High-throughput screening and bioinformatic analysis to ascertain compounds that prevent saturated fatty acid-induced β-cell apoptosis

Abstract

Pancreatic β-cell lipotoxicity is a central feature of the pathogenesis of type 2 diabetes. To study the mechanism by which fatty acids cause β-cell death and develop novel approaches to prevent it, a high-throughput screen on the β-cell line INS1 was carried out. The cells were exposed to palmitate to induce cell death and compounds that reversed palmitate-induced cytotoxicity were ascertained. Hits from the screen were analyzed by an increasingly more stringent testing funnel, ending with studies on primary human islets treated with palmitate. MAP4K4 inhibitors, which were not part of the screening libraries but were ascertained by a bioinformatics analysis, and the endocannabinoid anandamide were effective at inhibiting palmitate-induced apoptosis in INS1 cells as well as primary rat and human islets. These targets could serve as the starting point for the development of therapeutics for type 2 diabetes.

1. Introduction

A fundamental aspect of the current paradigm of type 2 diabetes pathogenesis is that fatty acids have deleterious effects on pancreatic β-cell function and survival. This has been demonstrated in numerous studies, both in vitro and in vivo, in which β-cells exposed to fatty acids exhibit defects in glucose-responsive insulin secretion and ultimately die, generally by apoptosis. Thus, there has been great interest in approaches to inhibiting deleterious effects on β-cells induced by fatty acids. Numerous compounds and genes that are involved in many distinct signaling pathways have been reported to protect β-cells from the effects of fatty acids [1–6], but those have been discovered by candidate approaches based on a priori knowledge and predictions about the mechanisms by which fatty acids affect β-cells. By assaying large numbers of compounds, an unbiased approach using high-throughput screening (HTS) has the potential to uncover previously unknown compounds and pathways that affect β-cell function and survival.

Here, we report the development of an assay for fatty acid-induced β-cell death and its use to screen a library of known drugs for those that inhibit β-cell death induced by fatty acids. The assay is based on INS1 cells, a commonly used cell line derived from a radiation induced, transplantable insulinoma that developed in inbred New England Deaconess Hospital albino rats [7,8]. This cell line is susceptible to apoptosis induced by palmitate and has been used to study pathways that contribute to that process [9].

To determine the identity of the targets of the positive compounds that inhibited lipotoxicity, we used an additional bioinformatics approach called the Kinase Inhibitors Elastic Net (KIEN) method [10]. Hits from the screen and the KIEN method analysis were tested in an increasingly more stringent testing funnel, ending with studies on primary rat and human islets treated with palmitate.

2. Materials and methods

2.1. Cell culture

INS1 832/13 [11] and INS1E cells [12] were cultured in DMEM (11 mM glucose, Corning, Manassas, VA, USA, with 10% fetal bovine serum (FBS), Sigma-Aldrich, St. Louis, MO, USA) and RPMI (11 mM glucose, Gibco, 5% FBS), respectively, and T6PNE cells were maintained in RPMI (5.5 mM glucose, Corning, 10% FBS). Cell lines were cultured with 1% penicillin/streptomycin (pen-strep, Gibco, Waltham, MA, USA) in 5% CO₂ at 37 °C.

2.2. Preparation of palmitate-BSA complex

Palmitate (150 mM) (Sigma-Aldrich) was prepared in 50% ethanol and precomplexed with 15% fatty acid-free BSA (Research Organics, Cleveland, OH, USA) in a 37 °C water shaker (PMID 25967901). Finally, BSA-precomplexed palmitate was used as a 12 mM stock solution for all assays. For some experiments, palmitate (50 mM) was dissolved in 90% ethanol and diluted to a final concentration of 0.5 mM in medium containing 0.67 to 0.75% fatty acid-free BSA [13].

2.3. Compound library screening with INS1 832/13 cells

To determine assay conditions, a dose response curve was generated with INS1 832/13 cells, ranging from 2500 to 24,000 cells per well of a 384 well plate in the presence and absence of 0.75 mM palmitate. The best signal to noise ratio was obtained with 10,000 cells per well. For HTS, cells were seeded at 10,000 cells per well in 384-well tissue culture plates (Greiner Bio-One, Monroe, NC, USA) in the presence of 25 mM glucose, 1% FBS and 0.75 mM palmitate and fatty acid-free BSA. 2025 Compounds (active compound in DMSO) or DMSO were added at a final concentration of 10 μM with the Echo (Labcyte, Sunnyvale, CA, USA). Each plate contained palmitate as negative control and BSA as positive one with DMSO. Two days after addition of compounds, ATP levels were measured. HTS was performed in the SBP Conrad Prebys Center for Chemical Genomics (CPCCG). Selected hits were purchased for secondary assays (Table 3).

Table 3.

Compounds for confirmatory assays.

Chemical name	Supplier	Location
HSP90 Inhibitor, CCT018159	MILLIPORE	Overijse; Belgium
PIM1/2 Kinase Inhibitor VI	MILLIPORE	Overijse; Belgium
Syk Inhibitor	MILLIPORE	Overijse; Belgium
PIKfyve Inhibitor	MILLIPORE	Overijse; Belgium
JNK Inhibitor V	MILLIPORE	Overijse; Belgium
Anandamide (20:3,n-6)	Sigma-Aldrich	Overijse; Belgium
AS605240 PI 3-Kg Inhibitor	Sigma-Aldrich	Overijse; Belgium
IRAK-1/4 Inhibitor I	Sigma-Aldrich	Overijse; Belgium
Lynestrenol	Sigma-Aldrich	Overijse; Belgium
Indirubin-3-oxime	Sigma-Aldrich	St. Louis, MO, USA
MAP4K4 inhibitor	SBP CPCCG	La Jolla, CA, USA
PIK3CB inhibitor	Cayman Chemical	Michigan, USA

2.4. Hit confirmation with siRNA and shRNA

Candidates from a list based on assay hits and bioinformatic analysis were studied using siRNA if expressed in the human islet cell line T6PNE [14]. Each mRNA was targeted by a mixture of 4 separate siRNAs that were pooled. Pooled siRNAs were manufactured by Ambion (Waltham, MA, USA). Transfection of T6PNE cells was done by mixing 2 μL of each siRNA (1 μM stock) and 20 μL of diluted Lipofectamine RNAi MAX (Invitrogen, Waltham, MA, USA) (1:100 in Opti-MEM) in a well of a 96-well plate (Thermo Fisher Scientific, Waltham, MA, USA), followed by incubation for 30 min at room temperature. T6PNE cells (1000 cells per well) diluted in 80 μL of RPMI supplemented with 10% FBS and 1% pen-strep were added to the transfection mix and incubated for 2 days at 37 °C, 5% CO₂. Transfected cells were incubated at 37 °C for an additional 2 days with tamoxifen, 0.25 mM palmitate or BSA followed by measurement of ATP level. For siRNA validation, cells were harvested for RNA purification and QPCR 2 days after transfection with target gene or scrambled siRNA. shRNAs were from (Sigma-Aldrich).

2.5. qPCR

RNA was purified using RNeasy Kits (Qiagen, MD, USA), then converted to cDNA using the qScript cDNA SuperMix (Quanta BioSciences, Beverly, MA, USA). qPCR was conducted on cDNA corresponding to 2 μg of RNA using an ABI 7900HT (Applied Biosystems, Foster City, CA, USA) and qPCR SuperMix (BioPioneer, San Diego, CA, USA). All mRNA values were normalized to 18S rRNA values and are expressed as fold change over vehicle treated control.

2.6. MAP4K4 inhibitor synthesis and testing

Compound 1 (6-(2-Fluoropyridin-4-yl)pyrido[3,2-d]pyrimidin-4-amine) was synthesized following the literature procedure (Compound 29 in Ref. 33) to >95% purity by LC-MS and ¹H NMR.

2.7. XTT assay

XTT assay reagent (Roche) incubated for 3.5 h in 96well plate as per the manufacturer’s directions and measured with ELISA reader (OPTI MAX, Molecular Device).

2.8. ATP Lite Assay

ATP level was measured using the ATPlite 1step Luminescence Assay System (PerkinElmer, Waltham, MA, USA) as per the manufacturer’s directions.

2.9. Assessment of apoptosis by Hoechst 33342 and propidium iodide assay

Apoptotic cell death was detected by fluorescence microscopy after staining with DNA binding dyes Hoechst 33342 (5 μg/ml; Sigma) and propidium iodide (5 μg/ml, Sigma) [15]. The percentage of viable, apoptotic and necrotic cells was determined after staining with the DNA-binding dyes Propidium Iodide (PI, 5 μg/ml, Sigma) and Hoechst 33342 (HO, 5 μg/ml, Sigma) [16]. This method is quantitative, and has been validated by systematic comparison against electron microscopy [17] and several other well-characterized methods, including fluorometric caspase 3 & 7 assays and determination of histone-complexed DNA fragments by ELISA [16,18–20]. A minimum of 500 cells were counted in each experimental condition. Viability was evaluated by two independent observers, one of them being unaware of sample identity. The agreement between findings obtained by the two observers was >90%.

2.10. Rat and human islets

Pancreatic islets were isolated from male Wistar rats (Charles River Laboratories, Wilmington, MA, USA), housed and handled following the rules of the Belgian Regulations for Animal Care and with approval of the ULB Ethics Committee, and cultured in Ham’s F-10 medium (Invitrogen) containing 10 mM glucose, no FBS and 0.75% fatty acid-free BSA. These culture conditions are based on earlier studies showing optimal function and survival of rat islet cells in vitro at 10 mM glucose [21].

Human islets were isolated from 7 organ donors (age 76 ± 3 years; 4 women/3 men; body mass index 27 ± 1 kg/m²; cause of death cerebral hemorrhage) in Pisa, Italy, with the approval of the Ethics Committee. The human islets were cultured in Ham’s F-10 medium containing 6.1 mM glucose, 2 mM GlutaMAX, 50 μM 3-isobutyl-1-methylxanthine, 1% charcoal absorbed BSA, 50 units/ml penicillin, and 50 μg/ml streptomycin [13]. Studies with human islets were performed at 6.1 mM glucose, as we have previously shown that glucose concentrations in the range of 5.6–6.1 are ideal to preserve human islet function in culture [22]. In contrast with clonal β-cell lines, lipotoxicity is not potentiated by increasing glucose concentrations in human islets [15]; hence these experiments were performed at 6.1 mM glucose. The percentage of β-cells in the human islet preparations, assessed following staining with anti-insulin antibody [23], was 63 ± 3%.

2.11. Statistical analysis

Data are presented as means ± SEM of n independent experiments. Comparisons were performed by two-sided paired t-test. p < 0.05 was considered statistically significant.

3. Results

3.1. Development of an assay for palmitate induced killing of INS1 cells

To develop an assay for fatty acid induced lipotoxicity, we first constructed a dose response curve for palmitate-induced cytotoxicity (Fig. 1). Dose-responsive loss of cells was observable in the wells (Fig. 1A) and that was confirmed by XTT assay (Fig. 1B). For HTS, we used ATP level as determined by ATPLite luciferase activity, which is well suited for screening and has been shown to be an accurate measure of cellular viability, and thus should reflect a decrease in cell number caused by apoptosis, although it does not directly measure apoptosis (Fig. 1C) [24]. The optimal assay condition was determined to be 0.75 mM palmitate in 1% FBS at 25 mM glucose. Consistent with the findings of others, substantially elevated glucose concentration was required for maximal induction of palmitate-induced apoptosis [25]. Under those conditions, the Z’ [26] for the assay was 0.70, well above the value considered appropriate for conducting HTS.

Fig. 1.

Development of the lipotoxicity assay. INS1 832/13 cells were incubated with the indicated concentration of palmitate for 2 days with 1% FBS and XTT assay was used to measure cell viability. (A) Light microscopic image of treated cells. (B) XTT assay in 11 mM and 25 mM glucose (N = 6). (C) ATPLite assay in 25 mM glucose (N = 16). Scale bar = 100um.

3.2. High-throughput screen for compounds that prevent palmitate-induced INS1 cell death

Four compound libraries comprising 2025 compounds (Table 1) were screened in 384 well plates. Hits were defined as those that had ATPLite activity greater than 2-fold above the mean of the negative controls on each plate (Fig. 2A). For initial confirmation of the hits, 40 positive compounds (Table 2) were cherry-picked from the screening libraries and were rescreened in quadruplicate (Fig. 2B). 35 of 40 hits replicated when rescreened using the primary assay based on ATPLite (Fig. 2B).

Table 1.

Libraries screened.

Library	Vendor	# of compounds
Kinase Inhibitors	EMD (Massachusetts, USA)	327
StemSelect	EMD (Massachusetts, USA)	303
Bioactive lipid collection	Enzo Life Science (New York, USA)	195
Prestwick chem library	Prestwick (San Diego, USA)	1200

Fig. 2.

High-throughput screen. 2025 compounds from the libraries listed in Table 1 were incubated for 2 days with INS1 cells, followed by determination of ATP level by ATPLite assay. (A) Primary assay. Each dot represents the data from a single compound. The positive control was cells treated with fatty acid free BSA (0.9%) (yellow dots, N = 112). The negative control was cells treated with 0.75 mM palmitate plus DMSO (red dots, N = 112). For the experimental group, cells were treated with 0.75 mM palmitate plus compound (10uM) for 2 days, after which the ATP level was measured. (B) Primary confirmatory assay. 40 hits from A were tested in quadruplicate using the primary assay. The red line is the mean of the negative control (red dot) and the blue line is the mean of the positive control (yellow dot). Green dots represent the mean of the value from each compound (0.75 mM palmitate + 10uM of each compound, N = 4). Error bars are standard deviation.

Table 2.

Confirmed hits. Primary assay hits that were confirmed as positive in a rescreen of the primary assay performed in quadruplicate.

	Name	Compound Name	Putative Target	CAS#
1	EMD StemSelect	HSP90 Inhibitor,	HSP90	171009-07-7
2	EMD StemSelect	GSK-3 Inhibitor IX	GSK-3	667463-62-9
3	Prestwick 2015	Pioglitazone	PPARg	111025-46-8
4	Prestwick 2015	Clofibrate	PPARa	637-07-0
5	Prestwick 2015	Lynestrenol	Progesterone receptor	52-76-6
6	Prestwick 2015	Pirenzepine dihydrochloride	M1 muscarinic acetylcholine rec.	29868-97-1
7	Prestwick 2015	Nisoldipine	1,4-dihydropyridine Ca++ channel	63675-72-9
8	Prestwick 2015	2-Aminobenzenesulfonamide	Carbonic anhydrase	3306-62-5
9	Prestwick 2015	Rifapentine	Bacterial RNA polymerase	61379-65-5
10	Prestwick Drugs	Talampicillin hydrochloride	DD-transpeptidase	39878-70-1
11	Bioactive Lipids	Anandamide (20:4, n-6)	Cannabinoid receptor	94421-68-8
12	Bioactive Lipids	Anandamide (20:3,n-6)	Cannabinoid receptor
13	Bioactive Lipids	Anandamide (22:4,n-6)	Cannabinoid receptor
14	Bioactive Lipids	Mead ethanolamide	Cannabinoid receptor	169232-04-6
15	Bioactive Lipids	(R)-Methanandamide	Cannabinoid receptor	157182-49-5
16	Bioactive Lipids	Arachidonamide	Cannabinoid receptor	85146-53-8
17	EMD Kinase Inhibitor	Alsterpaullone	CDK1, CDK5, Cyclin B	237430-03-4
18	EMD Kinase Inhibitor	Aloisine, RP106	GSK3b	496864-15-4
19	EMD Kinase Inhibitor	Kenpaullone	CDK1, CDK2, CDK5, GSK3b	142273-20-9
20	EMD Kinase Inhibitor	Syk Inhibitor	Syk	622387-85-3
21	EMD Kinase Inhibitor	Indirubin-3′-oxime	GSK3	160807-49-8
22	EMD Kinase Inhibitor	GSK-3 Inhibitor XIII	GSK3	404828-08-6
23	EMD Kinase Inhibitor	Indirubin Derivative E804	Src, Cdc2, Cdk2	854171-35-0
24	EMD Kinase Inhibitor	JNK Inhibitor V	JNK	345987-15-7
25	EMD Kinase Inhibitor	GSK-3b Inhibitor VIII	GSK3b	487021-52-3
26	EMD Kinase Inhibitor	GSK-3 Inhibitor IX	GSK3	667463-62-9
27	EMD Kinase Inhibitor	GSK-3 Inhibitor X	GSK3	740841-15-0
28	EMD Kinase Inhibitor	GSK-3b Inhibitor XI	GSK3b	626604-39-5
29	EMD Kinase Inhibitor	JAK Inhibitor I	JAK	457081-03-7
30	EMD Kinase Inhibitor	Tyrphostin AG 879	TrkA, HER-2	148741-30-4
31	EMD Kinase Inhibitor	YM-201636	PIKfyve	371942-69-7
32	EMD Kinase Inhibitor	AS605240	PI3-Kg	648450-29-7
33	EMD Kinase Inhibitor	SMI-4a	PIM1/2	327033-36-3
34	EMD Kinase Inhibitor	PIM1/2 Kinase Inhibitor VI	PIM1/2	587852-28-6
35	EMD Kinase Inhibitor	Chk2 Inhibitor	Chk2	724708-21-8
36	EMD Kinase Inhibitor	Indirubin-3′-monoxime, 5-Iodo-	GSK3b, CDK1, CDK5	331467-03-9
37	EMD Kinase Inhibitor	IRAK-1/4 Inhibitor I	IRAK-1/4	509093-47-4
38	EMD Kinase Inhibitor	Syk Inhibitor IV, BAY 61-3606	Syk	732938-37-8
39	EMD Kinase Inhibitor	SU9516	CDK2	666837-93-0
40	EMD Kinase Inhibitor	1-Azakenpaullone	GSK3b	676596-65-9

The compounds from the Prestwick and StemSelect libraries did not demonstrate any obvious pattern, with diverse structures and mechanisms of action being represented among the hits. The Bioactive Lipids library was remarkable in that there was a cluster of endocannabinoid hits. The majority of the hits in the screen came from the kinase inhibitor library.

3.3. Secondary assays

Ten interesting compounds (Fig. 3A) from the 35 that passed primary confirmation were selected for secondary assays in INS1E cells based on the presence of at least one of the following criteria: 1. Putative targets of the compound were expressed in FACS-purified primary rat β-cells and human islets using RPKM >0.5 as a cutoff [27,28]; 2. Targets were involved in pathways of known or potential biological relevance to β-cell survival or function. Because the ATPLite assay does not directly measure apoptosis, it was important to use a more specific secondary assay. Thus, selected compounds were tested in an assay in which apoptosis was specifically assessed by staining with Hoechst 33342 and propidium iodide as done previously [29]. Seven of the ten compounds tested inhibited apoptosis in that assay (Fig. 3A). The seven most active were selected for further study. They were examined for dose-responsive inhibition of apoptosis, with the finding that all exhibited dose-responsive inhibition of apoptosis (Fig. 3B). Substantial inhibition of apoptosis was achieved in some cases -up to 3.5-fold- but in no case was the inhibition complete, i.e., providing complete protection against palmitate-induced cell death.

Fig. 3.

Effect of selected HTS compounds on palmitate-induced beta-cell death. Apoptosis in INS-1E cells was measured using the Hoechst 33342 and propidium iodide assay (A) Cells were cultured for 1 day in medium containing 11 mM glucose, 1% FBS, 0.75% fatty acid-free BSA with 0.5 mM palmitate and 5 μM of compounds (N = 3). Black bar is BSA only and white bar is palmitate treated, with or without the addition of putative protective compounds. (B) Dose-response of protective compounds. The vehicle was DMSO. *p < 0.05 against palmitate-treated cells.

3.4. Bioinformatics identify the candidate targets of hits from the Kinase Inhibitor library

Examining the putative targets of the reproducibly positive compounds revealed that in virtually every case in which multiple compounds from the libraries that were screened were directed against the same putative target, only a subset of the compounds against a particular target were positive. This was particularly notable for the kinase inhibitor hits. For example, only 10 of 27 screened compounds that inhibited GSK3 were hits. This suggested that the activity of the compounds may have been due to activity on targets other than the putative target of the compounds.

To determine the identity of the targets of the positive compounds that were responsible for inhibition of lipotoxicity, we used a bioinformatics approach called the Kinase Inhibitors Elastic Net (KIEN) method [10]. This method combines data from a screen with data on off-target effects of kinase inhibitors to identify candidates for the actual kinase that was inhibited by screening hits. Specifically, the method integrates information contained in drug-kinase profiling, namely the residual catalytic activity A_k,i of the kinase k after interacting with compound i [30], with the cell viability v_i after in vitro treatment with the same compound. The cell response to drugs from the screen was used to build a linear regression model as v_i = β₀ + β₁A_1,i + β₂A_2,i + ⋯ + β_pA_p,i where p = 291 was the total number of kinases, i = 1, …, 139 identified the compound screened, and the residual catalytic activities A_k,i were treated as predictors of the cell responses v_i. Thus, two different types of data were integrated: kinase profiling data obtained through enzymatic assays that probe directly the interaction between drug and kinases, and in vitro cell response data that is the result of complex signaling that involves many pathways downstream of the affected kinases.

To keep the set of kinases to a minimum, we used the elastic net regression method, which modifies the least square minimization by imposing a penalty on the regression coefficients [31]. For each kinase k, the regression provided a coefficient score β_k that was interpreted as a measure of the sensitivity of cells to alterations in the activity of that kinase. The elastic net screening selected 247 kinases as predictors, forcing the remaining 44 to have β_k = 0. A negative beta coefficient means that inhibition of that kinase is associated with increased cell survival in the experiment, which is the effect we are interested in. The elastic net algorithm is designed to find the optimal balance between fitting the data and using the smallest set of kinases as variables. The regression had an R² = 0.57 (Fig. 4A). The top 10 relevant kinases according to the elastic net method are shown in Fig. 4B.

Fig. 4.

Bioinformatic analysis of kinase inhibitor screen. (A) Comparison between the cell viability ${\widehat{v}}_i$ obtained using the elastic net regression and the actual values v_i measured in the experiment. The coefficient of determination R² = 0.57 is the square of the Pearson correlation between the ${\widehat{v}}_i$ and the v_i. (B) Top 10 relevant kinases determined by the elastic net method.

3.5. siRNA as an orthogonal approach to target validation

Given that small molecule kinase inhibitors are promiscuous, it was important to use an independent assay to verify target identity. For that purpose, we employed siRNAs as an orthogonal approach to determine whether the putative targets of the small molecules were valid. While siRNAs against a specific target mRNA are known to have a significant number of off-target effects [32], those effects are due to an entirely different mechanism than those of small molecule kinase inhibitors. Thus, if both the small molecule and an siRNA against the same putative target are active, this provides a high degree of assurance that the target is valid.

A subset of the putative targets of the small molecule inhibitors, as well as targets identified by the bioinformatics approach, were inhibited using siRNAs. Each siRNA was validated by qPCR (Fig. 5A) Of the targets studied, only siRNA-mediated inhibition of MAP4K4 and PIK3CB repressed palmitate induced cell death in the ATPLite assay (Fig. 5B). To increase the level of confidence further, we employed lentivirally-encoded shRNAs, which confirmed the data from the siRNA studies (Fig. 6).

Fig. 5.

siRNA to target validation. (A) siRNAs were validated by qPCR. T6PNE cells transfected with target gene or scrambled siRNA were harvested for RNA after 2 days (N = 5 technical replicates). (B) Candidate target gene siRNAs were transfected into human T6PNE cells for 2 days followed by treatment with BSA or palmitate (0.25 mM) for 2 days (N = 4). ATP activity of palmitate treated siRNA is normalized to that of the same siRNA treated with BSA. *p < 0.005, **p < 0.001against scrambled siRNA control. Error bars are standard error.

Fig. 6.

shRNA-mediated inhibition of MAP4K4 and PI3CB. (A) T6PNE cells were infected with lentiviral vectors encoding two different shRNAs to MAP4K4 and two shRNAs to PIK3CB. Gene expression was measured by qPCR to determine effects of the shRNAs on MAP4K4 and PIK3CB mRNA levels. (B) Clones of cells expressing the indicated shRNA were treated with BSA or palmitate (0.25 mM) for 2 days, followed by measurement of ATP activity. The ATP activity for each experiment with an shRNA clone was normalized to the same clone treated with BSA alone (N = 4). *p < 0.05, *p < 0.0001 against control. Error bars are standard error.

3.6. Small molecule inhibitor of MAP4K4

MAP4K4 was identified as a potential mediator of palmitate-induced apoptosis through the bioinformatics approach, using data from the HTS. While the siRNA and shRNA studies indicated that this identification was accurate, we had not tested any small molecule MAP4K4 inhibitors for their ability to inhibit palmitate-induced apoptosis. Unfortunately, no commercially available compounds that exhibited selectivity for MAP4K4 were available. Therefore, we synthesized compound 1 (6-(2-Fluoropyridin-4-yl) pyrido[3,2-d]pyrimidin-4-amine) [33] which exhibits substantial selectively for MAP4K4 and excellent potency (IC₅₀ = 17nM). Compound 1 inhibited palmitate-induced apoptosis in INS-1E (Fig. 7A) and T6PNE [14] cells (Fig. 7B).

Fig. 7.

Effect of MAP4K4 inhibitor. (A) Apoptosis in INS-1E cells pre-treated for 2 h with 10 μM MAP4K4 inhibitor and treated for 1 day with 0.5 mM palmitate plus the MAP4K4 inhibitor (N = 5). (B) Dose response with MAP4K4 inhibitors in human T6PNE cells treated with 0.25 mM palmitate (N = 4). ATP activity measured after 48 h. *p < 0.05, **p < 0.001 against palmitate-treated cells. Error bars are standard error.

3.7. Effect of target inhibition by small molecules on palmitate-induced apoptosis in primary islets

Having demonstrated that inhibitors of MAP4K4 and PI3CB, as well as endocannabinoids, were cytoprotective in palmitate-mediated apoptosis in cell lines, we proceeded to extend our studies to primary islets. In rat islets, anandamide and the MAP4K4 inhibitor (compound 1) led to decreased apoptosis (Fig. 8A). Small molecule inhibition of PIK3CB actually caused increased apoptosis in rat islets even in the absence of palmitate (Fig. 8A). Thus, anandamide and the MAP4K4 inhibitor were tested in primary human islets, where both partially inhibited palmitate-induced apoptosis (Fig. 8B).

Fig. 8.

Apoptosis in primary islets. (A) Apoptosis in primary rat islets cultured for 3 days in medium containing 10 mM glucose, no FBS, 0.75% fatty acid-free BSA with 0.5 mM palmitate with DMSO or 5 μM anandamide, MAP4K4 inhibitor, PIK3CB inhibitor (N = 3). (B) Cell death in human islets cultured for 3 days in medium containing 6.1 mM glucose, 1% charcoal absorbed BSA with 0.5 mM palmitate and DMSO or 5 μM anandamide, MAP4K4 inhibitor (N = 3–4). The vehicle was DMSO. *p < 0.05 against palmitate-treated cells. Error bars are standard error.

4. Discussion

Despite much effort, our understanding of the mechanisms by which fatty acids induce pancreatic β-cell dysfunction and death is quite limited, with many pathways being implicated [27], but none having been definitively proven or developed into an effective therapy [34]. High-throughput phenotypic screening is attractive as an unbiased approach to discovering new targets that modulate biological processes. Here, we screened small molecule libraries containing fairly well-characterized compounds and identified a number that inhibited palmitate-induced cell death. It is important to note that the primary assay was based on a rat insulinoma cell line. Hits from the cell line assay were pursued in primary islets, including human, and most but not all, e.g., the PIK3CB inhibitor, confirmed.

Most of the hits from our screen came from a library of kinase inhibitors. Unfortunately for efforts to use small molecules as probes to study the function of specific molecular targets, many if not most small molecules act on multiple targets [35]. This is particularly true for kinase inhibitors, the vast majority of which interact with the ATP binding pocket [30,36]. To deal with the extreme promiscuity of kinase inhibitors and identify the true targets responsible for inhibition of lipotoxicity, we employed a bioinformatics approach. The KIEN method quantifies the role of multiple kinases in a biological process [10], and thus allowed us to find hidden targets because not all kinases are the principal targets of any of the tested inhibitors. In this study, we used it to identify MAP4K4 as an important single kinase in β-cell lipotoxicity. However, the method could also be extended to guide a combinatorial approach in which multiple kinases are inhibited by the amount suggested by the model. This could be approximated by an optimally designed combination of multiple kinase inhibitors under a model in which lipotoxicity is mediated by multiple signaling pathways that each involve distinct kinases.

Of the multiple kinases that were targeted by initial hits in the assay, MAP4K4 passed the multiple layers of confirmatory assays, including studies with human islets. MAP4K4 is widely expressed and plays a role in multiple biological processes, including inflammation [37], cellular proliferation [38], and metabolism [39]. In diabetes, MAP4K4 mRNA and protein expression was reported to be increased in islets from diabetic db/db mice [40]. With respect to human diabetes, it has been identified through GWAS studies as being associated with type 2 diabetes [41]. Studies on its role in cell death are contradictory, with reports of pro-apoptotic [42] as well as anti-apoptotic [38] effects in cancer cells. In β-cells MAP4K4 silencing attenuated detrimental effects of TNF-α on insulin secretion and signaling [43]. In our studies of islet cell lines and primary rat and human islets, inhibition of MAP4K4 with pharmacological inhibitors or by siRNA/shRNA led to increased survival, consistent with a pro-apoptotic role for MAP4K4 in β-cells.

Apart from the kinase inhibitors, the most striking result was that multiple cannabinoid receptor agonists were positive in the assay. Consistent with an on-target effect, anandamide (18:2,n-6), an inactive anandamide analog, was inactive. However, N-arachidonylglycine, an active anandamide analog that is resistant to cleavage by fatty acid amide hydrolase, was not a hit. This is in keeping with an earlier report on β-cell cytoprotection from lipotoxicity with oleoylethanolamide, which acted indirectly to attenuate palmitate toxicity through its internalization and hydrolysis by fatty acid amide hydrolase to release oleate [44], a known antagonist of palmitate toxicity [15,45]. Taken together, these results suggest that the true active compounds in the assay may be metabolites that act on targets other than cannabinoid receptors. This is supported by the fact that, while endocannabinoid receptors have been purported to be present in the pancreas [46], examination of recent RNA-Seq data finds that these receptors are not detectable or expressed at a very low level in primary human and rat islets or β-cells [27,28,47]. Thus, the primary target of the cannabinoid receptor agonists that is responsible for their ability to inhibit lipotoxicity remains unclear but there are many possible candidates, including free fatty acid receptors.

A drawback of this and all other studies of β-cell lipotoxicity is that none of the compounds led to complete inhibition of apoptosis. While many compounds and targets have been studied for effects on lipotoxicity, in no case has a single target been found to be sufficient to completely inhibit fatty acid-induced cell death [9,34]. The biological effects of fatty acids are complex, with multiple downstream pathways likely being responsible for their toxic effect on cells. Thus, no single intervention may be sufficient to completely inhibit the ability of fatty acids to induce cell death. Ultimately, protection of β-cells from lipotoxicity is likely to require a multi-faceted approach. The bioinformatics approach that we used here is well suited for the identification of candidate combinations, which would require additional testing in a high-throughput assay such as that used here.

Abbreviations:

MAP4K4

mitogen-activated protein kinase kinase kinase kinase 4

PIK3CB

phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit beta

KIEN

Kinase Inhibitors Elastic Net

GWAS

genome-wide association study