86


Kanslets. The SI values for the 2D and 3D β-cells were not statistically different from each other, but the value for the human Kanslets was statistically lower than the other 2 groups.

3.3. Insulin secretogogues: glybenclamide

Fig. 2. Glucose-stimulated insulin secretion. Human Kanslets along with β-cells cultured in 2D or as 3D spheroids were exposed to increasing doses of glucose from 1.4 to 33.6 mM for 60 min. All 3 groups responded appropriately to the increasing concentration of glucose by releasing more insulin, but the human Kanslets were the least sensitive. N = 9 trials, 4 repetitions each for human Kanslets and 3 trials each for β-cell groups. ⁎indicates statistically different values, p b 0.001.

hyperglycemia. All 3 groups responded appropriately to the increasing concentration of glucose by releasing more insulin (Fig. 2). Specifically, the human Kanslets were the least sensitive to glucose requiring exposure to 28 mM glucose before a statistically measurable increase in insulin concentration was measured (minimal detectible dose, Table 1). The β-cell spheroids had a minimal detectible dose of 16.8 mM glucose with an estimated EC50 of 16.24 mM glucose. In contrast, the 2D monolayer of β-cells demonstrated the greatest sensitivity to the glucose with the lowest dose necessary to identify a statistically significant increase of basal conditions and the lowest EC50 value (Table 1). The slopes of the 3 responses were quite different with the steepest slope found in the 2D culture followed by the 3D β-cell spheroids and Kanslets (Table 1). The stimulation index (SI) is the percent increase in insulin secretion when comparing the highest glucose concentration to the lowest. For the 3 conditions in 33.6 mM glucose, the mean SI values were 4.50 for 2D cell culture, 4.31 for 3D β-cell spheroids, and 1.94 for human

Glybenclamide is an anti-diabetic drug that belongs to the sulfonylurea class. Its primary mechanism of action is to block the ATP-regulated potassium channel and induce and/or stimulate electrical activity in the β-cell, which in turn enhances insulin secretion (Garcia-Barrado, Jonas, Gilon, & Henquin, 1996). At 1 μM glybenclamide, islets were shown to have an enhanced response to 15 mM glucose, thus a range of concentrations at and below 1 μM were tested (Garcia-Barrado et al., 1996). The human Kanslets secreted more insulin with increasing doses of glybenclamide, indicating that it was more efficacious on the Kanslets than either of the β-cell models (Fig. 3). In fact, all values of insulin secretion from the Kanslets were statistically greater than the 0 drug level. Conversely, there was no statistical increase in insulin secretion in the 2D cultured β-cells compared to the basal (0 drug) condition (Table 1). The 3D β-cell spheroids showed an increase in insulin secretion. While the minimal statistically detectible concentration was 10 nM, an EC50 value could not be calculated within the doses shown in Fig. 3. Additional experiments that included higher concentrations of glybenclamide resulted in an EC50 value of 384 nM. 3.4. Caffeine Caffeine acts to increase intracellular Ca2 + by inducing release through the Ca2 + channels of the endoplasmic reticulum, bypassing plasma membrane receptor binding and subsequent second messenger signaling. The Ca2+ transient is a required step in the pathway from glucose stimulation to exocytosis of the insulin-containing granules. Fig. 4 summarizes the results of all three cell models responding well to caffeine concentrations previously shown to increase insulin secretion (Ramachandran et al., 2014). The 2D β-cell culture and the 3D β-cell spheroids had nearly identical values for the minimal detectible dose and EC50 (Table 1). In contrast, only the highest dose of caffeine resulted in a statistically increased amount of released insulin in the Kanslet group, due to high variability. The Hill slope calculated from the dose/

Table 1 Quality assurance values for assays. For each test compound utilized in the study the concentration at which the response was statistically different from baseline (0 drug value) was called the minimal detectible dose and signifies the dose at which a high-throughput screen would be able to detect a difference. The EC or IC50 and the Hill slope provide some information regarding the potency of the compound. The Z' value indicates the quality of the assay along with the robustness of the response to the compound. A Z' value between 0.5 and 1 is considered an excellent assay, with values near or at 1 classified as an ideal assay. The human Kanslets failed to demonstrate a Z' value in the excellent or ideal range for any of the experiments run. The 2D cell culture model had a Z' value in the excellent range in response to caffeine only. In contrast, the 3D β-cell model ranked in the excellent range for every experiment except when stimulated by glucose and glybenclamide. Agonist/antagonist

Cell model

Minimal detectible dose

EC or IC50

Hill slope

Z'

Glucose (mM)

2D β-cell culture β-Cell spheroids Human Kanslets 2D β-cell culture β-Cell spheroids Human Kanslets 2D β-cell culture β-Cell spheroids Human Kanslets 2D β-cell culture β-Cell spheroids Human Kanslets 2D β-cell culture β-Cell spheroids Human Kanslets

5.60 16.80 28.00 na 10.00 1.00 0.30 0.30 10.00 0.03 0.10 0.03 30.00 10.00 30.00

5.85 16.24 53.69⁎ na 384.00 12.05# 0.39 0.31 na 0.06 0.14 0.03 7.69 2.44 8.09

31.66⁎,# 5.15 2.12 na 0.43 1.46 0.61 0.97 0.78 −0.26 −0.66@ −14.39 −0.23 −1.74 −0.94

−3.95 0.11 −0.25 −1.86 0.48 0.16 0.66 0.65 0.23 0.01 0.53 −3.12 0.07 0.72 −0.49

Glybenclamide (nM)

Caffeine (mM)

Nifedipine (μM)

Diazoxide (μM)

⁎ Indicates a statistically significant difference between the 2D β-cell culture and human Kanslets at p b 0.05. # Indicates a statistically significant difference between the 2D β-cell culture and 3D β-cell spheroids at p b 0.05. @ Indicates a statistically significant difference between 3D b-cell spheroids and human Kanslets at p b 0.05.


Fig. 3. Glybenclamide-stimulated insulin secretion. β-cells cultured in 2D or as 3D spheroids along with 3D human Kanslets were exposed to increasing doses of glybenclamide from 1 to 1000 nM for a 60 min static incubation at 37 °C and 5% CO2 in EBSS. The human Kanslets secreted statistically more insulin with each increasing dose of glybenclamide (# indicates p b 0.01). Likewise, the 3D β-cell spheroids showed an increase in insulin secretion. There was no statistical increase in insulin secretion in the 2D cultured β-cells compared to baseline. N = 3 trials, 4 repetitions each. ⁎indicates statistically different values between the 2D cell culture and the other groups, p b 0.005.

87

Fig. 5. Nifedipine-induced insulin secretion inhibition. A1 h exposure to nifedipine resulted in a graded decrease in insulin secretion from all 3 groups. The 3D β-cell spheroids showed the strongest level of inhibition with nifedipine concentrations of 3 μM and higher. The IC50 for nifedipine was between 0.03 and 0.14 for the 3 groups. ⁎indicates statistical significance at only the highest dose for the 3D β-cell spheroids and compared to the human Kanslets at p b 0.05. Results are average of 3 independent trials (4 repetitions at each dose).

response curve for the β-cell spheroids was near 1, the Kanslets and monolayer β-cells were lower (Table 1).

minimal detectible differences that were similar between groups (Table 1). The 3D β-cell spheroids showed the strongest level of inhibition with higher doses of nifedipine compared to the other two groups.

3.5. Insulin secretion inhibitors: calcium channel blockers

3.6. ATP-dependent K+ channel (KATP) activators

Calcium influx during action potentials is essential for the triggering of exocytosis and the subsequent insulin release (Pento, Kagan, & Glick, 1974). Blocking influx through cell surface voltage-dependent Ca2 + channels is a non-selective way to reduce insulin secretion at stimulatory glucose levels (Al-Mahmood, el-Khatim, Gumaa, & Thulesius, 1986). The voltage-dependent Ca2 + channel blocker, nifedipine, was tested in a dose range shown previously to inhibit the normal Ca2+ transient in response to 28 mM glucose (Qureshi, Dejene, Corbin, & Nunemaker, 2015). Exposure to increasing concentrations of nifedipine resulted in a graded decrease in insulin secretion from all 3 groups (Fig. 5). The IC50 for nifedipine was between 0.03 and 0.14 for the 3 groups with

Diazoxide was introduced into clinical use as an anti-hypertension medication that was later shown to inhibit insulin secretion by directly activating KATP channels and altering mitochondrial function, subsequently causes the inhibition of insulin release (Grimmsmann & Rustenbeck, 1998; Jijakli et al., 1996; Wajchenberg, 2007). All three groups tested responded appropriately to diazoxide at doses from 0.3 to 300 μM (Fig. 6). This dose range was chosen because it coincides with the standard test range for diazoxide for cultured beta cells (Zhang et al., 2013). However, the 3D spheroids and the Kanslets were the least effective with approximately a 60% reduction in insulin secretion at the highest dose (300 μM), while the 2D monolayer demonstrated an almost complete loss of insulin release. The calculated IC50 values were very close for all three groups, between 2.44 and 8.09 μM (Table 1). Likewise, the Hill slopes were similar.

Fig. 4. Caffeine-stimulated insulin secretion. All cell models were exposed to increasing doses of caffeine from 0.01 to 10 mM for a 60 min static incubation (4 reps at each dose, 3 separate trials). The cells responded well to caffeine exposures of 300 μM and above. The 2D β-cell culture and the 3D β-cell spheroids had identical values for the minimal detectible dose and the EC50 were similar. In contrast, only the highest dose of caffeine resulted in a statistically increased amount of released insulin in the Kanslet group. Only the 2D and 3D β-cell models were statistically different from each other at the highest dose tested (p b 0.05).

Fig. 6. Diazoxide-induced insulin secretion inhibition. All three groups tested responded appropriately to inhibition by diazoxide at concentrations of 0.3 to 300 μM. However, the 3D spheroids and the Kanslets were the least responsive, while the 2D monolayer demonstrated an almost complete loss of insulin release at 100 μM diazoxide. There were no statistically significant differences in the 3 models tested.

88


3.7. Quality analysis In order to quickly determine whether a test compound deserves further study based on a single large-assay screen, the quality of the screen must be exceptional. Several classical assay analysis methods have been identified and used in the pharmaceutical industry. The signal-to-noise calculation takes into consideration the amplitude of the test response over the background, but it also includes calculations for the standard deviation of the background. Cut points for S/N values are dependent on the assay type and are not widely published, but any value below 2 does not provide enough distinction to determine drug efficacy. Each plate used in an assay included a positive control, designed to elicit a maximal insulin secretion from the sample. In this study, the positive control consisted of a depolarizing solution in high glucose (35 mM K+ in 22 mM glucose). The signal-to-noise ratio of the depolarization signal when applied to 2D β-cells was quite high (27.60), and is consistent with the high sensitivity that the 2D cell culture model demonstrated to increased glucose concentrations in Fig. 2. Microplates with β-cell spheroids had the next highest S/N value at 9.76 with the human Kanslets with the lowest S/N value of 7.34. The Z' value not only characterizes the robustness of the response to the compound, it also evaluates the quality of the assay itself (Zhang et al., 1999). A Z' value between 0.5 and 1 is considered an excellent assay, with values near or at 1 classified as an ideal assay (Zhang et al., 1999). The human Kanslets failed to demonstrate a Z' value in the excellent or ideal range for any of the experiments run (Table 1). The 2D cell culture model had a Z' value in the excellent range in response to caffeine only (Table 1). In contrast, the 3D β-cell model ranked in the excellent range for every experiment except when stimulated by glucose and glybenclamide. 4. Discussion Conventional cell-based drug screening assays are prone to artifacts due to the reliance on monolayers of cultured cells that do not mimic the in vivo situation (Breslin & O'Driscoll, 2013; Xi, Yu, Wang, Xu, & Abassi, 2008). The greatest advances in the use of 3D culture systems for HTS has been completed in the cancer field where the 3D cell culture system has been shown to be more realistic spatially, with biochemical and cellular heterogeneity (Leong & Ng, 2014). The results of such studies have illustrated that different results were obtained when the same molecules were tested on 2D versus 3D cell culture (Walpita & Wagner, 2014). In fact, a large-scale 3D screen identified unique anti-cancer compounds compared to the same screen conducted on 2D cell culture (Howes, Richardson, Finlay, & Vuori, 2014). But even with the more common use of 3D cancer spheroids for drug screens, the researchers often already have a target molecule that they are monitoring (Suzuki et al., 2014). Further, they typically use hydrogels or other scaffolds to create the 3D tumor model (Loessner, Holzapfel, & Clements, 2014; Song, Park, & Gerecht, 2014). In contrast to the cancer field, few studies have used HTS methods and 3D cell culture models to identify molecules targeted to β-cell function or viability (Tran et al., 2014). The lack of such a strategy has been identified as a major limitation to identifying new promising drugs for diabetes care (Akash, Rehman, & Chen, 2013). In fact, even when fresh islets were used for drug screening, they were first dispersed into single cells grown in a 2D monolayer (Beck et al., 2013; Walpita et al., 2012). Our previous publications compared spheroids created from cadaveric islets to native human islets and the pharmaceutical industry standard of rodent islets (Ramachandran et al., 2014). The results showed that reaggregated islet clusters from human islets were superior to intact human islets as a model system for drug discovery. This was easily noted by comparing the stimulation index (SI) values. The SI is a widely used assay to determine the proper function of isolated islets with values above 3 being considered indicative of highly responsive islet

cells (Sakata, Egawa, Sumi, & Unno, 2008). In fact the SI values for the 2D and 3D were both over 4 and were not different from each other, while the value for the human Kanslets was below 2. The current study utilized the INS-1 832/13 cells as a screening tool (Hohmeier et al., 1997, 2000) and compared them to the human Kanslets. This cell line is based on rodent β-cells, but transfected with the human pro-insulin gene, resulting in a cell line that consistently responds to known human insulin secretagogues that act in vivo. This is the first study to show that this cell line could be used in a 3D highthroughput screen format without scaffolds or hydrogels. A successful screen should be predictive of the human in vivo condition. The 3D β-cell spheroids responded to the insulin secretion agonists and inhibitors in the predicted manner. In contrast, the cultured β-cells in 2D failed to respond to glybenclamide, a common treatment for type 2 diabetes. Interestingly, the human islets failed to demonstrate a response to increasing concentrations of glucose. This was a surprising result, as we had previously shown that the human Kanslet was a highly viable screening model (Ramachandran et al., 2014). However, that work was completed in the 96-well format. Moving to the 384-well format for islet cell testing required modifications described in the methods section. While the cultured cells were able to be converted to automated dispensing and testing, the human islet cells showed high donor-to-donor and within-donor variability that was magnified in the 384-well format. The Z' values calculated for this method did not reach ideal levels, but were better than many other reportedly successful screens (Albrecht et al., 2004; Li, Xu, Liu, Christoffersen, & Wang, 2013; Tammela et al., 2004; Zuck et al., 2004), which reinforces the difficulty in working with human islets or β-cell lines. In this study, the β-cell spheroids had the greatest percentage of excellent Z' scores. In contrast, the same cells in 2D had only one response that scored in the excellent range, and that was in response to caffeine, while freshly dispersed human islets had Z' scores that never reached the excellent or ideal benchmark. While the 3D β-cell method did provide improvements over the other two testing models, there is still optimization that needs to be completed, as the Z' value for the 3D β-cell spheroids in response to glucose was below an acceptable range (0.11), and for glybenclamide, the value was just below an excellent score. (0.48). Small alterations in the 3D β-cell spheroid screen may further optimize of the protocol such as testing on smaller diameter spheroids and increasing the drug exposure time from 60 min. A high-throughput screening method using a cell culture line, rather than relying on human donor cells, has many advantages, if it reflects the potential in vivo outcome. We have shown with a robust glucose response and reference drugs, that β-cell spheroids tested in a 384-well format provide a valid phenotypic screen for new diabetes drugs. The cost of the 3D screening protocol is not significantly different from the same screen in 2D. In fact, the assays were run using standard protocols and reagents. Only the plate to produce the 3D spheroids was different from current industry standards. The procedure could ultimately help to reduce costs and the high attrition rate currently inherent in finding and developing new diabetes drugs (Kimlin, Kassis, & Virador, 2013). Acknowledgements This work was funded by the University of Kansas Medical Center, and Likarda, LLC. References Akash, M. S., Rehman, K., & Chen, S. (2013). An overview of valuable scientific models for diabetes mellitus. Current Diabetes Reviews, 9(4), 286–293. Albrecht, H., Zbinden, P., Rizzi, A., Villetti, G., Riccardi, B., Puccini, P., ... Imbimbo, B. P. (2004). High throughput screening of beta-amyloid secretion inhibitors using homogenous time-resolved fluorescence. Combinatorial Chemistry & High Throughput Screening, 7(8), 745–756. Al-Mahmood, H. A., el-Khatim, M. S., Gumaa, K. A., & Thulesius, O. (1986). The effect of calcium-blockers nicardipine, darodipine, pn-200-110 and nifedipine on insulin

J. Amin et al. / Journal of Pharmacological and Toxicological Methods 82 (2016) 83–89 release from isolated rat pancreatic islets. Acta Physiologica Scandinavica, 126(2), 295–298. Beck, A., Vinik, Y., Shatz-Azoulay, H., Isaac, R., Streim, S., Jona, G., ... Zick, Y. (2013). Otubain 2 is a novel promoter of beta cell survival as revealed by sirna high-throughput screens of human pancreatic islets. Diabetologia, 56(6), 1317–1326. Breslin, S., & O'Driscoll, L. (2013). Three-dimensional cell culture: The missing link in drug discovery. Drug Discovery Today, 18(5–6), 240–249. Burns, S. M., Vetere, A., Walpita, D., Dancik, V., Khodier, C., Perez, J., ... Altshuler, D. (2015). High-throughput luminescent reporter of insulin secretion for discovering regulators of pancreatic beta-cell function. Cell Metabolism, 21(1), 126–137. DiMasi, J. A., Grabowski, H. G., & Hansen, R. W. (2014, Nov. 18). Cost to develop and win marketing approval for a new drug is $2.6 billion. Retrieved from Tufts University (website:). Ferrannini, E., & Mari, A. (2014). B-cell function in type 2 diabetes. Metabolism, 63(10), 1217–1227. Garcia-Barrado, M. J., Jonas, J. C., Gilon, P., & Henquin, J. C. (1996). Sulphonylureas do not increase insulin secretion by a mechanism other than a rise in cytoplasmic Ca2+ in pancreatic beta-cells. European Journal of Pharmacology, 298(3), 279–286. Grimmsmann, T., & Rustenbeck, I. (1998). Direct effects of diazoxide on mitochondria in pancreatic b-cells an on isolated liver mitochondria. British Journal of Pharmacology, 123(5), 781–788. Gubler, H. (2006). Methods for statistical analysis, quality assurance and management of primary hts data. Hihg-throughput screening in drug discovery, Vol. 1. (pp. 151–207). Weinheim, Germany: Wiley-VCh. Hectors, T. L. M., Vanparys, C., Pereira-Fernandes, A., Martens, G. A., & Blust, R. (2013). Evaluation of the ins-1 832/13 cell line as a beta-cell based screening system to assess pollutant effects on beta-cell function. PLoS One, 8(3), e60030. Hohmeier, H. E., & Newgard, C. B. (2004). Cell lines derived from pancreatic islets. Molecular and Cellular Endocrinology, 228(1–2), 121–128. Hohmeier, H. E., BeltrandelRio, H., Clark, S. A., Henkel-Rieger, R., Normington, K., & Newgard, C. B. (1997). Regulation of insulin secretion from novel engineered insulinoma cell lines. Diabetes, 46(6), 968–977. Hohmeier, H. E., Mulder, H., Chen, G., Henkel-Rieger, R., Prentki, M., & Newgard, C. B. (2000). Isolation of ins-1-derived cell lines with robust atp-sensitive k + channel-dependent and -independent glucose-stimulated insulin secretion. Diabetes, 49(3), 424–430. Howes, A. L., Richardson, R. D., Finlay, D., & Vuori, K. (2014). 3-dimensional culture systems for anti-cancer compound profiling and high-throughput screening reveal increases in egfr inhibitor-mediated cytotoxicity compared to monolayer culture systems. PLoS One, 9(9), e108283. Iversen, P. W., Benoit, B., Chen, Y. F., Dere, W., Devanarayan, V., Eastwood, B. J., ... Sittampalam, G. S. (2004). HTS assay validation. In E. L. C. a. t. N. C. f. A. T. Sciences (Ed.), Assay guidance manual [internet] (pp. 1–31) (Bethesda, MD: http://www.ncbi. nlm.nih.gov/books/NBK53196). Jijakli, H., Nadi, A. B., Cook, L., Best, L., Sener, A., & Malaisse, W. J. (1996). Insulinotropic action of methyl pyruvate: Enzymatic and metabolic aspects. Archives of Biochemistry and Biophysics, 335(2), 245–257. Kaniuk, N. A., Kiraly, M., Bates, H., Veranic, M., Volchuk, A., & Brumell, J. H. (2007). Ubiquitinated-protein aggregates form in pancreatic beta-cells during diabetes-induced oxidative stress and are regulated by autophagy. Diabetes, 56(4), 930–939. Kimlin, L., Kassis, J., & Virador, V. (2013). 3d in vitro tissue models and their potential for drug screening. Expert Opinion on Drug Discovery, 8(12), 1455–1466. Law, C. J., Ashcroft, H. A., Zheng, W., & Sexton, J. Z. (2014). Assay development and multivariate scoring for high-content discovery of chemoprotectants of endoplasmic-reticulum-stress-mediated amylin-induced cytotoxicity in pancreatic beta cells. Assay and Drug Development Technologies, 12(7), 375–384. Leong, D. T., & Ng, K. W. (2014). Probing the relevance of 3D cancer models in nanomedicine research. Advanced Drug Delivery Reviews, 15(79C-80C), 95–106. Li, Q. Y., Xu, M. K., Liu, G., Christoffersen, C. T., & Wang, M. W. (2013). Discovery of novel pde10 inhibitors by a robust homogeneous screening assay. Acta Pharmacologica Sinica, 34(8), 1116–1120.

89

Loessner, D., Holzapfel, B. M., & Clements, J. A. (2014). Engineered microenvironments provide new insights into ovarian and prostate cancer progression and drug responses. Advanced Drug Delivery Reviews, 15(79C-80C), 193–213. Lorenz, M. A., El Azzouny, M. A., Kennedy, R. T., & Burant, C. F. (2013). Metabolome response to glucose in the b-cell line INS-1 832/13. The Journal of Biological Chemistry, 288(15), 10923–10935. Lupi, R., & Del Prato, S. (2008). Beta-cell apoptosis in type 2 diabetes: Quantitative and functional consequences. Diabetes & Metabolism, 34(Suppl. 2), S56–S64. Pento, J. T., Kagan, A., & Glick, S. M. (1974). Influence of altered states of calcium homeostasis on insulin secretion in rats and rabbits. Hormone and Metabolic Research, 6(3), 177–180. Qureshi, R. M., Dejene, E. A., Corbin, K. L., & Nunemaker, C. S. (2015). Stress-induced dissociations between intracellular calcium signaling and insulin secretion in pancreatic islets. Cell Calcium, 75(0), 368–375. Ramachandran, K., Huang, H. H., & Stehno-Bittel, L. (2014). A simple method to replace islet equivalents for volume quantification of human islets. Cell Transplantation, 24(7), 1183–1194. Ramachandran, K., Williams, S. J., Huang, H. H., Novikova, L., & Stehno-Bittel, L. (2013). Engineering islets for improved performance by optimized reaggregation in a micromold. Tissue Engineering Parts A, 19(5–6), 18–26. Rupnik, M. (2009). The physiology of rodent beta-cells in pancreas slices. Acta Physiologica, 195, 123–138. Sakata, N., Egawa, S., Sumi, S., & Unno, M. (2008). Optimization of glucose level to determine the stimulation index of isolated rat islets. Pancreas, 36(4), 417–423. Song, H. H., Park, K., & Gerecht, S. (2014). Hydrogels to model 3d in vitro microenvironment of tumor vascularization. Advanced Drug Delivery Reviews(79C-80C), 19–29. Suzuki, H., Hirata, Y., Suzuki, N., Ihara, S., Sakitani, K., Kobayashi, Y., ... Koike, K. (2014). Characterization of a new small bowel adenocarcinoma cell line and screening of anti-cancer drug against small bowel adenocarcinoma. The American Journal of Pathology, S0002-9440(14) (00598-00597). Tammela, P., Alvesalo, J., Riihimäki, L., Airenne, S., Leinonen, M., Hurskainen, P., ... Vuorela, P. (2004). Development and validation of a time-resolved fluorometric immunoassay for screening of antichlamydial activity using a genus-specific europium-conjugated antibody. Analytical Biochemistry, 333(1), 39–48. Tran, K., Li, Y., Duan, H., Arora, D., Lim, H. Y., & Wang, W. (2014). Identification of small molecules that protect pancreatic β cells against endoplasmic reticulum stress-induced cell death. ACS Chemical Biology, 9(12), 2796–2806. Wajchenberg, B. L. (2007). Beta-cell failure in diabetes and preservation by clinical treatment. Endocrine Reviews, 28(2), 187–218. Walpita, D., & Wagner, B. K. (2014). Evaluation of compounds in primary human islet cell culture. Current Protocols in Chemical Biology, 6(3), 157–168. Walpita, D., Hasaka, T., Spoonamore, J., Vetere, A., Takane, K. K., Formina-Yadin, D., ... Wagner, B. K. (2012). A human islet cell culture system for high-throughput screening. Journal of Biomolecular Screening, 17(4), 509–518. Watkins, R. A., Evans-Molina, C., Blum, J. S., & DiMeglio, L. A. (2014). Established and emerging biomarkers for the prediction of type 1 diabetes: A systematic review. Translational Research, 164(2), 110–121. Wikstrom, J. D., Sereda, S. B., Stiles, L., Elorza, A., Allister, E. M., Neilson, A., ... Shirihai, O. S. (2012). A novel high-throughput assay for islet respiration reveals uncoupling of rodent and human islets. PLoS One, 7(5), e33023. Xi, B., Yu, N., Wang, X., Xu, X., & Abassi, Y. A. (2008). The application of cell-based labelfree technology in drug discovery. Biotechnology Journal, 3(4), 484–495. Zhang, J. H., Chung, T. D., & Oldenburg, K. R. (1999). A simple statistical parameter for use in evaluation and validation of high throughput screening assays. Journal of Biomolecular Screening, 4(2), 67–73. Zhang, Q., Ramrecheya, R., Lahmann, C., Tarasov, A., Bengtsson, M., Braha, O., ... Rorsman, P. (2013). Role of KATP channels in glucose-regulated glucagon secretion and impaired counterregulation in type 2 diabetes. Cell Metabolism, 18(6), 871–882. Zuck, P., Murray, E. M., Stec, E., Grobler, J. A., Simon, A. J., Strulovici, B., ... Ferrer, M. (2004). A cell-based beta-lactamase reporter gene assay for the identification of inhibitors of hepatitis c virus replication. Analytical Biochemistry, 334(2), 344–355.