Abstract

Background

   Proprotein convertase subtilisin/kexin type 9 (PCSK9) is a secreted
   protein that down-regulates hepatic low-density lipoprotein receptor
   (LDLR) by binding and shuttling LDLR to lysosomes for degradation. The
   development of therapy that inhibits PCSK9 has attracted considerable
   attention for the management of cardiovascular disease risk. However,
   only monoclonal antibodies of PCSK9 have reached the clinic use. Oral
   administration of small-molecule transcriptional inhibitors has the
   potential to become a therapeutic option.

Methods

   Here, we developed a cell-based small molecule screening platform to
   identify transcriptional inhibitors of PCSK9. Through high-throughput
   screening and a series of evaluation, we found several active
   compounds. After detailed investigation on the pharmacological effect
   and molecular mechanistic characterization, 7030B-C5 was identified as
   a potential small-molecule PCSK9 inhibitor.

Findings

   Our data showed that 7030B-C5 down-regulated PCSK9 expression and
   increased the total cellular LDLR protein and its mediated LDL-C uptake
   by HepG2 cells. In both C57BL/6 J and ApoE KO mice, oral administration
   of 7030B-C5 reduced hepatic and plasma PCSK9 level and increased
   hepatic LDLR expression. Most importantly, 7030B-C5 inhibited lesions
   in en face aortas and aortic root in ApoE KO mice with a slight
   amelioration of lipid profiles. We further provide evidences suggesting
   that transcriptional regulation of PCSK9 by 7030B-C5 mostly depend on
   the transcriptional factor HNF1α and FoxO3. Furthermore, FoxO1 was
   found to play an important role in 7030B-C5 mediated integration of
   hepatic glucose and lipid metabolism.

Interpretation

   7030B-C5 with potential suppressive effect of PCSK9 expression may
   serve as a promising lead compound for drug development of
   cholesterol/glucose homeostasis and cardiovascular disease therapy.

Fund

   This work was supported by grants from the National Natural Science
   Foundation of China (81473214, 81402929, and 81621064), the Drug
   Innovation Major Project of China (2018ZX09711001-003-006,
   2018ZX09711001-007 and 2018ZX09735001-002), CAMS Innovation Fund for
   Medical Sciences (2016-I2M-2-002, 2016-I2M-1-011 and 2017-I2M-1-008),
   Beijing Natural Science Foundation (7162129).

   Keywords: PCSK9, Small-molecule inhibitor, FoxO1/3, HNF1α, Lipid and
   glucose metabolism, Atherosclerosis
     __________________________________________________________________

Research in context.

Evidence before this study

   To date, PCSK9 inhibitors are novel therapeutics for the treatment of
   cardiovascular disease in addition to statins. Two monoclonal
   antibodies alirocumab and evolocumab targeting PCSK9 have been approved
   by FDA, but they are expensive and requiring subcutaneous
   administration, limiting their widespread use. As small-molecule
   inhibitors targeting protein-protein interaction of PCSK9 and LDLR are
   difficult to develop, more options are needed.

Added value of this study

   Here, we have developed a novel cell-based screening platform to
   discover inhibitors targeting PCSK9 transcription. Using a robust and
   pragmatic approach, we identified an effective small-molecule PCSK9
   inhibitor 7030B-C5. Our data showed that 7030B-C5 down-regulated the
   expression of PCSK9 at both mRNA and protein levels and increased the
   cellular LDLR protein level and its mediated cellular LDL-C uptake. In
   both C57BL/6 J mice and ApoE KO mice, oral administration of 7030B-C5
   significantly reduced hepatic PCSK9 expression/secretion and increased
   LDLR expression. Moreover, the compound profoundly reduced
   atherosclerosis progression, and showed dual benefits in lipid and
   glucose metabolism in an HNF1α and FoxO1/3-responsive
   elements-dependent manner.

Implications of all the available evidence

   A small-molecule compound, 7030B-C5, with PCSK9 inhibitory and
   anti-atherosclerosis activity may potentially serve as a drug lead
   compound for cholesterol/glucose modulation and cardiovascular disease.

   Alt-text: Unlabelled box

1. Introduction

   Increases in hepatic low-density lipoprotein receptor (LDLR) play a
   pivotal role in enhancing the clearance of plasma low-density
   lipoprotein cholesterol (LDL-C) and are associated with reducing the
   risk of cardiovascular disorder [41][1]. Proprotein convertase
   subtilisin/kexin type 9 (PCSK9) is a critical player in LDL-C
   metabolism through regulating the degradation of hepatic LDLRs
   [[42]2,[43]3]. PCSK9 is a liver-derived plasma protease which is
   initially synthesized as a 75 kDa precursor protein and converted into
   a 62 kDa mature form undergoing the autocatalytic cleavage process in
   the Golgi apparatus [[44]4,[45]5]. Circulating PCSK9 binds to the
   extracellular epidermal growth factor-like repeat A (EGF-A) domain of
   LDLR, causing its internalization [[46]2,[47]6]. Followed, the
   PCSK9/LDLR complex translocates to the endosome-lysosomal compartment
   where LDLR is degraded [[48]7,[49]8]. Reduced cell-surface expression
   of LDLR leads to increased circulating levels of LDL-C. In human
   subjects, early studies have shown that gain-of-function mutations in
   the PCSK9 gene lead to familial hypercholesterolemia [50][9], while
   loss-of-function mutations are associated with hypocholesterolemia and
   protection against coronary artery disease [51][10], [52][11],
   [53][12]. Thus, decreasing circulating PCSK9 level or activity to
   up-regulate hepatic LDLR level and to lower circulating LDL-C level
   will be beneficial for reducing the risk of cardiovascular disease in
   humans.

   The pivotal role of PCSK9 in the metabolism of LDL-C as well as the
   verified safety of PCSK9 inhibition led to the development of PCSK9
   inhibitors. Recently, two PCSK9 monoclonal antibodies (mAbs) alirocumab
   and evolocumab which were reported to disrupt the PCSK9-LDLR
   interaction have been approved by the U.S. Food and Drug Administration
   (FDA) [54][13]. Alirocumab and evolocumab can be potentially used as
   monotherapy or add-on therapies to statins in patients with familial
   hypercholesterolemia or atherosclerotic cardiovascular disease,
   effectively reducing LDL-C up to 70% and with well-tolerated profile
   [55][14], [56][15], [57][16], [58][17]. Besides, these monoclonal
   antibodies significantly suppress circulating PCSK9 levels and
   consistently reduce LDL-C levels in patients [59][18]. However, it is
   estimated that the cost of PCSK9 mAb inhibitors was ~$14,000 per person
   per year in the US which limits its widespread use [60][19]. The
   small-molecule inhibitors can offer potentially more convenient routes
   of administration and lower costs. However, small-molecule inhibitors
   specific to inhibit PCSK9 activity remains challenging because it is
   difficult for small molecules to blockade the fairly large size of the
   flat interface between the catalytic subunit of PCSK9 and the EGF-A
   domain of LDLR [61][20]. Other promising approaches have been explored
   to discover small molecules with potential to become a therapeutic
   option by inhibiting PCSK9 protein synthesis. PF-06446846 was
   identified by Petersen et al. through phenotypic screening of a
   small-molecule compound collection, which directly and selectively
   inhibits translation of PCSK9 by stalling the 80S ribosome in the
   proximity of codon region [[62]21,[63]22]. Orally-administered
   PF-06446846 reduces plasma PCSK9 and LDL-C levels without liver
   toxicity symptoms in vivo [64][23]. However, its development was
   discontinued for competitive reasons [65][24].

   The gene expression of PCSK9 is controlled by various transcriptional
   regulators such as sterol regulatory element binding protein (SREBP),
   hepatocyte nuclear factor 1α (HNF1α), and forkhead box O3 (FoxO3). A
   sterol regulatory element (SRE), which forms the target site for
   SREBPs, is present 330 bp upstream of the translation start site and
   functions as an important cis-element regulating PCSK9 transcription
   [66][25]. Cholesterol-lowering drugs, statins, which activate the
   SREBP2 pathway by inhibiting HMG-CoA reductase and subsequently
   activating the expression of LDLR, have also been proven to stimulate
   PCSK9 gene expression, thus diminish the beneficial effects of statin
   treatment [67][26]. Insulin and docosahexaenoic acid (DHA) could
   activate PCSK9 transcription via SREBP-1c [68][27], [69][28], [70][29].
   In addition to SREBP regulation, HNF1α, a liver-enriched transcription
   factor, was found to bind HNF1 element within the PCSK9 promoter and
   activate PCSK9 expression in hepatic cells [71][30] and livers
   [[72]31,[73]32]. HINFP was reported as a positive regulator of PCSK9
   transcription through facilitation of histone H4 acetylation at PCSK9
   promoter [74][33]. FoxO3, a forkhead transcription factor, has been
   identified as a negative regulator of PCSK9 gene expression through
   epigenetic modulation. FoxO3 interacts with the insulin-response
   element (IRE) within PCSK9 promoter, recruiting the sirtuin-6 (Sirt6)
   protein to deacetylate histones and reducing the promoter binding
   capacity of HNF1α, thereby suppressing the PCSK9 gene expression in
   hepatic cells [75][34]. In this work, we developed a strategy to
   discover small-molecule inhibitors of PCSK9 expression at transcription
   level, and described the identification and mechanistic
   characterization of 7030B-C5, an active small-molecule inhibitor of
   PCSK9 transcription. Oral administration of the compound significantly
   reduced atherosclerosis progression, and showed dual benefits in both
   lipid and glucose metabolism, providing a promising lead compound for
   further development of novel drug candidate for small-molecule PCSK9
   modulators.

2. Materials and methods

2.1. Cell cultures and cell viability assay

   HepG2 cells were cultured in Eagle minimal essential medium (MEM;
   Invitrogen, Carlsbad, CA, USA) supplemented with 10% (vol/vol) fetal
   bovine serum (FBS; Gibco, Grand Island, NY, USA), 1% sodium pyruvate
   and 1% nonessential amino acids (Invitrogen). HEK-293T cells were
   cultured in Dulbecco's Modified Eagle Medium (DMEM; Invitrogen)
   supplemented with 10% (vol/vol) FBS (Gibco). HepG2 cells were
   transfected with luciferase reporter constructs contained the PCSK9
   gene promoter sequence spanning the region from −2112 to −1 bp
   (nucleotide +1 corresponds to the A of ATG start codon, the same as
   below). The stable pGL4-PCSK9-P transfected HepG2 cells were
   established by selection with G418 (700 μg/mL, Invitrogen). The cells
   were cultured at 37 °C in a humidified 5% CO[2] incubator.

   Cell viability was determined using MTT assay. HepG2 cells were treated
   with vehicle or positive compounds (0.39–100 μM) for 4 h followed by
   incubation with MTT reagent (1 mg/mL) at 37 °C for 4 h. The medium was
   removed, and the purple formazan crystals were dissolved in DMSO. The
   absorbance was measured at 550 nm using a Victor X5 multilabel plate
   reader (PerkinElmer, Waltham, MA, USA).

2.2. Construction of luciferase reporter system and high-throughput screening

   Human PCSK9 gene promoter was amplified by PCR using HepG2 genomic DNA
   as the template and cloned into pGL4-Basic vector (Promega, Madison,
   WI, USA) at Xho I and Hind III sites to get pGL4-PCSK9-P. And
   pGL4-PCSK9-P plasmid was used as the template for the PCR amplification
   of PCSK9 gene promoter fragments (D1-D7). The HNF1α mutant (HNF1α-mu)
   and HINFP mutant (HINFP-mu) were generated from construct D5 using the
   Fast MultiSite Mutagenesis System (TransGen, Beijing, China). Primers
   used for plasmid construction are listed in Supplemental Table I.

   The cells were transfected with the human PCSK9 promoter plasmid using
   Lipofectamine 2000 (Thermo Scientific, Wilmington, DE, USA) according
   to the manufacturer's instructions. After transfection, the cells were
   treated with compounds or vehicle control (0.1% DMSO) for 24 h. The
   luciferase activities were measured using the Luciferase Assay System
   (Promega) and detected by a Victor X5 multilabel plate reader
   (PerkinElmer).

   A collection of 6328 compounds was screened in the pGL4-PCSK9-P stably
   transfected cells at a concentration of 25 μg/mL for the primary
   screening. The screening compound library and the positive compounds
   7030B-C5, 7031B-H9, 7045B-E7 and 7045B-F7 are from the center for
   National New Drug Screening (Institute of Medicinal Biotechnology,
   Chinese Academy of Medical Sciences, Beijing, China).

2.3. Transfection with small interference RNA (siRNA)

   Three pre-designed siRNAs targeted to human HNF1α mRNA, FoxO3 mRNA,
   HINFP mRNA and one silencer negative control siRNA were obtained from
   Ambion. Transfection of siRNAs was carried out using the Lipofectamine
   RNAiMax reagent (Thermo Scientific) according to the manufacturer's
   instructions. The cells were incubated with serum-free OPTI-MEM medium,
   and transfected with the negative control (siNC) or specific siRNA at a
   final concentration of 100 nM. After 48 h of transfection, the cells
   were treated with the vehicle or 7030B-C5 (12.5 μM) for 24 h and
   harvested for further analysis.

2.4. Quantitative real-time PCR

   Total RNAs were isolated from cultured cells or tissues with the SV
   total RNA isolation system (Promega) and quantified by Nanodrop
   spectrophotometry (Thermo Scientific). RNA was reverse transcribed
   using TransScript First-Strand cDNA Synthesis SuperMix (TransGen).
   Quantitative real-time PCR (qPCR) was performed in the Bio-Rad CFX96
   real-time system (Bio-Rad, Hercules, CA, USA) using the SYBR green
   Master Kit (Roche, Mannheim, Germany) as previously described [76][35].
   qPCR primers are listed in Supplemental Table II.
   Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used to normalize
   the mRNA levels of target genes.

2.5. Western blotting

   Whole-cell lysates were prepared using RIPA lysis buffer (50 mM Tris
   HCl, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, pH
   7.4) containing complete EDTA-free protease inhibitor cocktail (Roche).
   Protein samples were separated by 10% SDS-polyacrylamide gel
   electrophoresis (PAGE) and transferred onto a 0.45-μm polyvinylidene
   fluoride (PVDF) membrane (Millipore, Bedford, MA, USA). The membranes
   were incubated with primary antibodies to LDLR (Abcam, Cambridge, UK),
   PCSK9 (R&D Systems, Minneapolis, MN, USA) HNF1α (Cell Signaling
   Technology, Danvers, MA, USA), FoxO1 (Cell Signaling Technology), FoxO3
   (Cell Signaling Technology), HINFP (Novus, Littleton, CO, USA) or GAPDH
   (ZSJQ-Bio, Beijing, China), followed by horse radish peroxidase
   (HRP)-conjugated secondary antibodies (ZSJQ-Bio). An enhanced
   chemiluminescence detection system (Millipore) was used for blot
   detection. The quantification of detection protein amounts was
   determined by Image J software, normalized to GAPDH.

2.6. DiI-LDL uptake assays

   After incubated with positive compounds, HepG2 cells were incubated
   with 2 mg/mL 1,1′-dioctadecyl-3,3,3′,3′- tetramethylindocarbocyanine
   perchlorate (DiI)-labeled LDL (Biomedical Technologies, Stoughton, MA,
   USA) for 4 h at 37 °C. Then cells were harvested, washed and
   resuspended in phosphate buffered saline (PBS). DiI fluorescence
   measurements were performed using an Epics XL flow cytometer (Beckman
   Coulter, Miami, FL, USA) as described previously [77][36].

2.7. Chromatin immunoprecipitation assay (ChIP assay)

   Cells cultured in MEM medium were treated with vehicle or 7030B-C5
   (12.5 μM) for 24 h. The ChIP assay was carried out using the ChIP Assay
   Kit (Cell Signaling Technology) according to the manufacturer's
   instructions. After 7030B-C5 treating, the cells were fixed in 1%
   formaldehyde for 10 min at room temperature. The chromatin was sheared
   to an average length of 150–900 bp by sonication. The chromatin
   extracts were immunoprecipitated at 4 °C overnight with anti-HNF1α,
   anti-FoxO3, or control IgG (Cell Signaling Technology). Immunocomplexes
   were isolated by binding to protein A-agarose beads. Precipitated DNA
   was isolated after ethanol precipitation. Quantitative real-time PCR
   was performed to analyze the promoter binding levels determined using
   the specific PCSK9 primers 5′-TCCAGCCCAGTTAGGATTTG-3′ and
   5′-CGGAAACCTTCTAGGGTGTG-3′.

2.8. Animal experiments and quantification of atherosclerosis

   All experimental procedures involving animals were approved by the
   Institutional Laboratory Animal Care and Use Committee of Institute of
   Medicinal Biotechnology. C57BL/6 J mice, ApoE KO mice were purchased
   from Vital River Laboratory Animal Technology (Beijing, China).

   To investigate the oral acute toxicity of active compounds, male
   C57BL/6 J mice (~8 weeks old) fed normal chow were divided into four
   groups (5 mice/group), and intragastric injected with active compounds
   dissolved in H[2]O (100 mg/kg bodyweight/day, 300 mg/kg bodyweight/day,
   500 mg/kg bodyweight/day, 1000 mg/kg bodyweight/day, respectively).
   Survival and behavioral characteristics were immediately observed as
   well as 10 min, 15 min, 30 min, 1 h, and 2 h for the first day. Then,
   it was observed for once a day lasting for a total of 7 days.

   To determine the effect of active compounds on PCSK9 expression in
   vivo, male C57BL/6 J mice (~8 weeks old) fed normal chow were divided
   into three groups (5 mice/group), and intragastric injected with
   vehicle and 7030B-C5 (30 mg/kg bodyweight/day), 7031B-H9 (50 mg/kg
   bodyweight/day), 7045B-E7 (100 mg/kg bodyweight/day), 7045B-F7
   dissolved in H[2]O (100 mg/kg bodyweight/day) for 4 weeks,
   respectively. Mice were then sacrificed with 1% sodium pentobarbital
   (40 mg/kg) followed by collection of liver samples individually to
   determine PCSK9 and LDLR expression by western blot and real-time PCR.

   To study the effects of active compounds on atherosclerosis, male ApoE
   KO mice (~8 weeks old) were fed a high-fat diet (HFD) containing 0.15%
   cholesterol and 20% lard or a regular rodent chow as a control. And the
   male ApoE KO mice (~8 weeks old) fed a HFD were divided into three
   groups (10 mice/group), and intragastric injected with vehicle and
   7030B-C5 (10 mg/kg bodyweight/day, 30 mg/kg bodyweight/day), 7031B-H9
   (50 mg/kg bodyweight/day, 100 mg/kg bodyweight/day). After 8 weeks of
   feeding, blood samples were collected for determination of lipid
   profiles by 7100 automatic biochemical analyzer (HITACHI, Tokyo,
   Japan). After 12 weeks of feeding, the animals were sacrificed followed
   by collection of mouse aorta, liver and blood samples. Liver samples
   were used to determine hepatic protein expression by Western blot.
   Liver tissues were fixed in 4% neutral buffered formalin and embedded
   in OCT freezing medium. The tissues were sliced and stained with
   hematoxylin-eosin (H&E) staining for morphological analysis, Oil Red O
   to detect lipid accumulation. Blood samples were collected and detected
   as described above. And serum PCSK9 levels were determined by an ELISA
   assay kit (Sino Biological, Beijing, China). Aortas were collected and
   stained with Oil Red O solution. To determine lesions in aortic root,
   frozen sections of aortic root were prepared and then stained with Oil
   Red O solution. The lesions in en face aortas and aortic root were
   quantitatively determined by computer-assisted image analysis method
   (Image J) and showed as percent lesion area [78][37].

2.9. Fast protein liquid chromatography (FPLC)

   Mice serum was fractionated using FPLC (Akta FPLC; GE Healthcare,
   Pittsburgh, PA, USA) using Superose 6 Increase 10/300 GL columns.
   Phosphate buffered saline (PBS; pH 7.4) eluent buffer was first passed
   through a 0.22 μm filter, and then the column was equilibrated at a
   rate of 0.25 mL/min. Serum sample aliquots of 500 μL were then injected
   into the column. Elutions were then carried out at a flow rate of
   0.25 mL/min. 63 fractions, 0.3 mL each, were collected in separate
   eluant eppendorf tubes for subsequent cholesterol content analysis by
   7100 automatic biochemical analyzer (HITACHI).

2.10. Hepatic transcriptome analysis

   Total RNA from ApoE KO mice livers were extracted using TRIzol reagent
   (Invitrogen). Subsequently, mRNA was purified from total RNA using the
   Oligotex mRNA Midi Kit (Qiagen, Hilden, Germany). Poly(A)+ mRNA was
   obtained using magnetic poly(A)+ Dynabeads (Invitrogen). For RNA-seq,
   the cDNA libraries were prepared with fragmented mRNA according to a
   TruseqTM RNA sample prep kit (Illumina, California, USA). The cDNA was
   then end-repaired, adenylated, and ligated to the sequencing adapters
   provided by Illumina. Finally, the libraries were enriched by PCR
   amplification and sequenced using an Illumina HiSeq 2000 platform
   (Majorbio Bio-Pharm, Shanghai, China). Sequence data has been deposited
   with the NCBI BioProject database under accession PRJNA602107.

2.11. Statistical analyses

   All experiments were repeated at least three times and representative
   results are shown in figures. The values are expressed as the mean ±
   SEM. Student's t-test was applied for comparing two groups and one-way
   ANOVA for multiple groups, followed by Bonferroni's correction as
   applicable (GraphPad Prism Software; Graph-Pad). P value of < 0.05 was
   considered significant. Error bars denote SEM.

3. Results

3.1. Discovery of novel PCSK9 inhibitors using cell-based high-throughput
screening (HTS) assays

   In order to establish a luciferase reporter-based HTS assay to find
   modulators targeting PCSK9 gene transcriptional expression, a 2112-bp
   fragment of PCSK9 gene promoter region was directionally inserted into
   the upstream of luciferase reporter gene of pGL4-Basic vector to
   construct the recombinant plasmid pGL4-PCSK9-P ([79]Fig. 1a).
   Subsequently, the HTS assay was built by stably transfecting plasmid
   pGL4-PCSK9-P into HepG2 cells and quantitatively assessed by Z’ factor
   [80][38] using berberine (BBR) as a positive control. BBR is a known
   inhibitor of PCSK9 transcription [81][30] ([82]Fig. 1b and Suppl Fig.
   1), which regulated PCSK9 expression through the modulation of
   transcriptional factors SREBP2 and HNF1α in hepatic cells. In our
   assay, BBR significantly repressed PCSK9 transcriptional activity in a
   dose-dependent manner, with an IC[50] of 10.26 μM (Suppl Fig. 1c).
   Besides, anacetrapib, a CETP inhibitor, which was reported to inhibit
   transcriptional activation of the PCSK9 gene by reducing the expression
   of mature form of SREBP2 [83][39], was used to evaluate the established
   in vitro HTS assay as well. The results showed that anacetrapib could
   also significantly reduce the PCSK9 transcriptional activity in a
   dose-dependent manner, with the IC[50] of 33.16 μM (Suppl Fig. 1d). In
   addition, the HTS assay achieved a good signal-to-background ratio with
   a low percent coefficient of variation, indicating that the model is
   suitable for high-throughput screening (Suppl Table 3).

Fig. 1.

   [84]Fig. 1
   [85]Open in a new tab

   Identification of novel PCSK9 inhibitors using cell-based
   high-throughput screening (HTS) assays. (a) The construction of
   recombinant plasmid pGL4-PCSK9-P. Human PCSK9 promoter region spanning
   −2112 to −1 bp, relative to the ATG start codon, was amplified by PCR,
   verified by DNA sequencing and cloned into pGL4-Basic vector between
   the Xho I and Hind III sites to produce pGL4-PCSK9-P. (b) The samples
   in compound library were screened by the established cell-based HTS
   assay for their capability of inhibiting PCSK9 transcription. In total,
   6328 compounds at 25 μg/mL were screened. The red dot represented PCSK9
   inhibitors 7030B-C5, 7031B-H9, 7045B-E7 and 7045B-F7 we identified. The
   blue dot represented positive control berberine (BBR). (c) HTS hit
   triage workflow resulting in the identification of 7030B-C5, 7031B-H9,
   7045B-E7 and 7045B-F7, with the number of compounds selected at each
   step. (d) The inhibitory activity and chemical structure of a novel
   PCSK9 inhibitor 7030B-C5, 7031B-H9, 7045B-E7 and 7045B-F7. The stable
   pGL4-PCSK9-P transfected HepG2 cells were treated with 7030B-C5,
   7031B-H9, 7045B-E7 or 7045B-F7 for 24 h and the luciferase activities
   were measured. The data represent the mean ± SEM of at least three
   independent experiments. (For interpretation of the references to color