High throughput small inhibitory RNA screening is an important tool that could help to more rapidly elucidate the molecular mechanisms of chemical cornea epithelial injury. Herein, we present the development and validation of exposure models and methods for the high throughput screening of hydrogen fluoride- and chloropicrin-induced cornea epithelial injury.
Toxicant-induced ocular injury is a true ocular emergency because chemicals have the potential to rapidly inflict significant tissue damage. Treatments for toxicant-induced corneal injury are generally supportive as no specific therapeutics exist to treat these injuries. In the efforts to develop treatments and therapeutics to care for exposure, it can be important to understand the molecular and cellular mechanisms of these injuries. We propose that utilization of high throughput small inhibitory RNA (siRNA) screening can be an important tool that could help to more rapidly elucidate the molecular mechanisms of chemical cornea epithelial injury. siRNA are double stranded RNA molecules that are 19-25 nucleotides long and utilize the post-transcriptional gene silencing pathway to degrade mRNA which have homology to the siRNA. The resulting reduction of expression of the specific gene can then be studied in toxicant exposed cells to ascertain the function of that gene in the cellular response to the toxicant. The development and validation of in vitro exposure models and methods for the high throughput screening (HTS) of hydrogen fluoride- (HF) and chloropicrin- (CP) induced ocular injury are presented in this article. Although we selected these two toxicants, our methods are applicable to the study of other toxicants with minor modifications to the toxicant exposure protocol. The SV40 large T antigen immortalized human corneal epithelial cell line SV40-HCEC was selected for study. Cell viability and IL-8 production were selected as endpoints in the screening protocol. Several challenges associated with the development of toxicant exposure and cell culture methods suitable for HTS studies are presented. The establishment of HTS models for these toxicants allows for further studies to better understand the mechanism of injury and to screen for potential therapeutics for chemical ocular injury.
Toxicant-induced ocular injury is a true ocular emergency because chemicals have the potential to rapidly inflict significant tissue damage. Unfortunately, treatments for toxicant-induced corneal injury are only generally supportive as no specific therapeutics exist to treat these injuries. The current treatment strategy is non-specific and primarily includes topical therapeutic treatments such as lubricants, antibiotics, and cycloplegics followed by anti-inflammatories (e.g., steroids) once the cornea has re-epithelialized1,2. Despite the best current therapeutic treatment options available, long-term prognosis is generally poor due to progressive corneal clouding and neovascularization2,3.
Animal models have traditionally been used to investigate chemical toxicity and understand mechanisms of injury. However, animal studies are time consuming and expensive. There are also efforts to reduce animal testing. For example, REACH legislation (EC 1907/2006) in the European Union has provisions intended to reduce animal testing. The provisions include a requirement that companies share data in order to avoid animal testing and obtaining approval from the European Chemicals Agency prior to performing proposed tests on animals. Under the provisions of REACH, animal testing should be a last resort. There is also the European Cosmetics Regulation (EC 1223/2009) that phased out the testing of cosmetics in animals. When animal studies are conducted, they are guided by the principles of 3Rs (Refinement, Reduction, and Replacement), which provide a framework for performing more humane animal research, reducing the number of animals used, and using non-animal alternatives where possible. For these reasons, the field of toxicology has sought to adopt in vitro assays that can provide insight into molecular mechanisms of toxicity and can be done in higher throughput4. This is a functional toxicology approach where toxicants are defined by their function rather than solely by their chemistry. Taken a step further, functional toxicogenomics seek to understand the role(s) that specific genes play in the effects of toxicants5. With the application of siRNA technology, screens to investigate gene function in the molecular and cellular responses to toxicants can be done at high throughput. siRNA are double stranded RNA molecules that are 19-25 nucleotides long that take advantage of the post transcriptional gene silencing pathway present in all mammalian cells6. These are synthetically made and designed to target a specific gene. When introduced into a cell, the siRNA is processed and one strand, the guide strand, is loaded into the RNA-induced silencing complex (RISC). The siRNA directs the RISC to a complementary region in an mRNA molecule, and the RISC degrades the mRNA. This results in the reduction of expression of the specific gene. The resulting reduction of expression of the specific gene can then be studied in toxicant exposed cells to ascertain the function of that gene in the cellular response to the toxicant. Such an approach has been used to further understand the mechanisms of ricin susceptibility and the AHR-dependent induction of CYP1A17,8.
The Chemical Terrorism Risk Assessment (CTRA) list and the toxic industrial chemicals (TIC) listings have itemized select chemicals based on their toxicity and potential to be released during a terrorist, warfare, or industrial accident event9. We are applying an siRNA high throughput screening (HTS) toxicogenomic approach to the study of CTRA list toxicants, which have been identified to be at high risk of use in a terrorist incident. Traditional toxicology seeks to understand the adverse effects that chemicals have on living organisms; however, we have a further desire to understand the mechanisms of injury for the purpose of informing the development of therapeutics and therapeutic approaches, and possibly, to discover molecules which can be targeted for therapeutic development. This effort in some ways may be considered analogous to the use of high throughput siRNA screening and cell based assays in the drug discovery process10. A major difference would be that drug discovery typically seeks a singular target for therapeutic discovery whereas in our approach it is somewhat unlikely that there would be a singular target with high therapeutic value for the treatment of toxicant exposure. We anticipate that any effective treatment paradigm for toxicant exposure would require a multi-faceted approach to achieve high therapeutic value, and toxicogenomic data may vitally inform an effective treatment paradigm.
Benchtop automation brings high throughput methodology to laboratories outside the pharmaceutical or biotech industries. The in vitro studies at our institute have historically been traditional assays which are low throughput11,12,13. In the past few years, our laboratory has transitioned to the use of benchtop robotics to perform high throughput siRNA screening. Herein, we present the refinement of ocular cell models and the development of in vitro exposure methods for hydrogen fluoride (HF) and chloropicrin (CP) suitable for high throughput siRNA screening. Our goal is to identify molecules that regulate cellular injury in response to these toxicants. The targets of the siRNA library we selected include G protein-coupled receptors, protein kinases, proteases, phosphatases, ion channels, and other potentially druggable targets. HF and CP were selected for study by cross-referencing CTRA list agents with the ToxNet reports of industrial accidents to find those that present the greatest risk of ocular injury via vapor exposure9,14. CP (chemical formula Cl3CNO2, CAS number 76-06-2) was originally used as a tear gas in WWI15. It is currently used as an agricultural fumigant and functions as a nematicide, fungicide, and insecticide16. Hydrogen Fluoride (HF) is used in processes including alkylation in oil refineries and electrochemical fluorination of organic compounds17. HF (chemical formula HF, CAS number 62778-11-4) is a gas but in its aqueous form is hydrofluoric acid (HFA, CAS number 7664-39-3). Therefore, we elected to use HFA in our in cell exposure models. The SV40 large T antigen immortalized human corneal epithelial cell line SV40-HCEC was selected for study. Cell viability and the inflammatory marker IL-8 were selected as endpoints because targets that are involved in cellular injury should be reflected in the cell death and the inflammatory response. Specifically, if a target were to play a protective role in toxicant exposure, cell death and/or inflammatory cytokine production should increase when the target expression is inhibited by siRNA. The opposite would be true for targets that play a negative role. Also, chronic inflammation appears to play a role in cornea pathology after exposure, and intervention in cell death pathways may improve clinical outcome2,18.
1. Cell Culture Maintenance
2. Plate Cells for Experimentation
3. Transfect Cells with siRNA
4. Refeed the Cells the Following Day
5. Positive Control Addition
6. HF Exposure of Cultured Cells
CAUTION: HFA is corrosive and acutely toxic.
7. CP Exposure of Cultured Cells
CAUTION: CP is acutely toxic and an irritant.
8. Sample Collection and Cell Viability Assay
9. Measure IL-8 Concentration in Cell Culture Supernatants
Exposure Method Development
We refined and evaluated the suitability of the human corneal epithelial cell line SV40-HCEC for use in HTS studies. SV40-HCEC were immortalized using the SV-40 large T antigen and were a gift from Dhanajay Pal23. There were too many variables explored in exposure methodology development to present concisely herein, and so, only some samples of findings we believe may be more broadly applicable to multiple toxicants are presented.
Toxicant Exposure by Spike or Medium Exchange
We initially sought to expose cells by spiking 5 µL of a working concentration of toxicant diluted in water, saline, or medium into 95 µL of medium already present into each well of cells in a 96-well plate to achieve the final desired exposure concentration. SV40-HCEC were exposed using this spike expose/refeed methodology to 1 X LD50 of HFA for 24 h (1 X LD50 = 0.02% HFA by this method when the HFA spike was diluted in medium), and wells were refed with fresh medium 10 min later. After 24 h, the cells were overlaid with MTT substrate as described above for the cell viability assay. Images were captured by photomicroscopy prior to solubilization of the formazan crystals with DMSO. Close examination of wells showed that a large degree of cell death was localized in one spot, and this is in spite of the fact that the plate was shaken for 15 s immediately after the spike was added to the entire plate (Figure 1a). This phenomenon was observed when the spike was added to the bottom of the well or when added at the meniscus. A higher magnification image of this area using phase contrast shows that cells are still present in the area unstained by MTT (Figure 1b). We also tested a medium exchange expose/refeed methodology where medium was removed from the wells, replaced with cell culture medium containing the 1 X LD50 of HFA (1 X LD50 = 0.035% HFA by this method) and wells were refed with fresh medium 10 min later. After 24 h, the cells were overlaid with MTT substrate as described above. This method achieved an apparently more even cell death response of cells within each well (Figure 1c and 1d). Unexposed controls are provided for reference and show no localized areas of cell death (Figure 1e and 1f). For CP exposures, the spike method did not result in a large degree of cell death localized to one spot within each well; however, the cell death response was more consistent from one iteration to the next using the medium exchange exposure method (data not shown).
Figure 1: Cell death induced by spike and medium exchange exposure methods. All images were captured 1.5 h after addition of MTT substrate. Panel (a) is a bright field image of SV40-HCEC exposed to HFA spike expose/refeed methodology. Panel (b) is a higher magnification of the same well in panel (a) using phase contrast. Panel (c) is a bright field image of SV40-HCEC exposed by medium exchange expose/refeed methodology. Panel (d) is a higher magnification of the same well in panel (c) using phase contrast. Panels (e) and (f) are from an unexposed well, bright field and phase contrast higher magnification, respectively. Please click here to view a larger version of this figure.
Toxicant Stability in Medium
It was observed that once diluted in medium, the apparent cytotoxicity of HF and CP changes and then stabilizes. HFA was diluted in medium to 0.0036% and allowed to incubate at 37 °C for 1-60 min prior to exposing SV40-HCECs as described above. For HFA, the cytotoxicity increases within the first 5 min that HFA is diluted in medium and is then stable for at least an hour (Figure 2a). One way ANOVA followed by Dunnett's Multiple Comparison Test showed that % cell viability at all time points were significantly different from the 1 min time point. There were no significant differences between the 2.5 to 60 min time points. CP is organic and is first diluted to 5% in DMSO. It was then diluted in medium to 0.0008% and allowed to incubate at 37 °C for 5-120 min prior to exposing SV40-HCECs as described above. Once diluted in medium, the cytotoxicity of CP decreases over 60 min and then stabilizes for at least the next h (Figure 2b). One way ANOVA followed by Dunnett's Multiple Comparison Test showed that % cell viability at all time points were significantly different from the 5 min time point. There were no significant differences between the 60 to 120 min time points.
Figure 2: Toxicant loss of cytotoxicity in medium dilutions. The data are expressed as the average of the percent viability ± SD (n=6). Panel (a) shows the stability of 0.0036% HFA cytotoxicity when diluted in medium and incubated at room temperature for 1-60 min prior to exposing SV40-HCECs as described above. Panel (b) shows the stability of 0.0008% CP cytotoxicity when diluted in medium and allowed to incubate at 37 °C for 5-120 min prior to exposing SV40-HCECs as described above. Please click here to view a larger version of this figure.
Expose/Refeed or Expose Only
Another factor investigated in exposure methodology was whether to: A) add toxicant for 10 min, wash, refeed, remove plates from the chemical fume hood, and incubate 24 h in a standard cell culture incubator (expose/refeed method); or B) add toxicant, leave it on, remove plates from the chemical fume hood, and incubate 24 h in a standard cell culture incubator (expose only method). Safety is an important factor as it may not be permissible to remove plates of cells that contain diluted toxicant from the chemical fume hood because the toxicant will off-gas and expose personnel. A related consideration is the number of plates to be exposed at a given time since the more plates there are, there will be a greater volume off-gassing. In our HTS methods, we expose 48 x 96-well plates in a given study day. We placed exposed cell culture plates into a sealed bag and then used colorimetric gas detector tubes to assess CP and HF off-gassing from exposed plates. We found that exposing 48 plates to 1 x LD50 of HFA did not result in any detectable HF off-gassing (data not shown). For CP exposures, we found that there was enough CP off-gassing from 48 plates exposed to 1 x LD50 to present at least some safety concern (data not shown). Both expose/refeed and expose only methods were evaluated for HFA exposures. SV40-HCEC cells were exposed to HFA and viability was assessed by MTT assay 24 h after exposure. Dose response iterations were performed on different days. Compared to the expose/refeed method (Figure 3a), we found that the expose only method produced more consistent cell viability assay results (Figure 3b). Therefore, this method was selected as part of the final exposure methodology. For CP, any differences in the consistency of cell viability assay results between the two different exposure methodologies could not be evaluated since safety concerns precluded removing the CP exposed plates from the chemical fume hood for 24 h incubation in a standard cell culture incubator.
Figure 3: Exposure consistency with expose/refeed and expose only methods. The data are expressed as the average of the percent viability ± SD (n=6). (a) HFA was diluted in medium, incubated at room temperature for 10 min, and then used to expose cells by the expose/refeed method. (b) HFA was diluted in medium, incubated at room temperature for 10 min, and then used to expose SV40-HCEC by the expose only method. Please click here to view a larger version of this figure.
Cell Model Refinement
Medium Constituents
We adhered to the culture conditions established by the originating investigators for the passage and maintenance of these cell lines; however, we modified them for cells in experimentation. The cell culture medium prescribed by the originating investigators for the SV40-HCEC cells does not include hydrocortisone, but we utilized a low level of hydrocortisone (0.5 µg/mL) during exposure studies to suppress the occasional spike in cytokine production seen in naive SV40-HCEC (data not shown) which adversely affected data analysis from one iteration to the next during the refinement studies.
Phenotypic Stability
During our cell model refinement studies, we found that the SV40-HCEC begin to have decreased cytokine production in response to toxicant exposure starting around passage number 60 (data not shown). For this reason, we maintained cryostocks of low passage cells, executed all studies using cells under passage number 60, and monitored cytokine production during the screening process.
Plate Layout
An important source of experimental variability in high throughput studies is edge effects of multi-well plates. We found that the cytokine responses of cells in edge wells were not consistent with, or equivalent to, interior wells in both HFA and CP studies (data not shown). Median polish of the data, a statistical exercise used to minimize the edge effects on a data set, did not sufficiently resolve the issue (data not shown). Therefore, no cell plate edge wells were used for experimentation or controls (Figure 4).
Figure 4: Library and cell plate layouts. The grey shaded areas in the Library Plate Layout represent wells that contain siRNA. The white wells are empty. The 40 targets from rows A-D of the Library Plate are used in the "Top Plate Set", and the 40 targets from rows E-H of the Library Plate are used in the "Bottom Plate Set". Wells B2-B6 are used for negative pool siRNA controls. Wells C2-C6 are non-transfected. Wells B7-B11 are used for either cardamonin or SKF 86002 positive control drugs as described, and wells C7-C11 are used for DMSO vehicle control. Please click here to view a larger version of this figure.
In Vitro Model Validation for HTS
We used Z' factor analysis to determine suitability of SV40-HCEC for use in HTS studies as this statistical test is used to determine assay quality for HTS24. For HF Z' factor analysis, cells were seeded at 1.25 x 103 cells/well, transfected with negative pool siRNA (as described above) 24 h after seeding, and exposed 48 h after transfection to 0.0036% HFA (1 x LD50), n=3. For CP Z' factor analysis, cells were seeded and transfected as described above, and exposed 48 h after transfection to 0.0008% CP (1 x LD50), n=5. A greater n was used for the CP studies owing to the greater variability in this exposure model. Z' factor was calculated for IL-8 expression data from unexposed vs. exposed cells and negative pool siRNA transfected unexposed vs. exposed cells for CP and HFA exposures. We also analyzed non-transfected cells to determine if there was an effect on assay quality due to siRNA transfection. Many iterations of the validation studies were performed while refining exposure and culture conditions in an effort to maximize the Z' factor results. SV40-HCEC passed validation and most iterations had Z' factor values greater than 0.50 for HF exposures (Table 1) and CP exposures (Table 2). Each replicate in these tables was from cells plated on different days.
Exposure Groups | Z’-Factor Analysis | ||
Rep. #1 | Rep. #2 | Rep. #3 | |
HF-Exposed Negative Control vs. Unexposed Negative Control | 0.68 | 0.5 | 0.56 |
HF-Exposed vs. Naïve | 0.4 | 0.68 | 0.56 |
Table 1: SV40-HCEC HFA Exposure Z` Factor Analysis of IL-8
Exposure Groups | Z’-Factor Analysis | ||
Rep. #1 | Rep. #2 | Rep. #3 | |
CP-Exposed Negative Control vs. Unexposed Negative Control | 0.6 | 0.18 | 0.46 |
CP-Exposed vs. Naïve | 0.42 | 0.29 | 0.42 |
Table 2: SV40-HCEC CP Exposure Z’ Factor Analysis of IL-8
Primary siRNA Screening Results of HFA Injured Cells
Transfection optimization studies were performed prior to screening, and siRNA targeting cyclophilin b was used for these studies. An in situ RNA detection assay was used to measure cyclophilin b mRNA levels, and a high content analyzer was used to quantify the results. We achieved near 90% knockdown of cyclophilin b and typically no more than 5-10% cell loss with 4 pmol of siRNA/well and 0.3 µL/well transfection reagent (data not shown). Higher amounts of siRNA actually resulted in poorer cyclophilin b knockdown (data not shown). Similar levels of knockdown were achieved with 1.2 pmol/well of siRNA and 0.09 µL/well transfection reagent, but we elected to use 4 pmol per well in order to address any targets that could require greater amounts of siRNA to achieve effective knockdown. Target knockdown kinetics were also evaluated, and it was observed that target knockdown was maximal by two days post-transfection (data not shown). Target knockdown consistency across the plate was also validated (data not shown). The anti-inflammatory drugs cardamonin and SKF 86002 are used as positive controls for the inhibition of cytokine production for HFA and CP exposure, respectively. These were selected by testing a variety of compounds with known anti-inflammatory properties in exposed cells. The dose utilized in screening studies was selected by dose response optimization studies.
The high throughput siRNA screening strategy we utilized is industry-standard and involves three major steps: 1) primary screening, 2) library deconvolution, and 3) target validation. In primary screening, the phenotype of each target is evaluated utilizing a mixture of different siRNA sequences as a single reagent to target each gene. Targets are then down-selected to library deconvolution. In this step, each of the different siRNA sequences that are pooled to target each gene in primary screening are tested separately. Targets undergo another down-selection into target validation in which the target phenotype is confirmed with drugs, phenotype restoration, and/or siRNA from another vendor. For our primary screen, we elected to perform a focused sub-genomic screen rather than a genome wide screen for reasons of economy. Microarray data from HF and CP injured mouse corneas and Ingenuity Pathway Analysis (IPA) were used to focus our target library for the primary screen (manuscripts in preparation). Briefly, mouse corneas were exposed to toxicant vapor using a vapor cup similar to previously described methods25. Exposed cornea buttons and controls were harvested at various time points post exposure. RNA was isolated, and the quality and amount of RNA monitored. All microarray experiments were performed according to the manufacturer's protocol26. Raw signal intensities were normalized and analyzed by principal component analysis (PCA) to determine the significant sources of variability in the data. Genes were rank ordered by statistical significance. The data were mined and compared to gene lists of preconfigured siRNA libraries. From these gene lists, we selected 3120 targets for screening. The siRNA library was screened in SV40-HCEC cells (see Figure 5 for a flow chart of the HTS protocol for the primary screen). Endpoint assessments included IL-8 levels in cell culture medium by no-wash bead-based assay and cell viability by MTT assay. The cells were exposed to 0.0036%HFA, as previously described, which is approximately 1 x LD50. Cell viability was assessed by MTT assay 24 h after exposure. The effect of each siRNA pool on cell viability was analyzed for statistical significance relative to the exposed negative pool average for each plate using strictly standardized mean difference (SSMD)27. A dual-flashlight plot of the SSMD and cell viability fold-change is shown in Figure 6. The hit selection thresholds chosen for targets that exacerbated cell death or improved cell viability were SSMD ≤ -1.28 and SSMD ≥ 1.28, respectively.
Figure 5: Primary Screen HTS protocol Flow Chart. This flow chart shows the essential steps that are performed on each day for the HTS protocol. Please click here to view a larger version of this figure.
Figure 6: Dual-Flashlight Plot of the siRNA Effect on Cell Viability of HFA-Exposed SV40-HCEC Cells. Results shown for each target are the average of six replicates (n=6). Cell viability data are expressed as a fold-change relative to the exposed negative pool control siRNA, and statistical significance was analyzed by SSMD. Targets which had a SSMD ≤ -1.28 or ≥ 1.28 were selected as hits. Please click here to view a larger version of this figure.
The effect of each siRNA pool on HFA-induced IL-8 production was analyzed for statistical significance relative to the exposed negative pool average for each plate. IL-8 in cell culture supernatant was assessed assay 24 h after exposure. A dual-flashlight plot of the SSMD and IL-8 fold-change is shown in Figure 7. The hit selection thresholds chosen for targets that decreased IL-8 production were a 40% decrease and SSMD ≤ -1.0. The hit selection thresholds chosen for targets that increased IL-8 production were a five-fold increase and SSMD > 1.28.
Figure 7: Dual-Flashlight Plot of the siRNA effect on IL-8 Production by HFA-Exposed SV40-HCEC Cells. Results shown are the average of six replicates (n=6). IL-8 expression level data are expressed as a fold-change relative to the exposed negative pool siRNA, and statistical significance was analyzed by SSMD. Hits were defined as targets which had a 40% decrease in IL-8 levels and SSMD ≤ -1.0, or a five-fold increase in IL-8 levels and SSMD > 1.28. Please click here to view a larger version of this figure.
Herein we describe our methods and results on the development of a high throughput cornea epithelial cell screening model for the study of HF and CP injuries. We also present the results from the primary siRNA screen for HF injury. There were many challenges to the development of HTS models for the study of TIC injuries. Methods that we could find in the literature related to the study of HF, HFA or CP in cell culture models were of little help. Most in vitro studies on the fluoride ion involve oral cells and utilized sodium fluoride and not HF or HFA (for some examples, see28,29). Some methodology on the in vitro exposure of bronchial epithelial cells to CP has been published by Pesonen et al.30. Ultimately, we found no specific methods in the literature related to the HTS study of HF, HFA or CP, and the vast majority of challenges and variables related to the study of these toxicants in HTS required development. Particular attention was paid to achieving the high degree of exposure consistency across time and numerous screening plates necessary for a successful HTS study.
One type of challenge encountered was related to the application of toxicant to cell culture systems. Some toxicants are either reactive with or unstable in aqueous solutions. The toxicants we selected for study are not regarded as having these instability issues. An agricultural study investigating photohydrolysis of CP showed that a 0.001 M solution of CP in water had a half-life of 31.1 h when exposed to an 1100 lux xenon light source (sunlight intensity of an overcast day) for 10 days, 12 h per day31. There was no measurable hydrolysis over the same 10 day period when the solution was stored in the dark. The proton and fluoride ion of HFA are clearly stable in water. There are, however, additional stability considerations with regard to toxicant dilutions made in cell culture medium. We found that there was an initial change of cytotoxicity to some degree when the toxicants were first diluted in cell culture medium that then stabilizes after a brief period of time. We believe that this is due to the toxicant binding or complexing with constituents in the medium. Cell medium contains calcium, magnesium, protein, etc. that are known to interact with the fluoride ion32,33. CP is known to have oxidative reactions with biological thiols, but it may also bind DNA, proteins, and other nucleophiles34,35. Once these reactions and interactions have completed or reached equilibrium, the amount of available toxicant stabilizes and thus the apparent cytotoxicity stabilizes. We initially explored the use of saline and water dilutions of toxicants for working stocks used for exposures to avoid toxicant interactions with medium constituents in working dilutions. The use of saline would also preserve the acid component of HFA injury for exposures performed by medium exchange. We found that the cell death responses to toxicants in water and saline dilutions were more variable than the method that used cell culture medium dilutions with medium exchange (data not shown). Regarding the spike exposure method with HFA, the variability may be due to inconsistency in the spike dispersing in the well leading to cellular targets competing with medium constituents for the available fluoride ion. The reasons for the variability in HFA exposures by medium exchange with HFA saline dilutions are entirely unclear. In any event, exposure methods using saline dilutions for HFA were abandoned and not explored for CP. The toxicity of our HFA dilutions in medium probably closely mimics studies using sodium fluoride because the cell medium buffers the free proton in our HFA dilutions.
Other challenges we encountered were related to refinement of the cell culture models for use in HTS. We found it necessary to modify the cell culture medium constituents for cells in experimentation, specifically hydrocortisone and antibiotics. We utilized a low level of hydrocortisone (0.5 µg/mL) during exposure studies to suppress the occasional spike in cytokine production seen in naive SV40-HCEC. This low amount of hydrocortisone did not adversely affect the responses of the cells to toxicant and was below levels typically used to suppress cytokine production by cultured cells in response to stimulant36,37. The exact reasons why naive SV40-HCEC cells occasionally produce excess cytokines are not known, but it is possible that the cells are variably sensitive to the minor stressors that occur during cell passaging and plating. Many labs routinely include antibiotics in cell culture maintenance medium, but some antibiotics, tetracyclines in particular, have been shown to exhibit anti-inflammatory effects on cultured cells including corneal epithelial cells38. In general, our laboratory avoids the use of antibiotics whenever possible since these can potentially confound studies. We also investigated the phenotypic drift over multiple passages of our cells in culture. It is common for immortalized cells to undergo genetic and/or phenotypic drift over multiple passages, if not maintained in the exponential phase, if allowed to reach confluency, or if exposed to certain environmental stressors39,40,41. Therefore, it is critical during a high throughput screen that cell lines are properly maintained and that the screen is completed using cell passage numbers that maintain the known phenotypic response. Multi-well plate edge effects is the phenomenon that cells grown in the outer perimeter wells of multi-well plates often do not respond the same or with the same consistency as cells grown in interior wells42. There are statistical exercises, such as median polish, that can be utilized to minimize edge effects on a data set43. However, there is no statistical exercise that can eliminate edge effects on a particular target or control if it is always tested in an edge well, and varying the plate layout between replicates to move targets or controls off of edge wells is not logistically practical or cost effective. It is our opinion that the decision to include the use of edge wells in cell-based HTS assays should be made cautiously. We found it necessary to eliminate the use of cells grown in edge wells for controls and targets.
Our HTS cell model and exposure methodology for in vitro HF injury was used to successfully complete the primary siRNA screen of 3120 targets. SSMD was used for hit selection because it was proposed specifically to measure the magnitude of difference between a control and an siRNA27. Our results for cell viability show that there was a linear relationship in the dual-flashlight plot between SSMD and fold-change for virtually all targets. For this reason, we used the single criterion of SSMD for hit selection for this endpoint. The IL-8 data differed in that some targets had weak but consistent effects while others showed strong but inconsistent effects. Therefore, we used both fold-change and SSMD for hit selection criteria for this endpoint. The SSMD values used for hit selection for both cell viability and IL-8 data were chosen because cutoffs between 1 and 1.645 (or -1 and -1.65) achieve low false discovery and non-discovery rates44. There were some targets that were hits in both the cell viability and IL-8 data. In these cases, preference for inclusion in the deconvolution library was given to those targets that exerted their effects in the same direction of improving overall health (improved viability and decreased IL-8) or exacerbating injury (increased cell death and increased IL-8). Overall, we selected a total of 250 targets between the cell viability and IL-8 endpoints to make up the deconvolution library for HF injury. The primary screen for in vitro CP injury is in progress.
In summary, we have developed exposure and culture conditions suitable for the HTS study of ocular injury by HF and CP. Several variables in exposure and culture conditions were refined and optimized, and the models were validated to be suitable for HTS studies by Z' factor analysis. We have further proven the utility of these models by completing the primary screen of an siRNA library in the HF model. There are many injuries and diseases that are of no or limited interest to the pharmaceutical or biotech industry, toxicant injury by CP and HF included, and the advent of benchtop robotics facilitates the HTS study of injuries and diseases in virtually any laboratory. High throughput genomic data sets can greatly contribute to the understanding of the roles that specific genes play in injury or disease which can help to more efficiently focus follow on in vitro or in vivo studies.
The authors have nothing to disclose.
This research was supported by the National Institutes of Health CounterACT Program Interagency Agreement# AOD13015-001. We would like to thank Stephanie Froberg and Peter Hurst for their efforts and expertise on video production.
Bravo liquid handing platform | Agilent or equivalent | G5409A | |
Bravo plate shaker | Agilent or equivalent | Option 159 | |
Bravo 96LT disposable tip head | Agilent or equivalent | Option 178 | 96-channel large tip pipetting head unit |
Bravo 96ST disposable tip head | Agilent or equivalent | Option 177 | 96-channel small tip pipetting head unit |
Bravo 384ST disposable tip head | Agilent or equivalent | Option 179 | 384-channel small tip pipetting head unit |
Bravo 96 250 μL sterile barrier tips | Agilent or equivalent | 19477-022 | |
Bravo 384 30 μL sterile barrier tips | Agilent or equivalent | 19133-212 | |
Bravo 384 70 μL sterile barrier tips | Agilent or equivalent | 19133-212 | |
EnSpire multimode plate reader | Perkin Elmer or equivalent | 2300-0000 | AlphaLISA assay detector with high power laser excitation |
IL-8 (human) AlphaLISA Detection Kit | Perkin Elmer or equivalent | AL224F | no-wash bead-based assay |
ProxiPlate-384 Plus white 384-shallow well microplates | Perkin Elmer or equivalent | 6008359 | |
Lipofectamine RNAiMAX | Invitrogen or equivalent | 13778500 | Transfection reagent |
Opti-MEM 1 Reduced Serum Medium | Invitrogen or equivalent | 31985070 | |
TrypLE Express | Gibco or equivalent | 12605010 | Cell detachment solution |
IncuCyte Zoom | Essen Instruments or equivalent | ESSEN BIOSCI 4473 | Incubator-housed automated microscope |
Chloropicrin | Trinity Manufacturing or equivalent | N/A | Acute toxicity and irritant |
DMEM-F12 cell culture medium | Invitrogen or equivalent | 11330-057 | Contains HEPES |
Fetal bovine serum | Invitrogen or equivalent | 1891471 | |
Human epidermal growth factor (cell culture grade) | Invitrogen or equivalent | E9644-.2MG | |
Recombinant human insulin (cell culture grade) | Invitrogen or equivalent | 12585-014 | |
Penicillin-Streptomycin solution (cell culture grade) | Invitrogen or equivalent | 15140122 | |
Hydrocortisone (cell culture grade) | Sigma or equivalent | H0888-10G | |
Glucose (cell culture grade) | Sigma or equivalent | G7021 | |
PBS (cell culture grade) | Sigma or equivalent | P5493 | |
siRNA | Dharmacon or equivalent | various | |
Thiazolyl blue tetrazolium bromide | Sigma or equivalent | M5655 | MTT assay substrate |
siRNA buffer | Thermo or equivalent | B002000 | |
96-well cell culture plates | Corning or equivalent | CLS3595 | |
T150 cell culture flasks | Corning or equivalent | CLS430825 | |
BSL-2 cell culture hood | Nuaire or equivalent | NU-540 | |
300 mL robotic reservoirs | Thermo or equivalent | 12-565-572 | |
96 baffled automation reservoirs | Thermo or equivalent | 1064-15-8 | |
500 mL sterile disposable storage bottles | Corning or equivalent | CLS430282 | |
Microplate heat sealer | Thermo or equivalent | AB-1443A | |
Microplate heat sealing foil | Thermo or equivalent | AB-0475 | |
Cardamonin | Tocris or equivalent | 2509 | Anti-inflammatory, used as positive control |
SKF 86002 | Tocris or equivalent | 2008 | Anti-inflammatory, used as positive control |
DMSO | Sigma or equivalent | D8418 | |
48% hydrofluoric acid | Sigma or equivalent | 339261 | Corrosive and acute toxicity |
1000 μL Single channel pipettors | Rainin or equivalent | 17014382 | |
200 μL Single channel pipettors | Rainin or equivalent | 17014391 | |
20 μL Single channel pipettors | Rainin or equivalent | 17014392 | |
1000 μL 12-channel pipettors | Rainin or equivalent | 17014497 | |
200 μL 12-channel pipettors | Rainin or equivalent | 17013810 | |
20 μL 12-channel pipettors | Rainin or equivalent | 17013808 | |
Pipettor tips 1000 μL | Rainin or equivalent | 17002920 | |
Pipettor tips 200 μL | Rainin or equivalent | 17014294 | |
Pipettor tips 20 μL | Rainin or equivalent | 17002928 | |
Chemical fume hood | Jamestown Metal Products | MHCO_229 | |
384-well sample storage plates | Thermo or equivalent | 262261 | |
Sodium chloride | Sigma or equivalent | S6191 | |
50 mL conical tubes | Thermo or equivalent | 14-959-49A | |
Serological pipettes 50 mL | Corning or equivalent | 07-200-576 | |
Serological pipettes 25 mL | Corning or equivalent | 07-200-575 | |
Serological pipettes 10 mL | Corning or equivalent | 07-200-574 | |
Serological pipettes 5 mL | Corning or equivalent | 07-200-573 | |
SV40-HCEC immortalized human corneal epithelial cells | N/A | N/A | These cells are not commercially available, but can be obtained from the investigators cited in the article |
Sceptor Handheld Automated Cell Counter | Millipore or equivalent | PHCC20060 | |
GeneTitan Multi-Channel (MC) Instrument | Affymetrix or equivalent | 00-0372 | |
Affymetrix 24- and 96-array plates | Affymetrix or equivalent | 901257; 901434 | |
Draegger tube HF | Draeger or equivalent | 8103251 | |
Draegger tube CP | Draeger or equivalent | 8103421 | |
Draegger pump | Draeger or equivalent | 6400000 | |
Clear Plate seals | Resesarch Products International or Equivalent | 202502 | |
Reagent reservoirs | VistaLab Technologies or equivalent | 3054-1000 | |
Xlfit | IDBS or equivalent | N/A | Excel add-in used for automated curve fitting |