Identification of small molecules that reduce growth of human lung adenocarcinoma cell lines
We screened a library of 189, 290 small molecules for agents that would inhibit growth of human lung adenocarcinoma cell lines. The cell lines that were used in the primary screen have been well characterized and are shown in (Paez et al. 2004
; Pao et al. 2004
; Pao et al. 2005
; Sordella et al. 2004
; Tracy et al. 2004
). Three of the cell lines (H1650, H3255 and H1975) contain EGFR
alleles found commonly in human lung adenocarcinoma cell lines (delE746-A750, L858R and L858R-T790M, respectively). These cell lines display different sensitivity to the EGFR tyrosine kinase inhibitor gefitinib (). Growth of H3255 is the most sensitive to gefitinib (IC50
= 10 nM); this cell line also has amplification of EGFR
(Tracy et al. 2004
). In contrast, growth of H1975 (with the L858R-T790M allele) is inhibited by gefitinib only at high concentrations (IC50
= 10 μM). Introduction of allele-specific EGFR
siRNAs into H1975 or H3255 induces apoptosis, suggesting that both lines require continuous expression of the EGFR
oncogene for survival ((Sordella et al. 2004
); R. S, unpublished). The H1650 cell line shows intermediate sensitivity to gefitinib (IC50
= 1 μM). The H2030 cell line contains the G12C allele of KRAS
; growth of this line is inhibited only at high concentration of gefitinib (IC50
= 10 μM). Two lines of non-tumor cells were used to help eliminate compounds that showed general cytotoxicity: primary human bronchiolar epithelial cells (NHBE) and human lung fibroblasts (WI-38). We sequenced exons 18–24 of EGFR
and exon 2 of KRAS
and did not detect any mutations in either of these two genes in the control cells (results not shown). NHBE and WI-38 cells showed intermediate sensitivity to gefitinib (IC50
= 1 μM).
Characteristics of human lung adenocarcinoma cell lines and control cells
Our screening strategy is outlined in . The primary screen was performed at a single drug concentration of 10 μM against H1650, H1975, H2030 and H3255. Cell growth (number of viable cells) remaining at the end of the assay was determined using the vital dye Alamar blue. This growth assay does not differentiate between cytostatic or cytotoxic compounds. Before conducting the screen, we validated the screening platform and assay by screening a library of approximately 3, 000 compounds against 5 cell lines (H1650, H1975, H2030, H3255 and HPL1D) on two separate days. The robustness and reproducibility of the assay were assessed using the Z′ statistical parameter (Zhang et al. 1999
). The high signal: noise ratio of the assay and the Z′ values of 0.5–0.8 are consistent with a robust assay (Table S1
Outline and summary of cell-based small molecule screen
Compounds that reduced growth of at least one of the adenocarcinoma cell lines at 10 μM by μ 50% were considered “positives”. Using this criterion, we identified 6, 552 novel small molecules for further analysis (). Out of this group, 3, 112 compounds were re-synthesized and tested again for their ability to reduce growth and the activity of 669 compounds was confirmed. Dose response studies were then performed with the panel of adenocarcinoma cell lines and two control lines. In parallel, this group of 669 compounds was tested for solubility in growth media, and compound purity and structure were tested by LC-MS. Forty nine compounds had either low solubility in growth media or contained impurities that raised questions about the identity and structure of the molecules. We were therefore able to titrate the activity of only 620 of the candidates. Fifty nine compounds inhibited the growth of at least one tumor cell line with IC50 μ ≤ 1 μM. We then selected molecules as candidates for further study based on the following criteria: 10-fold better growth inhibition of at least one adenocarcinoma cell line over control cells; ease of making chemical derivatives and affinity reagents; a lack of structural similarity to ATP; and strong structure-activity relationships (SAR) of analogs that were present in the library. Of the 59 compounds with IC50 μ 1μM for inhibition of growth of at least one lung cancer line, 9 showed a 10-fold selectivity (based on IC50 values) for tumor cells over the control cells. None of these compounds resembled ATP in structure and there were no analogs of 5 of these 9 compounds in the library, preventing an estimate of SAR. Therefore, only 4 compounds met all of the criteria outlined above.
The growth-inhibitory profiles of these compounds are shown in . To simplify discussion, the four compounds selected for further study were given the acronym LCS (lung cancer screen)-1 to 4. All four impaired the growth of at least two adenocarcinoma cell lines with an IC50 < 1 μM. Each of the lung cancer lines contains a known mutation affecting either EGFR or KRAS, yet the pattern of growth inhibition does not correlate with these mutations. This is most evident in the cases of LCS-2 and LCS-4. LCS-2 shows no activity against the H3255 line but it is highly active against both the H1650 and H1975 lines, which also have EGFR mutations, and it is quite active against the KRAS mutant line, H2030. Similarly, LCS-4 is moderately active against H1975 and H2030 but shows no activity against the two other cell lines with EGFR mutations (H1650 and H3255). LCS-1 and LCS-3 inhibited growth of all four lines, but with different patterns of activity: LCS-1 is at least ten-fold more active against H1975 than against H1650 and shows intermediate strength against H3255 and H2030; LCS-3 is most active against H1650. LCS-1 has moderate inhibitory effect on the control cell lines, but LCS-2, and -3 have at least ten-fold lower activity on NHBE cells. Since LCS-4 is inactive against two tumor cell lines, it was presumed not to have generalized activity against common growth mechanisms and was not tested with the control cells.
Cellular activity and chemical properties of candidate small molecules
Taken together, these findings are most compatible with the idea that the active compounds are targeting the products of genes, either wild-type or mutant, on which the growth-inhibited cancer cell lines have become dependent on as a result of still unidentified mutations specific to each line. The affected proteins might or might not be components of the EGFR or KRAS signaling machinery. A satisfactory explanation of the observed patterns of sensitivity of tumor cell lines to these small molecules will probably require a fuller description of the genotypes of the cell lines and identification of the molecular targets of the compounds.
As shown in , the four candidates are all of low molecular weight and are predicted to cross the plasma membrane readily, based on the membrane partition coefficient (cLogP). The cLogP value of a compound is the logarithm of its partition coefficient between n-octanol and water ((log Coctanol
)) and is a well-established measurement of the hydrophilicity of a compound (Ghose et al. 1999
). Compounds that readily permeate the plasma membrane have a cLogP value of < 5.
2-phenylpyridazin-3(2H)-ones as growth inhibitors of human lung adenocarcinoma cell lines
The primary HTS data provided useful information about the relationship of structure to activity (SAR) in the vicinity of the 2-phenylpyridazin-3(2H)-one scaffold of LCS-1. The SAR data are based on the differential activity of a cluster of 123 derivatives tested on the four cell lines. The SAR data strongly suggest that the pyridazine-3(2H)one moiety of LCS-1 requires an electron-deficient environment on the pyridazine ring with closely packed substitutions such as halogens and halogen mimetics as electron withdrawing groups tolerated at positions R2 and R3. Substitutions of electron-rich donors at these positions eliminated biological activity. Surprisingly, methoxy group substitutions on the R1 position did not affect activity, whereas a hydroxyl group at the same position completely abolished activity. This observation offered the first potential point of attachment for a linker that joins an active compound to a solid support for affinity chromatography. The phenyl moiety of LCS-1 was found to tolerate the most substitutions found within the cluster identifying the R5 position as a second potential point of attachment for a linker. The other candidates LCS-2, LCS-3 and LCS-4, were not represented by as many analogs in the library. As shown in , LCS-1 is not related to ATP in structure and is therefore unlikely to be an ATP-competitive inhibitor. Based on its chemical structure, LCS-1 appeared to be easily amenable to chemical modification. LCS-1 was therefore the first compound chosen for further study.
Structure and growth inhibitory activity of LCS-1, LCS-1.28 and LCS-1.34
Synthesis of a focused library of LCS-1 derivatives
Based on the SAR profile obtained from primary HTS data, we synthesized a focused library of 41 LCS-1 analogs to further explore the effects of substitutions on the R2
positions on the pyridazine ring and to assess the feasibility of linker attachments at the R1
position of the pyridazine ring and at the R5
position of the phenyl moiety of the molecule. Supplementary Table 2
summarizes the screening data with 10 cell lines, including two control cell lines. All analogs were soluble within the range of concentrations used for dose-response studies. One of the new analogs showed slightly increased potency compared to the parental molecule, 13 were inactive and the rest inhibited growth with IC50
values with a potency similar to LCS-1 (Supplementary Table 2
). Chemical modifications on the R2
positions were found to tolerate only halogens or halogen mimetics, consistent with the preliminary SAR data.
The growth-inhibitory activity profile of two active analogs compared with the parental compound LCS-1, is shown . Chloro-to-bromo substitutions on the R2 and R3 positions on the pyridazine ring resulted in the analog LCS-1.28 (4,5-dibromo-2-m-tolylpyridazin-3(2H)-one). This analog had a similar activity profile to the parental molecule. Chlorination of the ortho and the meta positions on the phenyl ring of LCS-1 produced analog LCS-1.34, which was slightly more potent than LCS-1 and LCS-1.28 in inhibiting growth of H1975.
Reduction of growth of additional lung adenocarcinoma cell lines by LCS-1, LCS-1.28 and LCS-1.34
We expanded the panel of human lung adenocarcinoma cell lines used to test LCS-1 to include a total of 16 cell lines (8 with EGFR mutations and 8 with KRAS mutations). LCS-1 inhibited the growth of 9 of these cell lines (some with EGFR and some with KRAS mutations) at concentrations 10- to 40-fold lower than levels required to inhibit growth of NHBE or WI-38 cells (). When tested against one or two of the LCS-1 derivatives, the tumor cell lines were, in general, similarly sensitive to LCS-1, LCS-1.28 and LCS-1.34. However, NHBE and WI-38 cells were two- to three-fold more sensitive to LCS-1.28 and LCS-1.34 than to LCS-1.
Growth inhibitory profile of LCS-1, LCS-1.28 and LCS-1.34
Determination of the half-life of LCS-1, LCS-1.28 and LCS-1.34 in human liver microsomes
To determine which of the three most active substituted phenyl pyridazinone compounds would be the best for in vitro and in vivo assays, we measured their stability in human liver microsomes. LCS-1 had a half-life of 6.6 min and this was similar to the half-life of LCS-1.28 (7.1 min). LCS-1.34 exhibited a 13-fold increase in half-life (82.2 min) in human liver microsomes over the parental molecule, and it was therefore chosen for all further experiments.
Inhibition of growth by the LCS-1.34 may be due to both inhibition of proliferation and induction of apoptosis
To determine the cellular mechanism by which compounds related to LCS-1 reduced the number of cells, we examined the effect of the most stable analog, LCS-1.34, on DNA synthesis and apoptosis. H358, H1975 and H3255 cells were treated with increasing concentration of LCS1.34 for 24 h then DNA synthesis was assessed by thymidine incorporation for the final 2 h. DNA synthesis was reduced in a concentration-dependent manner to 47 ± 3% and 42 ± 17% of control values (0.25 μM LCS-1.34) in H358 and H1975, respectively (). H3255 was the least sensitive of the three cell lines, showing only a 16% reduction in thymidine incorporation (84 ± 9% of control) after treatment with 0.25 μM LCS-1.34 (). However, at 1 μM, there was a further reduction in DNA synthesis to 30 ± 8% of control in H3255 cells, with more profound reductions in the other two cell lines.
LCS-1.34 inhibits DNA synthesis and induces apoptosis
To determine the effect of LCS-1.34 on cell death, H358, H1975 and H3255 cells were treated for 48 h with LCS-1.34, and annexin-FITC conjugates bound to the surface of treated cells was measured by FACS. LCS-1.34 induced apoptosis in the three cell lines to different extents (): LCS-1.34 at 0.5 μM increased the number of H358 and H1975 cells undergoing apoptosis by three-fold (from 8 to 24 % in H358 cells and from 3.5 to 11% in H1975 cells). This concentration of LCS-1.34 did not affect the number of apoptotic cells in the H3255 cultures; however, at a higher concentration (1 μM), the number increased 2-fold, from 9 to 19%. At this higher concentration, LCS-1.34 induced an approximately 6-fold increase in the number of apoptotic cells in H358 and H1975 cultures.
MAPK and PI 3-kinase pathways are inhibited by LCS-1.34
To gain additional insight into the mechanism by which the LCS-1-related compounds induced cell death, we analyzed the phosphorylation status of several signaling proteins that mediate proliferation and survival. We used a proteome profiler antibody array (R&D Systems) that allowed us to measure the phosphorylation status of 18 kinases, including three major MAPK family members (ERK, JNK and p38) and components of the PI 3-kinase pathway (AKT, GSK-3 and p70 S6 kinase). H1975 cells were treated for 24 h with 0.5 μM LCS-1.34. Whole cell-extracts were prepared and protein phosphorylation analyzed. Phosphorylation of ERK2, the three AKT isoforms, and p70 S6 kinase was reduced in cells that were treated with LCS-1.34 (). Similar results were obtained in H3255 cells (data not shown). LCS-1.34 did not, however, alter phosphorylation of any of the p38 or JNK isoforms (data not shown).
LCS-1.34 blocks phosphorylation of cytoplasmic components of signaling pathways but not phosphorylation of EGFR family members
EGFR family members are not direct targets of the phenyl pyridazinone compounds
Given the effect of LCS-1 and its derivatives on some cell lines with EGFR mutations, we asked if phosphorylation of EGFR or other members of the EGFR family (ERBB2 and ERBB3) was affected by these compounds using a human phospho-RTK antibody array. For these experiments, we used H3255 cells, which have relatively higher levels of phospho-EGFR. Treatment of cells for 24 h with 0.5 μM LCS-1.34 did not affect total phosphorylation of EGFR, ERBB2 or ERBB3 whereas erlotinib dramatically reduced phosphorylation of all three receptors (). In addition, we used a phosphorylation site-specific antiserum to show that treatment of H1975 cells with LCS-1.34 did not alter the phosphorylation of EGFR at Y1068 ().