Chemical inhibition of LSF induces a delay in mitotic progression with condensed but unaligned chromosomes
In asynchronous populations of hepatocellular carcinoma cells, the LSF small molecule inhibitors (FQIs) resulted in cells delayed with G2/M (“4n”) DNA content [16]. In order to characterize this cell cycle-related phenotype in detail, we utilized HeLa cells, which are similarly sensitive to FQI1 treatment and readily able to be synchronized for in depth cell cycle analysis. First, the dose- and time-dependence of this phenotype was examined, using cellular DNA content as a readout. Cells were synchronized at G1/S with a double thymidine block (Fig. 1a) and analyzed for cellular DNA content throughout the subsequent cell cycle. Upon release from the G1/S block, the population of vehicle-treated cells proceeded synchronously through S phase, with DNA content in between 2n and 4n levels (Fig. 1b, 4.5 h; see Additional file 2 for alignments of each individual peak distribution). The control cells then proceeded to mitosis and largely re-entered G1 again at 8.5 h in the experiment shown. At 0.9 and 1.8 μM, FQI1-treated cells were initially delayed in returning from G2/M to G1, remaining with 4n DNA content (Fig. 1b, 8.5 h; Additional file 2), an observation consistent with previous studies [16]. Following this delay, a mix of phenotypic outcomes was observed at 16 h after release from G1/S, with FQI1-treated cells having divided to a 2n DNA content, initiated cell death pathways (subG1 DNA content), or retained their duplicated DNA content. At the highest FQI1 concentration tested (3.6 μM), progression of cells away from 2n content appeared less complete (note skewing of the peak to a lower shoulder even at 8.5 h), and never fully reached 4n DNA content. Subsequently, a large fraction of the population converted to subG1 DNA content by 16 h post release from the G1/S block (Fig. 1b; Additional file 2).
Given limitations in interpretation of population-wide data, it was critical to analyze the phenotype(s) on a per cell basis upon inhibition of LSF. Using HeLa cells stably expressing fluorescently labeled histone H2B, chromosomal DNA was visualized by time-lapse microscopy as each cell passed through mitosis. Asynchronous H2B-YFP-expressing cells were treated with increasing concentrations of FQI1 and imaged immediately thereafter. At 1.8 μM (Fig. 1c) mitotic progression was delayed with condensed, unaligned chromosomes. Cells subsequently appeared to exit mitosis without proper chromosome segregation, resulting in a multinuclear (4n) G1 state (Fig. 1c) [16]. In contrast, cells treated with vehicle progressed through mitosis normally (Fig. 1c and d). Mitotic time (time from nuclear envelope breakdown (NEB) to anaphase) was dose-dependent, increasing with increasing concentrations of FQI1 (Fig. 1d). At the lower concentrations of FQI1, cells exited mitosis aberrantly after the mitotic delay (Fig. 1c, d and e). However for cells treated with 3.6 μM FQI1, mitotic time for those cells that entered mitosis could not be determined using this experimental design (Fig. 1e). Cells would arrest with condensed but unaligned chromosomes but never exited mitosis during the course of the time lapse experiment, either normally or aberrantly, whether entering mitosis at the beginning or end of the experimental window of 11 h (Addition File 3, Fig. S1A).
In some cell lines, LSF is necessary for upregulation of thymidylate synthase expression and therefore efficient transition through S phase [24]. Since S phase defects can ultimately lead to mitotic defects, DNA damage was monitored by measuring phosphorylated H2AX (γ-H2AX) chromosomal foci. Cells treated with 1.8 μM FQI1 or vehicle during synchronization were analyzed for γ-H2AX by immunofluorescence 8 h after release from the G1/S block (Fig. 1f). UV-irradiated cells, a positive control, demonstrated extensive γ-H2AX staining. In contrast, only comparable, low levels of phosphorylated H2AX were visualized in both the vehicle- and FQI1-treated cells, consistent with low levels of DNA damage known to occur in cancer cells [25]. This was consistent with our previous studies suggesting that adequate thymidylate synthase expression in tumor cells was likely achieved even with reduction in LSF activity [15, 26].
Overall, upon treatment of cells with FQI1, we observed defects in chromosome alignment and segregation, resulting in mitotic delay and multi-nucleation. The lack of elevated DNA damage signals supports the hypothesis that FQI1-mediated mitotic defects are due to a direct requirement for LSF in regulating proper progression through mitosis, and in particular in progressing to metaphase, in which the condensed chromosomes are fully aligned.
LSF small molecule inhibition during cell synchronization reduced expression of mitotic regulators
The lack of chromosomal alignment is not a phenotype expected to be caused by upregulation of cyclin B expression, which was previously suggested from experiments in asynchronous cells to be a major cause of the FQI1-mediated mitotic delay [16]. Thus, we investigated the effects of FQI1 treatment on the expression of a number of mitotic regulators, using synchronized cells to more carefully assay cell cycle expression. HeLa cells were again synchronized with a double thymidine block (Fig. 2a). Cell populations treated in this manner pass through the subsequent cell cycle in a wave that is generally, although not precisely, synchronous. Only a short window of time is spent in mitosis, on average 50 min for these HeLa cells (Fig. 1d). Initial analyses when harvesting for flow cytometry for DNA content indicated that around 8.5 h the cells had just returned to G1 phase, since some were still in mitosis (4n DNA content, Fig. 1b; Additional file 2). By fluorescence microscopy, a range of mitotic figures were observed in separate experiments harvested either at 8 or 8.5 h (Additional File 3, Fig. S1B). Therefore, for subsequent analyses of RNA or protein levels in mitotic cells, cells were monitored visually starting just prior to 8 h, and harvested only once the control population was observed to significantly round up on the dish, indicating mitotic entry. RNA levels were measured initially in vehicle-treated cells at the G1/S border (0 h) and in mitosis (approximately 8 h). Results are shown for the mitotic regulators Aurora kinase B (AURKB) and Cyclin Division Cycle 20 (CDC20) [27, 28], as expression levels of these genes were altered by FQI1 (see below). In the untreated, control cells, mitotic expression of CDC20 was elevated approximately 6-fold compared to RNA levels at G1/S (Additional File 3, Fig. S1C), as expected. However, AURKB RNA levels in vehicle-treated cells increased only 1.2 fold in mitosis, consistent with dysregulated expression in these cancer cells. Incubation with 1.8 μM FQI1 during the synchronization protocol resulted in reduction of both AURKB and CDC20 RNA levels compared to the control cells 8 h post release (Fig. 2b). For analysis of protein levels, cells were also harvested in mitosis (around 8 h), and levels were quantified after immunoblotting using the LI-COR Odyssey instrumentation, which provides direct measurements of fluorescence levels over a 4-log intensity range. In Fig. 2c and d are shown both snapshots of the immunoblot images and the independent, direct quantitative measurements, respectively. Consistent with transcript reduction, AURKB and CDC20 protein levels were also reduced in a dose-dependent manner at the time control cells were visually entering mitosis (Fig. 2c and d), whereas LSF protein levels were unchanged, as expected (Fig. 2c). The impact of the downregulation of AURKB was tested by monitoring phosphorylation of an AURKB substrate. Phosphorylation of Histone 3 on Serine 10 [29] was reduced by FQI1 in a dose-dependent manner (Fig. 2c and d).
In stark contrast to the previous results in asynchronous cells [16], in the synchronized cells treated with up to 1.8 μM FQI1, Cyclin B protein levels were similar to those in control cells, and certainly not increased (Fig. 2c), despite obvious delays in mitotic progression at such concentrations (Fig. 1b-d). At the higher concentrations of FQI1 during the synchronization procedure, Cyclin B levels were actually downregulated (Fig. 2c). In order to investigate downregulation of cyclin B further, we analyzed levels of Cyclin B in synchronized cells with vehicle versus a high concentration of FQI1 (5 μM) when treated only following, rather than during, the synchronization process (Additional File 3, Fig. S1D). In this case, Cyclin B levels peaked around mitotic entry at similar levels in both control and FQI1-treated cells. In control cells, Cyclin B levels declined as cells continued through the cell cycle, as expected, but Cyclin B remained elevated in the FQI1-treated cells, which remained arrested in mitosis during this protocol [16]. First, these findings demonstrate that elevated levels of cyclin B observed in asynchronous cell populations are the result, not the cause of the mitotic arrest induced by FQI1, unlike what was previously suggested [16]. Furthermore, these results suggest that the decrease in Cyclin B levels when 3.6 μM FQI1 is present during cell synchronization (Fig. 2c) must result from FQI1 effects during previous cell cycles. In particular, because cyclin B is required for mitotic entry, lower levels of cyclin B suggested that cells treated with higher concentrations of FQI1 during synchronization were not efficiently proceeding into the final mitosis after release from the G1/S block (see also below). This is consistent with the bulk cellular DNA profiling curves (Fig. 1b; Additional file 2), since this analysis cannot distinguish between “2n” as early G1 versus G1/S or “4n” as G2 versus M. This complicates straightforward interpretation of the bulk expression results, with ambiguity as to whether decreased levels of AURKB and/or CDC20 were the cause of the mitotic cell cycle defects, or also the consequence of cell cycle defects resulting from lower LSF activity during the first passage through mitosis in the synchronization procedure.
For a stringent examination of whether diminished AURKB and CDC20 gene expression resulted from lack of cell cycle progression of LSF inhibited cells or from diminished expression of these genes in mitosis in the presence of FQI1, we analyzed RNA in synchronized, LSF-inhibited cells only from cells demonstrably in mitosis, isolated by standard mitotic shakeoff methodology. A reproducible decrease in CDC20 (Additional File 3, Fig S2B), but not AURKB (Additional File 3, Fig. S2A), RNA was observed in this experiment. We also sought to identify candidate LSF target genes by identifying binding sites for LSF near the genes. Given the lack of a sufficiently robust antibody against LSF for chromatin immunoprecipitation (ChIP), a stable HEK cell line inducibly expressing HA-tagged LSF [15] was used for the ChIP-sequencing analysis. Gene ontology analysis of genes whose transcription initiation sites were located within 20 kb of the ChIP peaks only revealed a small number of statistically significant functional or pathway categories, with a focus on categories involving nucleosomes, chromatin, transcriptional regulation, and splicing regulation (Additional File 3, Table S1), none of which seemed particularly revealing with regards to the phenotypes observed here. Multiple HA-LSF binding peaks were observed around the AURKB gene (Additional File 3, Fig. S2C), and binding of LSF was validated both at the AURKB promoter and around 3000 bp upstream of the transcription start site by quantitative PCR (Additional File 3, Fig. S2D). In contrast, no HA-LSF binding peaks were observed within 20 kb of the CDC20 gene. Taken in combination, whether LSF activates AURKB expression in these, or other, cells remains unresolved. The mitotic shakeoff experiment does suggest that LSF regulates CDC20 expression, either from distant binding sites, or indirectly. Global gene expression data from cells treated with FQI1 only between G1/S and mitosis did not identify dysregulation of RNA encoding any other mitotic regulators [30].
Despite not pinpointing mitotic genes directly transcriptionally regulated by LSF, these results did provide molecular biomarkers in this synchronized cell system for responsiveness to the LSF inhibitor FQI1.
RNAi mediated knockdown of LSF phenocopies inhibition of LSF with the small molecule inhibitor FQI1
Specificity of small molecule inhibitors to their intended target is a key requirement so that biological consequences of inhibitor effects can be mechanistically attributed to the target of interest. Knowledge of specificity is of even more importance in developing such inhibitors for use in the clinics. FQI1 inhibits LSF DNA-binding and protein-binding activities, whereas it does not impact activity of a number of other transcription factors, both with disparate and similar structural domains [15, 31]. However, in order to demonstrate that the overall cellular consequences of FQI1 treatment were specific consequences due to inhibiting LSF, a direct comparison with specific removal of LSF was required. Although LSF has a long half-life, of approximately 24 h [32], we identified an siRNA that resulted in robust and durable knockdown of LSF RNA and protein, at least for 48–96 h (Fig. 3b-d, Additional File 3, Fig. S3A-B). In addition, since certain siRNAs can cause nonspecific reduction in mRNA encoding MAD2 [33], which controls the spindle assembly checkpoint and therefore the rate of progression through mitosis, we verified that the selected siRNA targeting LSF did not inadvertently reduce MAD2L1 RNA levels (Additional File 3, Fig. S3C).
In order to compare downstream biomarkers from FQI1 and LSF siRNA treatments, AURKB or CDC20 RNA levels were measured following RNAi mediated knockdown of either LSF or a non-expressed control. Cells were transfected with siRNAs, to initiate protein knockdown, 30 h prior to synchronization. RNA and protein expression were analyzed at two time points - when control cells were arrested at G1/S (0 h) following the synchronization protocol and when these cells were largely in mitosis after release from the block (approximately 8 h) (Fig. 3a, Additional File 3, Fig. S4A, Additional file 2). For ease of comparison, RNA levels were plotted relative to the level in the control siRNA sample at each time point. At 20 nM siRNA, significant knockdown of LSF-encoding RNA (TFCP2, Fig. 3c) was achieved over this time course. LSF protein levels were also decreased significantly at all three concentrations of LSF siRNA, but to the greatest degree with 20 nM LSF siRNA (top immunoblot, Fig. 3d). Consistent with the results generated with the LSF small molecule inhibitor at 8 h after G1/S release (Fig. 2b), AURKB and CDC20 RNA levels were significantly reduced upon knockdown of LSF (Fig. 3c). Immunoblotting of lysates harvested at the time of mitotic entry of the control cells (approximately 8 h) confirmed a dose-dependent reduction in AURKB and CDC20 protein levels after siRNA-mediated knockdown of LSF (Fig. 3d and e), consistent with the findings upon inhibition of LSF with FQI1 (Fig. 2c and d). As expected, phosphorylation of AURKB substrate Serine 10 of histone H3 [29, 34] was reduced (Fig. 3d). Finally, as with FQI1 treatments, Cyclin B levels were also reduced in a dose-dependent manner. With LSF siRNA treatment, all mitotic protein levels were decreased even at the lowest amount of the siRNA, including that of cyclin B, reflective of the DNA profiling data in which a significant fraction of cells never proceeded from 2n DNA content after release from the G1/S block, unlike with the control siRNA (Additional File 3, Fig. S4A; Additional file 2). This effect was similar, although more pronounced, than that observed in experiments using FQI1. Overall, LSF siRNA phenocopied the molecular consequences of FQI1 on protein expression, whether due to direct transcriptional effects, and/or consequences of cell cycle dysregulation.
To determine whether LSF knockdown resulted in similar mitotic phenotypes to those observed with FQI1, synchronized YFP-H2B-expressing HeLa cells were transfected with siRNAs targeting LSF or a non-expressed control. A single thymidine block protocol was sufficient for synchronization (Fig. 4a), as mitotic progression is viewed on a cell-by-cell basis. This protocol results in a partial cell cycle synchronization, with 84% of the control cells rounding up and undergoing mitosis in a smaller window of time following release from the block (Fig. 4d; Additional file 1; Additional File 3, Fig. S4B). Surprisingly a larger fraction of the LSF siRNA-treated cells visually rounded up than underwent DNA condensation subsequent to release from the G1/S block (Fig. 4d). Representative time-lapse images of cells treated with the highest concentration (20 nM) of either LSF targeting siRNA or control siRNA highlight dramatic changes in mitotic progression. Control cells exhibited progression through normal mitotic phases within an hour, as expected (Fig. 4b and c). However, cells with diminished LSF levels exhibited an extensive delay with condensed chromosomes that never achieved stable alignment, generally followed by defective cellular division and multinucleation (Fig. 4b). In addition, some cells remained in mitosis with condensed chromosomes throughout the entire time lapse analysis. Quantitation documented that mitotic time was dramatically increased when LSF levels were reduced (Fig. 4c). We note that 5 and 10 nM of LSF siRNA resulted in slightly longer times for mitotic progression than did 20 nM, which is likely due to the inability of a number of cells at the 20 nM LSF siRNA treatment (the most perturbed cells) to fully progress through mitosis. When siRNA-transfected cells were imaged by time lapse microscopy after a thymidine block, there was an inverse correlation between higher levels of LSF knockdown and the number of cells undergoing nuclear envelop breakdown (Additional File 3, Fig. S4B).
By histone H2B fluorescence from the time lapse data, the most striking mitotic outcome for individual cells treated with LSF siRNA after the extended delay in mitosis appeared to be mitotic slippage (Fig. 4b). The mitotic slippage outcomes were confirmed by immunofluorescence of synchronized cell populations, co-stained for DNA and α-tubulin (Additional File 3, Figs. S5A-B). Quantitation of these immunofluorescence data demonstrated significant increases in cells with condensed, but nonaligned chromosomes, incomplete cytokinesis, and multinucleation upon LSF knockdown. These phenotypes mimicked those observed with FQI1 treatment using a parallel treatment and synchronization protocol, and therefore analyzed at 40 h after FQI1 treatment was initiated (Additional File 3, Fig. S5C-D). Using this synchronization protocol with FQI1, outcomes from mitotic slippage reflecting the eventual degradation of cyclin B during mitotic arrest could be assessed, unlike in the time-lapse experiment in which the maximum time of exposure to FQI1 was 11 h. Finally, treatments with both FQI1 and LSF siRNA yielded mitotic cells with cellular protrusions (Additional File 3, Fig. S5E).
Also consistent with the results from FQI1 treatment, knockdown of LSF did not induce phosphorylated H2AX (γ-H2AX) foci, as monitored the beginning of mitosis in a synchronized cell population (Fig. 4e). This result suggests that the effects of inhibiting LSF on mitotic progression are not due to defects induced indirectly in S phase.
LSF has a widely expressed paralog, LBP1A [35], whose activity is also inhibited by FQI1 (T. Grant, unpublished observations). Unlike LSF, however, LBP1A has not yet been implicated in cancer [1, 2], which is consistent with our results that knockdown of LSF alone caused the same molecular and phenotypic outcomes as those caused by FQI1. Nonetheless, to determine the potential contribution of LBP1A to FQI1 outcomes, we investigated the mitotic phenotypes upon LBP1A siRNA treatment. An identified siRNA resulted in robust and durable knockdown of LBP1A-encoding RNA (Additional File 3, Fig. S6A), with no reduction in MAD2L1 RNA levels (Additional File 3, Fig. S6B). Using two distinct assays, knockdown of LBP1A did reduce overall cell proliferation in the cell population to a limited extent, but much less so than did knockdown of LSF (Fig. 5a-b). Despite this slight decrease in cell number over time relative to the control, LBP1A knockdown did not lead to any significant drop in cell viability between 72 and 96 h post-transfection, unlike the dramatic drop in viability evident upon LSF knockdown at this time interval (Fig. 5b). Furthermore, LBP1A knockdown did not observably inhibit mitotic progression as measured by cellular DNA profiling (Additional File 3, Fig. S6C). As a more sensitive assay, time-lapse microscopy did detect a subtle increase in average mitotic time upon LBP1A knockdown (1.5–fold, Fig. 5c-d), but with only a small percentage of the cells being demonstrably affected. This limited extent of mitotic delay is consistent with the overall limited reduction in cell proliferation upon LBP1A knockdown. Notably, no abnormal mitotic phenotypes were observed either by time-lapse or immunofluorescent microscopy, however. Given the minimal cellular consequences upon inhibiting LBP1A, we conclude that inhibition of LSF activity is what drives the dramatic FQI1-mediated mitotic defects.
Induction of cellular senescence following inhibition of LSF
In addition to mitotic delay resulting from LSF inhibition, some cells undergoing synchronization while inhibiting LSF were arrested at other points in the cell cycle, as suggested above by the reduction in cyclin B expression. Cellular DNA profiling of both 3.6 μM FQI1-treated and LSF siRNA-treated cells being synchronized with a double thymidine block captured cells that no longer progressed from the 2n state through S phase upon release from the G1/S block (Fig. 1b, Additional File 3, Fig. S4A; Additional file 2), and time-lapse microscopy showed that a considerable fraction of the cells treated with LSF siRNA during a thymidine block never underwent nuclear envelope breakdown during 10–12 h after release from the G1/S block (Additional File 3, Fig. S4B). Mitotic defects, caused by multiple distinct insults, can lead to senescence after G1 re-entry with either 2n or 4n DNA content [36, 37]. Thus, we hypothesized that mitotic defects from decreasing LSF levels or activity during previous cell divisions resulted in senescence. To test this hypothesis, cells were synchronized as before by a double thymidine block in the presence of FQI1, or LSF siRNA, and analyzed for senescence by monitoring β-galactosidase activity at low pH [38] at a time point when control cells entered mitosis. Both reduction in LSF levels and inhibition of LSF activity resulted in significantly greater numbers of β-galactosidase-positive cells (stained blue) compared to the respective controls (Fig. 6a-b). Although the absolute degree of senescence varied between experiments, and was generally more pronounced with LSF siRNA treatment, overall there was a consistent 3- to 5-fold increase in senescent cells with increasing amounts of LSF inhibition. Only treatment with 0.9 μM FQI1 was not sufficient to induce senescence (Fig. 6c-d). These data show that inhibition of LSF can result in senescence of cancer cells, and support the hypothesis that reduced LSF levels or activity during previous cell cycle(s) can predispose cells to senescence.