ABSTRACT
The LMO2/LDB1 macromolecular complex is critical in hematopoietic stem and progenitor cell specification and in the development of acute leukemia. This complex is comprised of core subunits of LMO2 and LDB1 as well as single-stranded DNA-binding protein (SSBP) cofactors and DNA-binding basic helix-loop-helix (bHLH) and GATA transcription factors. We analyzed the steady-state abundance and kinetic stability of LMO2 and its partners via Halo protein tagging in conjunction with variant proteins deficient in binding their respective direct protein partners. We discovered a hierarchy of protein stabilities (with half-lives in descending order) as follows: LDB1 > SSBP > LMO2 > TAL1. Importantly, LDB1 is a remarkably stable protein that confers enhanced stability upon direct and indirect partners, thereby nucleating the formation of the multisubunit protein complex. The data imply that free subunits are more rapidly degraded than those incorporated within the LMO2/LDB1 complex. Our studies provided significant insights into LMO2/LDB1 macromolecular protein complex assembly and stability, which has implications for understanding its role in blood cell formation and for therapeutically targeting this complex in human leukemias.
INTRODUCTION
In hematopoiesis, lineage-specific transcription factors control specification of the hematopoietic stem cell (HSC) for diverse cell types. At the top of this developmental hierarchy are 9 factors that directly affect the HSC itself: BMI1, RUNX1, GATA2, LMO2, TAL1, LDB1, MLL, GFI1, and ETV6 (1). These master regulators are conserved among all vertebrates and have been experimentally characterized in mice, zebrafish, and humans (2). Knockout of any one of the genes encoding these factors causes the loss of all forms of hematopoiesis, both embryonic and adult, by perturbing the creation, survival, or self-renewal of primitive and definitive HSCs. Examining this gene list, there are three emerging themes. First, the factors are part of a transcriptional network with autoregulation and interregulation (3). Second, the factors are frequently coopted in human leukemias by various genetic mechanisms such as chromosomal translocation (4). Third, for our study, all the factors function as part of multisubunit protein complexes. Four of the factors listed above act in concert within a macromolecular complex, the LMO2/LDB1/TAL1/GATA2 (or the LDB1/LMO2) protein complex. There are diverse data supporting the idea that these proteins are bound together, including coimmunoprecipitation (co-IP), copurification followed by mass spectrometry, electrophoretic mobility shift assays, and cooccupancy at target genes by chromatin immunoprecipitation (5–9).
Assembly of the LDB1/LMO2 complex depends upon specific interactions between LMO2 and class II basic helix-loop-helix (bHLH) proteins, LMO2 and GATA factors, and LMO2 and LDB1. There are multiple bHLH and GATA paralogs capable of binding LMO2, so multiple versions of the LMO2-associated complex exist depending upon the expression of the subunits. LMO2 is an 18-kDa protein with two zinc-binding LIM domains, LIM1 and LIM2. LIM1 folds to create an interface for binding class II bHLH proteins such as TAL1 and LYL1 (10). LIM2 has an interface that binds GATA factors 1 to 3. A portion of LIM1 also serves as an interface for binding to the LIM interaction domain (LID) of LDB1. LDB1 has a self-association domain through which LDB1 may dimerize or multimerize (11). The class II bHLH proteins heterodimerize with class I bHLH proteins such as E2.2, E12, E47, and HEB (12). The bHLH proteins and GATA proteins can be part of the same complex allowing the LDB1/LMO2 complex to bind adjacent E boxes and GATA sites (8, 9, 13, 14). Such motifs bound by LMO2/LDB1 complexes have been described in erythroid progenitor cells at various gene targets, including the beta globin gene promoters and the locus control region (LCR) (5, 14, 15). The self-association domain of LDB1 mediates looping and proximity between the beta globin LCR and beta globin proximal promoters, a seminal example of enhancer-promoter communication (11, 16–18).
Several iterations of the LDB1/LMO2 complexes are drivers in leukemia. In fact, LMO2 and TAL1 were originally cloned from chromosomal translocations in T-cell acute lymphoblastic leukemia (T-ALL) (19). LMO2 was also the target of insertional activation in gammaretroviral gene therapy-induced T-ALL (20, 21). Mouse modeling and the characterization of the LMO2-associated complexes have been highly informative in dissecting the pathogenesis of LMO2-induced T-ALL, underscoring the role for specific bHLH and GATA factors as requisite cooperating drivers (22–25). We recently confirmed by purification of FLAG-LDB1 and mass spectrometry that the LMO2/LDB1 complex in T-ALL closely resembles the complex hypothesized to function in normal HSCs (7).
Regardless of the variation in bHLH or GATA factors or the cofactors that these transcription factors may recruit, the subunits of LMO2 and LDB1 are constant. We probed the LMO2/LDB1 interaction and discovered a discrete motif within the LDB1 LID that was essential for LMO2 binding. We consistently observed an increase in steady-state abundance of LMO2 with coexpression of LDB1 and a decrease in abundance with the coexpression of LDB1ΔLID (7). This effect on LMO2 steady-state protein was observed in multiple leukemic cell types. To more closely analyze the effects on protein stability, we sought to understand the kinetics of turnover of LMO2 and its partner proteins. To this end, we devised a pulse-chase technique through the use of multiplexed lentiviral expression of Halo-tagged proteins (26). We discovered that there is a hierarchy of protein turnover for the subunits of the complex, with LDB1 being the most stable protein. Furthermore, we discovered that every subunit, including both direct and indirect binding partners of LDB1, was stabilized by LDB1. These findings have remarkable implications for the assembly of this important macromolecular complex and underscore identification of LDB1 as the major core subunit that could be targeted in leukemias.
RESULTS
Halo tagging of LMO2 and its binding to LDB1.Analysis of LMO2 turnover by quantitative Western blotting after cycloheximide treatment showed half-lives in the range of 6 to 10 h for endogenous LMO2 (in K562, MOLT4, and LOUCY cells) and exogenous LMO2 in Jurkat leukemia cells (Fig. 1A and B). LMO2 turnover was markedly prolonged in the presence of LDB1, and LDB1’s decay was not observed within the same time frame. We were at the detection limits of the cycloheximide chase assay, where toxicity was a confounding issue. Therefore, we developed an alternative approach to analyze LMO2 protein turnover in live cells without metabolic perturbation and without toxins. We synthesized recombinant LMO2 with an amino-terminal Halo tag (26). We expressed Halo-LMO2 in K562 cells, where the recombinant protein had enhanced steady-state abundance with LDB1 coexpression (Fig. 1C, lanes 9 and 10), implying direct binding with LDB1. This was confirmed by coimmunoprecipitation (co-IP) of Halo-LMO2 with FLAG-LDB1 (Fig. 1D, lane 9). Previously, we established R320LITR within LDB1 as the key interacting residues and found that single-residue substitutions (i.e., I322A) within RLITR reduced LMO2 binding to LDB1 (7). Halo-LMO2 was bound and recovered with FLAG-LDB1, but recovery was reduced by FLAG-LDB1ΔLID or LDB1(R320L→AAAA). We have previously found that these mutant LDB1 proteins could still dimerize and pull down full-length LDB1/LMO2 complexes (Fig. 1C and D, lanes 9 to 14). Nevertheless, the specific pattern in which Halo-LMO2 interacted with LDB1 matched that of the wild-type (WT) protein.
Limitations of conventional methods for determining LMO2 stability. Initial characterization of Halo-tagged LMO2 was performed. (A) Conventional cycloheximide chase experiment. Whole-cell lysates were prepared from the time points shown, subjected to 4% to 15% gradient SDS-PAGE, transferred, and blotted with anti-HA antibody (Ab). The top panel shows Jurkat cells lentivirally expressing LMO2-HA alone, and the bottom panel shows LMO2-HA coexpression with FLAG-LDB1. (B) Graph showing half-lives (t1/2) of LMO2-HA representing the cycloheximide chase experiment data shown in panel A compared with t1/2 data derived from cycloheximide chase in K562 cells following endogenous LMO2 (using anti-LMO2 monoclonal antibody). (C) Immunoblot of whole-cell lysates prepared from K562 cells transduced with empty vector (EGFP-puro), FLAG-LDB1 WT, or mutant LDB1 proteins: ΔLID, RLIT→AAAA, and I322A. The last 3 are deficient in binding LMO2. Cells coexpressed empty vector (EBFPII-Hygro) (lanes 3 to 8) or Halo-LMO2 (lanes 9 to 14). Immunoblots are shown with anti-LDB1 polyclonal Ab, anti-FLAG, anti-LMO2, anti-Halo, anti-TAL1, anti-GFP (as a lentiviral expression control since all transductions were performed by the use of lentiviral vectors), and anti-VCP (valosin-containing protein) (loading control). (D) Cell lysates from the experiment performed as described for panel C were subjected to immunoprecipitation with anti-FLAG monoclonal antibody, eluted with sample buffer, and immunoblotted as described above.
Quantification of lentivirally expressed Halo proteins.We developed novel lentiviral vectors with unique fluorescent markers and antibiotic resistance genes to enable multiplexed expression of LMO2 and its partners (see Fig. S1 in the supplemental material). Since overexpression could influence the interpretation of co-IP and protein turnover studies, we extensively analyzed the protein levels achieved by these lentiviral vectors. First, we compared the protein levels of lentivirally expressed Halo-LMO2 with those of endogenous LMO2 in various cell lines by the use of an anti-LMO2 monoclonal antibody (Ab) which recognized both proteins. Immunoblotting of equivalently loaded protein lysates showed that Halo-LMO2 was expressed at or below levels of endogenous LMO2 in multiple leukemic cell lines (Fig. 2A). We extended these studies to other proteins expressed by our lentiviral system. We cloned Halo-CTCF and Halo-SOX2 into our lentiviral vectors and compared their expression levels in transduced Jurkat cells to the levels seen with a U2OS cell line in which the Halo tag was knocked into the CTCF gene (27). As shown in Fig. 2B, lentivirally expressed Halo-CTCF at our usual multiplicity of infection (MOI) of 1 to 2 resulted in protein levels in Jurkat cells equivalent to those seen with Halo-CTCF expressed from its endogenous promoter in U2OS cells (Fig. 2B, compare lane 1 to lanes 2 to 4). Steadily increasing the MOI above our usual working concentrations caused an increase in lentiviral expression of Halo-CTCF by 2-fold above the endogenous levels (Fig. 2B, compare lane 1 to lanes 5 to 9). These proteins were also compared by labeling the cells with fluorescent Halo ligand, R110, and visualizing protein lysates after SDS-PAGE (i.e., “in-gel” fluorescence [Fig. 2B, middle panel]). Similarly to the results of the immunoblotting performed with anti-CTCF (top panel), Halo protein fluorescence showed that the lentiviral expression seen with Halo-CTCF was equivalent to the Halo-CTCF expression from the knock-in (Fig. 2B, compare lanes 1 and 2). The coexpression of Halo with Halo-CTCF decreased the in-gel fluorescence of Halo-CTCF (Fig. 2B, middle panel, lanes 2 and 3). Lentivirally expressed Halo-SOX2 protein was also present at levels similar to Halo-CTCF. As proposed previously by Cattoglio et al. (28), the U2OS knock-in line was applicable for relative quantifications of other Halo-tagged proteins by flow cytometry. Thus, the mean fluorescence intensities of the stable Halo protein-expressing cell lines were compared, and the results showed that lentivirally expressed Halo-LMO2, Halo-CTCF, and Halo-SOX2 were expressed at levels lower than that of Halo-CTCF emanating from the knock-in U2OS line (Fig. 2C). Cattoglio et al. had performed absolute quantification of Halo-CTCF in U2OS by several techniques, allowing us to derive the copy number of our lentivirally expressed proteins in Jurkat cells (Fig. 2D). Confocal microscopy showed that most (95%) of these Halo-LMO2 molecules were localized in the nucleus (Fig. 2E; see also Fig. S2). In summary, on the basis of immunoblotting, in-gel fluorescence after SDS-PAGE, and flow cytometry, we concluded that the level of expression of our recombinant Halo-tagged proteins was within a physiological range.
Halo-tagged proteins expressed from recombinant lentiviruses have physiological abundance and localization and can be precisely quantified in lysates or live cells. (A) Cell lysates were prepared from untransduced Jurkat cells; Jurkat cells transduced with Halo-LMO2; untransduced K562 cells; K562 cells transduced with Halo-LMO2; and KOPT-K1, LOUCY, and U937 cells. Lysates were subjected to 4% to 15% SDS-PAGE, transferred, and blotted with anti-LDB1, anti-LMO2 monoclonal antibody, anti-Halo, anti-TAL1, and anti-VCP (gel loading control). Molecular weight standards were run on the same gel and are shown at the right. (B) The top panel shows an immunoblot of lysates prepared from U2OS with a Halo knock-in at CTCF (lane 1) or lysates from Jurkat cells transduced with the constructs shown in the grid above (lanes 2 to 14). Empty vector control is shown as EBFPII-Hygro (lane 2). Lanes 2 to 9 show lysates from Jurkat cells lentivirally expressing Halo-CTCF at increasing MOIs. Lanes 10 to 14 show lysates from Jurkat cells lentivirally expressing Halo-SOX2 at increasing MOIs. The top panel shows gradient SDS-PAGE, transfer, and blotting with anti-CTCF antibody. The middle panel shows direct in-gel Halo fluorescence. Live cells prepared as described for panel A were labeled with cell-permeative fluorescent ligand R110, lysed, and subjected to gradient SDS-PAGE. The gel was visualized for green fluorescence as described in Materials and Methods. (C) Live Jurkat cells expressing Halo-tagged proteins were labeled with R110 Halo ligand, washed, and subjected to flow cytometry. Histograms show FITC fluorescence of the various cells in comparison to the labeled U2OS cells, which have a Halo tag knocked into the CTCF gene. (D) Table showing calculated copy numbers of Halo-tagged proteins based on the absolute values derived from the Halo knock-in cell line. Lentiviral Halo CTCF and Halo SOX2 were quantified from cell populations infected at intermediate and low MOIs, respectively. (E) Confocal microscopy of Jurkat cells lentivirally expressing Halo-LMO2 and labeled with Halo ligand, R110, and nuclear stain (Syto 17 Red). The right panel shows a merged image. Voxel quantification showed that 95% of Halo-LMO2 was nuclear.
LMO2 turnover is mediated by a ubiquitin-proteasomal system and is inhibited by LDB1.To analyze LMO2 protein turnover, we pulsed cells expressing Halo-LMO2 or Halo alone with the Halo ligand R110 and analyzed cellular fluorescence (i.e., chase) through the fluorescein isothiocyanate (FITC) channel (Fig. 3A). Immunoblotting of the transduced cells showed expression of hemagglutinin-LDB1 (HA-LDB1), HA-LDB1ΔLID, Halo, and Halo-LMO2 in Jurkat cells (Fig. 3B). A plot of the decay of positive cells after a 90-min pulse of R110 for untagged Halo protein and for Halo-LMO2 in the presence or absence of bortezomib, a specific 26S proteasomal inhibitor used in proteomic analysis of ubiquitinated moieties and also currently used to treat T-ALL (29, 30), is shown in Fig. 3C. Bortezomib was tested with or without coexpression of HA-LDB1 or HA-LDB1ΔLID, which cannot bind LMO2. The curves fit a typical first-order exponential decay, allowing calculation of half-lives (t1/2) as shown in Fig. 3D. Untagged Halo protein showed very slow protein turnover (Fig. 3C, black curve), whereas Halo-LMO2 had a t1/2 value of 6.6 h, similar to the t1/2 calculated from cycloheximide experiments. Coexpression of HA-LDB1 increased the Halo-LMO2 t1/2 to 20.6 h (P = 1.12E−5). Similarly, bortezomib increased the Halo-LMO2 t1/2 to 20.2 h. Halo-LMO2 was degraded faster with coexpression of HA-LDB1ΔLID (t1/2 = 4.0 h, P = 1.26E−3). In summary, the presence of LDB1 stabilized LMO2 as measured by Halo pulse-chase analysis. Halo-LMO2 turnover was reduced by bortezomib, implicating the ubiquitin-proteasomal pathway as the mechanism of degradation. Also, LDB1ΔLID, which is deficient in LMO2 binding but capable of homodimerization with endogenous full-length LDB1, increased the degradation of LMO2, a dominant-negative effect previously observed in multiple leukemic cell lines (7).
Halo-LMO2 pulse-chase analysis for monitoring turnover rates. (A) Schematic showing the procedure for Halo pulse-chase analysis of protein turnover. Cells are transduced with lentiviral vector expressing a Halo-tagged protein, typically, for this study, vector-expressed EBFPII-Hygro. Cells were placed under antibiotic selection, and Halo protein expression was verified at this stage by immunoblotting. Positive cells were pulsed with Halo ligand R110. Cells were 100% FITC positive (FITC+) at this stage. Flow cytometry was used to flow the FITC+ population. (B) Jurkat cells were transduced with empty vector (EBFPII-Hygro), Halo, or Halo-LMO2 along with HA-LDB1 or HA-LDB1ΔLID. Lysates were prepared and immunoblotted as described above. Blots shown are from experiments performed with anti-LDB1, anti-HA, anti-LMO2, anti-Halo, anti-GFP (expression control), and anti-VCP (gel loading control). (C) Graph showing decay of fluorescent cells over time. y-axis data show the percentages of positive cells found at t0. x-axis data represent hours elapsed during chase. These were all Jurkat cells expressing Halo, Halo-LMO2, Halo-LMO2 plus HA-LDB1, Halo-LMO2 plus bortezomib, and Halo-LMO2 plus HA-LDB1ΔLID. Points and error bars represent results from three independent experiments. (D) Bar graph showing a plot of Halo-LMO2 t1/2 derived from Halo pulse-chase analysis. Means and errors represent results from three independent experiments. P values are from two-tailed paired t test comparisons with data represented by the yellow bar (Halo-LMO2 with empty vector cotransduction).
Specific LMO2 lysines required for stabilization are also critical for binding to LDB1.The stabilization of LMO2 by bortezomib implicated the ubiquitin proteasomal system in LMO2 degradation. To identify the lysine residues within LMO2 that are potential sites for ubiquitination, we converted the 10 lysine codons in the cDNA to encode arginine. Unexpectedly, the level of lysineless mutant Halo-LMO2 [denoted K(0)] was not substantially increased at steady state (Fig. 4C) and it showed significantly faster turnover than the LMO2 wild type (WT) (t1/2 = 4.0 h versus 6.2 h (P = 1.06E−3) (Fig. 4D). We discovered that LMO2 K(0) was compromised in binding LDB1 as evidenced by reduced coimmunoprecipitation by LDB1(I322A) (Fig. S3). Our previous work had shown that there was a hydrophobic pocket within LMO2 that could accommodate I322 of LDB1 (Fig. 4A). We previously identified two leucines forming this pocket that were essential for stabilizing LMO2, a result that was confirmed by Halo pulse-chase [see Halo-LMO2Halo-LMO2 (L64A, L71A) in Fig. 4]. We noted there were two lysines, K74 and K78, in proximity to the LMO2 hydrophobic binding pocket interfacing with LDB1 R320LITR. A variant protein with only these two key lysines converted and the remaining 8 lysines retained showed significantly faster turnover, t1/2 = 3.9 h versus the t1/2 of Halo-LMO2 K(0) (Fig. 4C and D) (P = 1.76E−3). We also tested the reciprocal variant, where we left K74 and K78 intact and converted the remaining 8 lysines to arginine. As shown in Fig. 4D, this LMO2 variant, Halo-LMO2 K(0)(K74, K78), had a measured t1/2 of 5.5 h, representing a statistically insignificant (P = 0.107) difference from the measured t1/2 of Halo-LMO2 WT. We then tested single substitutions at K74 and K78. Halo-LMO2 K(0)(K74) had a measured t1/2 of 4.8 h, representing a significant (P = 7.28E−3) reduction compared to WT Halo-LMO2 whereas the t1/2 of Halo-LMO2 K(0)(K78), t1/2 = 5.1 h, was not significantly different (P = 0.09) (Fig. 4D). The LMO2 variant exhibited small changes in t1/2, but the changes were consistent over replicate experiments. Intriguingly, K74 was conserved within all nuclear LIM-only proteins whereas K78 was unique to LMO2 (Fig. 4B). Both K74 and K78 restored binding of the lysineless LMO2 to LDB1 (data not shown).
LMO2 binding and degradation depend on lysines in the LDB1 binding pocket. (A) Schematic showing a model of LMO2 (yellow backbone) bound to the LID domain (orange backbone) with relevant residues (discussed in Results) marked by arrows. (B) Alignment of LMO1 to LMO4 and ISL2. Arrows show residues important for LMO2 stability and LDB1 binding as discussed in Results. (C) Immunoblot of Jurkat cells transduced with empty virus (EBFPII-Hygro) or with the construct shown in the grid. In descending order, blots were probed with anti-LMO2, anti-Halo, anti-LDB1, anti-GFP (expression control), and anti-VCP (gel loading control). (D) Bar graph showing t1/2 of various Halo-LMO2 proteins. Error bars represent results from triplicate independent experiments. Comparisons to Halo-LMO2 were performed by paired two-tailed t tests, with the P values shown above the bars. N.S., not significant.
Lysineless proteins are degraded by the proteasome after ubiquitination of their amino termini. In order to show that the N terminus of this version of LMO2 was critical for ubiquitin modification (31, 32), we inserted a native LMO2 sequence translated from the longest transcript of the distal LMO2 promoter, creating a superstable protein, Halo-N+LMO2 K(0)(K74, K78), with a measured t1/2 value of 25 h (Fig. 4D) (P = 4.47E−3); the comparable lysineless Halo-N+LMO2 showed a t1/2 of 14.4 h. In summary, we identified K74 and K78 within LMO2 as essential for LDB1 binding and for normal levels of protein turnover.
The effect of LIM domain proteins on Halo-LMO2 turnover.Next, we examined the turnover of Halo-LMO2 in Jurkat, KOPT-K1, and K562 leukemia cells, which have various levels of LDB1 and LMO2. Jurkat cells are derived from T-ALL and express endogenous LMO1 but not LMO2; KOPT-K1 cells have a chromosomal translocation that results in overexpression of endogenous LMO2; and K562 cells are aneuploid chronic myelogenous leukemia cells, resemble hematopoietic stem and progenitor cells (HSPCs), and express abundant endogenous LMO2 and LDB1 (Fig. 5A) (33). Halo-LMO2 t1/2 levels were comparable in Jurkat and K562 cells, measured at 6.2 h and 6.4 h, respectively. The half-live times for superstable Halo-N+LMO2 K(0)(K74, K78) in those two cell lines were similarly prolonged at t1/2 = 25 and t1/2 = 20.9, respectively (Fig. 5B). In contrast, Halo-LMO2 t1/2 measured 1.3 h in KOPT-K1 cells.
The effect of LIM domain proteins on LMO2 turnover. (A) Immunoblot of Jurkat, KOPT-K1, and K562 cells transduced with empty virus (EBFPII-Hygro) or the construct shown in the grid probed with anti-LMO2, anti-LDB1, anti-GFP (expression control), and anti-VCP (gel loading control). (B and D) Bar graphs showing t1/2 of Halo-LMO2 (B) or Halo-LMO2 N+K(0) K74+ K78+ (D) proteins. Comparisons to Halo-LMO2 were performed by paired two-tailed t tests, with the P values shown above the bars. All experiments were performed three times. (C) Immunoblot of Jurkat cells expressing the various constructs shown in the grid probed with anti-LMO2, anti-HA, anti-Myc. HA and Myc epitope tagged proteins are labeled in the grid.
The fast turnover in KOPT-K1 cells suggested to us that Halo-LMO2 was competing with high levels of endogenous LMO2 (Fig. 5A, lanes 5 to 8) for the LDB1 LID. The LMO2 abundance seen with the K562 cells was approximately equivalent to that seen with that KOPT-K1 cells; however, Halo-LMO2 turnover in K562 cells, was not as fast perhaps due to the increased expression of endogenous LDB1 in the K562 cells in comparison to the KOPT-K1 cells (Fig. 5A, lanes 9 to 12). Competition among LIM domain proteins is an important determinant of neuronal cell type specificity in the spinal cord (34). To test this competition model and its effect on turnover, we measured Halo-LMO2 t1/2 and the effects of coexpression of the following competing nuclear LIM domain proteins: LMO2-HA, LMO1-HA, LMO4-HA, LHX9-HA, and ISL2-HA. These HA-tagged proteins were found to be expressed at various levels in Jurkat cells (Fig. 5C, lanes 4 to 8), but their forced coexpression increased the turnover of Halo-LMO2 (Fig. 5D). These results with respect to t1/2 normalized to the level of expression achieved suggested an approximate order of levels of affinity of LIM domain proteins for LDB1 LID. LMO2-HA was the most competitive followed by LMO1, LMO4, LHX9, and ISL2. The LIM domain proteins that enhanced Halo-LMO2 turnover showed greater conservation of the key residues that we had identified for LID binding, L64, L71, K74, and K78 (Fig. 4B). All the LIM proteins tested had L64 conserved; however, only LMO1 and LMO2 had L71 (Fig. 4B). LMO4 and LHX9 have a cysteine residue in place of K78 but have a conserved K74 at the comparable position. Fitting this logic, ISL2, the protein that had no effect on Halo-LMO2 turnover, suggesting that ISL2 was the weakest competitor for LID binding, has an arginine residue in place of K74 and a threonine residue in place of K78 (Fig. 5D).
We also coexpressed other known LMO2 binding partners and measured their effects on LMO2 turnover. TAL1 increased Halo-LMO2 t1/2 to 8.9 h (P = 0.017), but LYL1 showed no change from WT levels (6.9 versus 7.0 h, P = 0.75). Coexpression of both Myc-GATA2 and Myc-GATA3 significantly decreased Halo-LMO2 t1/2 to 4.9 h (P = 0.013) and 4.8 h (P = 0.011), respectively. Myc-GATA3 expression was weak but had a substantial effect on Halo-LMO2. Finally, Halo-LMO2 had a measured t1/2 of 7.7 h with HA–single-stranded DNA-binding protein 2 (SSBP2) coexpression, a statistically insignificant change from WT turnover.
LDB1 is a long-lived protein in leukemia cells.The stabilization of LMO2 to ∼20 h suggested that LDB1 itself may be long-lived. LDB1 proteins deficient in LMO2 binding (i.e., ΔLID, RLIT→AAAA, and I322A) caused faster turnover of Halo-LMO2 with coexpression, suggesting that these mutant LDB1 proteins may themselves turn over faster that the wild-type protein. To directly test this idea, we tagged LDB1 at its amino terminus with the SNAP tag and transduced SNAP-LDB1 into Halo-LMO2-expressing Jurkat cells. The SNAP tag has chemistry distinct from that of the Halo proteins, allowing us to simultaneously probe SNAP-LDB1 and Halo-LMO2 within the same cells (35). As shown in the immunoblots, SNAP-LDB1 proteins were expressed at similar levels to each other and to endogenous LDB1 (Fig. 6A, lanes 4 to 7). Pulse-chase analysis showed that Halo-LMO2’s turnover was increased by coexpression of SNAP-LDB1, t1/2 = 5.9 h to 14.6 h (P = 1.81E−3) (Fig. 6B). SNAP-LDB1ΔLID accelerated the turnover of Halo-LMO2 from 5.9 to 3.7 h (P = 2E−4) (Fig. 6B), similarly to LDB1ΔLID (Fig. 3D). This dominant-negative effect on steady-state LMO2 has been a consistent finding across multiple leukemic cell lines (7), and data in Fig. 3 and 6 show that it represents a kinetic effect on LMO2 turnover. SNAP-LDB1(RLIT→AAAA) also showed increased Halo-LMO2 turnover from 5.9 to 3.9 h (P = 8.68E−4). SNAP-LDB1(I322A) did not change significantly with respect to Halo-LMO2 turnover, 5.9 versus 6.4 h (P = not significant [NS]). As shown previously, the I322A mutant was only partially deficient in LMO2 binding. SNAP pulse-chase analysis showed that the SNAP-tagged LDB1 proteins all showed similar turnover rates, with t1/2 ranging from 18.6 to 24.7 h (Fig. 6B). These results show that LDB1 is a long-lived protein that imparts its stability to its direct interaction partner, LMO2. We repeated its pulse-chase analysis with Halo-LDB1, and its stability was consistent across diverse cell lines, measuring t1/2 of 23.6 to 27.6 h in Jurkat, KOPT-K1, and K562 cells (Fig. 6C), respectively.
LDB1 is a long-lived protein that imparts its stability to bound LMO2. Data represent results of multiplex Snap-LDB1/Halo-LMO2 or Halo-LDB1 pulse-chase analyses. (A) Immunoblots of Jurkat cells transduced with the constructs shown in the grid probed with anti-LDB1, anti-SNAP, anti-LMO2, anti-Halo, anti-GFP (expression control), and anti-VCP (gel loading control). The empty virus control was EBFPII-puro. (B) Bar graph showing t1/2 of Halo-LMO2 or SNAP-LDB1. Halo-LMO2 comparisons were performed by paired two-tailed t tests, with the P values shown above the bars. Error bars represent results from triplicate independent experiments. (C) Bar graph showing Halo-LDB1 t1/2 determined by Halo pulse-chase assay. (D) Immunoblot of Jurkat cells transduced with the constructs shown in the grid and probed with anti-LDB1, anti-HA, anti-GFP (expression control), anti-VCP (gel loading control). (E) Bar graph showing Halo-LDB1 t1/2. Error bars represent results from triplicate independent experiments. Comparisons to Halo-LDB1 WT were performed by paired two-tailed t tests, with the P values shown above the bars.
Prior studies had implicated K134 and K365 residues within LDB1 as affecting its degradation (17, 36). Compared to LDB1 WT, which had t1/2 of 27.7 h, LDB1(K134R) and LDB1(K365R) half-lives were prolonged, t1/2 = 77.2 h and t1/2 = 48.2 h, respectively (Fig. 6E). Immunoblots with anti-LDB1 showed two closely migrating bands, the more slowly migrating band being enhanced in abundance with N-ethylmaleimide (NEM) (Fig. 6D). This more slowly migrating migrating band was not observed in blots for LDB1 (K134R), suggesting the addition of a monoubiquitin at this residue (36). Halo-LDB1 turnover was inhibited by bortezomib, giving an extrapolated t1/2 of 203 h (Fig. 6E). Thus, LDB1 appears to be highly regulated by the ubiquitin-proteasome system, in addition to and in spite of being a very stable protein.
Halo pulse-chase analysis of SSBP, GATA, and LMO factors.In MEL and CHO cells, LDB1 stabilization was previously shown to be dependent on single-stranded DNA-binding protein 2 (SSBP2) (37). In contrast to those studies, here, LDB1 abundance did not increase with forced expression of SSBP2 or SSBP3 in any of the leukemic lines analyzed (data not shown). We directly tested the turnover of SSBP2, SSBP3, and SSBP4 by Halo pulse-chase analysis. Each SSBP paralog tested, including SSBP2, SSBP3, and SSBP4, had faster turnover than LDB1, measured at t1/2 = 5.1 h, t1/2 = 6.8 h, and t1/2 = 7.6 h, respectively (Fig. 7A and B). SSBP2 and SSBP3 showed longer half-lives with LDB1 coexpression (Fig. 6C and data not shown). Stabilization of SSBP2 and SSBP3 was not seen with coexpression of LDB1ΔLCCD, representing the interaction domain between SSBP proteins and LDB1 (data not shown). Next, we analyzed the turnover of GATA factors 1 to 3. Each GATA factor had a turnover faster than LDB1, but there were major differences between the factors. Halo-GATA1 had a t1/2 of 6.2 h, Halo-GATA2 had a t1/2 of 2.0, and Halo-GATA3 had the fastest turnover with a t1/2 of 1.2 h (Fig. 7D). These results were in agreement with the half-lives observed in cycloheximide chase experiments (38). Next, we analyzed LIM domain partners of LDB1 other than LMO2. Halo-LMO1 measured t1/2 = 9.7 h, Halo-LMO4 measured t1/2 = 5.4 h, and Halo-ISL2 measured t1/2 = 2.9 h (Fig. 7E). In summary, the data showed that every subunit of the LDB1/LMO2 complex analyzed had a shorter half-life than LDB1.
Halo-SSBP, GATA, and alternative LIM protein turnover. (A) Immunoblot of Jurkat cells transduced with the constructs shown in the grid probed with anti-Halo, anti-LDB1, anti-LMO2, anti-SSBP2, anti-TCF3, anti-GFP (expression control), and anti-VCP (gel loading control); (B to E) bar graphs showing t1/2 derived from Halo pulse-chase experiments. Error bars represent results from triplicate independent experiments. Comparisons to proteins shown in yellow bars were performed by paired two-tailed t tests, with the P values shown above the bars.
TAL1 and LYL1 are stabilized by the LMO2/LDB1 complex.TAL1 and LYL1 are necessary cooperating drivers in LMO2-induced leukemia (23, 25, 39). These class II bHLH proteins are known binding partners of LMO2. The binding interface between TAL1 and LMO2 requires the presence of F238 within the second helix of the bHLH domain (40), which is conserved as F201 within helix-2 of LYL1 (Fig. 8A). We tested the turnover of Halo-TAL1 and Halo-LYL1 and specific aspartic acid and glycine substitutions at F238 and F201, respectively. The immunoblots showed the expression of the Halo-tagged proteins (Fig. 8B and D). Halo-TAL1 had a t1/2 of 4.2 h, and Halo-LYL1 had a t1/2 of 1.8 h (Fig. 8C and E). LMO2-HA coexpression did not significantly stabilize TAL1 (t1/2 = 5.6 h with LMO2 versus t1/2 = 4.2 h without LMO2, P = 0.215) but stabilized LYL1 (t1/2 = 4.3 h versus 1.8 h, P = 0.015). HA-LDB1 coexpression markedly stabilized Halo-TAL1 and Halo-LYL1 to t1/2 = 19.9 h and t1/2 = 20.5 h, respectively. This effect was observed only in the presence of LMO2, and no stabilization was observed with coexpression of HA-LDB1ΔLID (Fig. 8C and E). Thus, LDB1’s stabilization effect was not observed without LMO2 binding. To test the bHLH requirement for LMO2 binding, we analyzed Halo-TAL1(F238D), Halo-TAL1(F238G), Halo-LYL1(F201D), and LYL1(F201G), all of which were compromised in LMO2 binding in coimmunoprecipitation assays (data not shown). As expected, LMO2 did not stabilize these variant proteins to the extent observed with the equivalent wild-type versions. Each mutant bHLH protein had a measured t1/2 comparable to that of its WT counterpart. HA-LDB1 coexpression increased the t1/2 of Halo-TAL1(F238D) to 10.7 h (P = 0.014). Similarly, Halo-LYL1(F201D) was stabilized by HA-LDB1 coexpression and showed a t1/2 of 3.7 h (P = 0.012). Thus, aspartic acid substitutions at the crucial phenylalanine residues (F238 in TAL1 and F201 in LYL1) completely abrogated LMO2-induced stabilization of LYL1 but had no effect on TAL1. These substitutions significantly abrogated LDB1-induced stabilization. The F238D and F201D mutants may still retain some LMO2 binding, especially since LDB1 stabilizes LMO2 and increases its steady-state abundance. In contrast, glycine substitutions at the same residues completely abrogated the effects of both LMO2 and LDB1. In summary, Halo-TAL1 and Halo-LYL1 half-lives in Jurkat cells were partially stabilized by LMO2 coexpression. Their half-lives are markedly prolonged by LDB1 coexpression but only if the proteins have intact LMO2 binding.
TAL1 and LYL1 are stabilized by the LMO2/LDB1 complex. (A) Alignment of TAL1 and LYL1 basic-helix-loop-helix domains, with arrows showing residues required for LMO2 binding. (B) Immunoblot of Jurkat cells transduced with the constructs shown in the grid and probed by anti-LDB1, anti-HA (in two panels cropped to show HA-LDB1 and LMO2-HA), anti-LMO2, anti-Halo, anti-GFP (expression control), and antitubulin and anti-VCP as gel loading controls. (C) Bar graph showing Halo-TAL1 t1/2, with grid underneath showing coexpression of empty vector (EBFPII-puro or mScarlet-Hygro) or indicated construct. (D) Immunoblot of Jurkat cells transduced with constructs in grid with the same panels of Westerns as described for panel A. (E) Bar graph showing Halo-LYL1 t1/2, with the same pattern as that described for panel C.
Complex assembly and function.Our results implied that intact binding interactions between all of the components created a stable macromolecular complex. We analyzed whether this assembly occurred in cells and whether complex assembly had a functional effect on transcription. Each component of our complex was expressed using a lentiviral vector with unique fluorescence and drug selection (Fig. S1). We included empty vector controls (Fig. 9A) as indicated. We transduced components pairwise with or without FLAG-LDB1 (F-LDB1) to test abundance (Fig. 9A) and binding (Fig. 9B) by co-IP with anti-FLAG monoclonal antibody. The measured half-lives uniformly correlated with steady-state abundances of Halo-tagged proteins detected by Western blotting. The results of the experiments performed as described for Fig. 9A extend this correlation to endogenous untagged or minimally tagged (i.e., single-HA-tagged) proteins. SSBP2 was poorly expressed in Jurkat cells, so SSBP3 was transduced instead; our prior experiments had shown comparable peptide counts for SSBP3 and SSBP2 by tandem mass spectrometry of purified F-LDB1 complexes (7). HA-SSBP3 was stabilized by LDB1 but not by coexpression of LMO2 (see lanes 6 to 9, Fig. 9A). Consistent with the Halo pulse-chase results, TAL1 and LYL1 were maximally stabilized by the coexpression of both LMO2 and LDB1 (Fig. 9A, lanes 10 to 13 for TAL1 and lanes 18 to 21 for LYL1).
LMO2/LDB1 complex assembly is required for complete phenotypic modulation of target gene expression. (A) Immunoblots of Jurkat cells transduced with the constructs shown in the grid. Each lentiviral construct had a unique color and antibiotic selection. The corresponding empty vector controls (EGFP-puro, mScarlet-Hygro, and Cerulean-blast) are shown. Proteins were tagged with FLAG or HA as shown. Anti-GFP recognizes EGFP, EBFPII, and Cerulean. Scarlet was tagged with V5. Anti-VCP was used as a gel loading control. (B) Cell lysates from the transduced cells shown in panel A were subjected to FLAG immunoprecipitation and immunoblotted as described here. Anti-heavy chain and anti-light chain data are shown for capture control. (C) Bar graph showing RNA-seq reads per kilobase per million (RPKM) for CEBPE of Jurkat cells transduced in the same pattern as that described for panel A. Every bar graph label corresponds to a lane number in the immunoblots in panels A and B. For example, bar graph labeled 2 represents the sample shown in lane 2 of 9A (Jurkat with empty vector transduction). (D) Heat map showing top 50 genes that are differentially expressed in pairwise comparisons of empty vector-transduced Jurkat (lane 2 of 9A) versus Jurkat transduced with all components of the complex. Columns 2 and 3 differ in the bHLHs transduced (TAL1 and LYL1, respectively). Comparisons represent lane 17 versus lane 2 for TAL1 and lane 25 versus lane 2 for LYL1.
Complex assembly was analyzed by anti-FLAG immunoprecipitation of F-LDB1. Jurkat cells have abundant endogenous TAL1, which was immunoprecipitated by F-LDB1 only in the presence of LMO2 (Fig. 9B, lanes 2 to 5). Endogenous TAL1 co-IP was augmented by coexpression of SSBP3 (Fig. 9B, lanes 6 to 9). Forced expression of LYL1 did not effectively outcompete endogenous TAL1 for LMO2/LDB1 binding, whereas SSBP3 and LYL1 coexpression reduced steady-state TAL1 and TAL1 co-IP (Fig. 9B, lanes 21 and 25). Next, we analyzed the effects of complex formation on gene expression. We performed a pairwise comparison of transcriptome sequencing (RNA-seq) results for Jurkat cells transduced with all complex components (i.e., LMO2, LDB1, SSBP3, and TAL1 or LYL1 [Fig. 9, lanes 17 and 25]) versus cells transduced with empty virus (Fig. 9, lane 2), generating a ranked list of differentially expressed genes. Most of the genes on this list were maximally activated or repressed by coexpression of the full complex but not by expression of partial complex components. CEBPE is shown as an example (Fig. 9C). All the genes requiring full LDB1/LMO2 complex expression are shown in the heat map, which included other bona fide targets such as NKX3-1 and ALDH1A2, both of which are representative of authentic and direct physiological TAL1 target genes (Fig. 9D).
Halo pulse-chase analysis can be used to screen for modifiers of degradation.Next, we asked whether the stable leukemia lines expressing various Halo-tagged proteins can be used in a screen to identify modifiers of stability. Deubiquitinases (DUBs) of the LMO2-associated proteins would stabilize LMO2 complex formation and could be important therapeutic targets in leukemias that are dependent upon LMO2. Also, the number of genes encoding DUBs was suitable for a targeted screen, totaling ∼80 genes versus ∼400 genes encoding E3 enzymes (41). We assembled a lentiviral short hairpin RNA (shRNA) library against 70 DUB genes, of which 44 (63%) were expressed in Jurkat cells. We transduced pooled shRNAs directed against each DUB into individual Jurkat lines stably expressing Halo-LMO2, Halo-LDB1, Halo-SSBP2, Halo-SSBP3, Halo-TAL1, or Halo-LYL1. After transduction, we analyzed the cells for their growth and for effects on the Halo-tagged proteins. We devised three criteria to identify an important hit: (i) a reduction in the percentage of R110 fluorescence at t0 in cells transduced with a DUB-specific shRNA compared to scrambled shRNA; (ii) a reduction in absolute Halo signal (i.e., mean fluorescence intensity [MFI]) at t0; and (iii) a reduction in Halo signal after a 5-h chase (Fig. 10A). The outcomes of this screen are shown in Fig. 10B (see also Fig. S4). We identified a set of shRNAs against a DUB, ALG13, that met all 3 criteria for every subunit of the complex for Halo-LMO2, Halo-LDB1, Halo-SSBP2, and Halo-SSBP3 and that met 2 criteria for Halo-TAL1 and Halo-LYL1 (Fig. 10B). Other DUBs that potentially affected some of the subunits met 2 of 3 criteria, including OTUD7B, USP3, and USP4 (Fig. S4). ALG13 is a DUB with an unusual structure. ALG13 has an amino-terminal glycosyltransferase domain (42) followed by the DUB domain found in the ovarian tumor (OTU) class of DUBs and a Tudor domain followed by a proline-rich domain (43). The OTU family of DUBs had several hits that met our criteria for various subunits (Fig. 10B, ALG13 and OTUD7B). Pooled shRNAs against the DUBs identified were validated in a secondary screen and a time course for Halo-LMO2 degradation. As shown in Fig. 10C, shRNA knockdown of ALG13, OTUD7B, USP3, and USP4 accelerated the degradation of Halo-LMO2 compared to scrambled shRNA control. The ALG13 shRNA pool was comprised of 5 shRNAs, which we tested individually in the same assay. Four of the 5 shRNAs against ALG13 caused increased turnover of Halo-LMO2 (data not shown). To further validate the role of ALG13 in LMO2 degradation, we performed the Halo pulse-chase analysis in K562 cells by forcing the expression of full-length ALG13 (1,137 amino acids [aa]) or a catalytically inactive ALG13 mutant and measuring the resultant t1/2. We deleted the DUB domain creating ALG13ΔDUB (deleted catalytic DUB domain) but could not rule out drastic effects on folding of the protein, so we engineered a point mutant, ALG13 C242R. Interestingly, alanine substitution at the catalytic cysteine residue can enhance the affinity for ubiquitin in OTU DUBs so an arginine is the better residue for use in substitution to evaluate a catalytically inactive DUB (44). We measured a t1/2 of Halo-LMO2 of 6.4 h in the empty vector control, but with forced expression of full-length ALG13, we measured an extended t1/2 of 7.6 h (P = 0.009 for comparison to empty vector control). In contrast, we measured t1/2 of 6.7 h (P = NS) and 6.3 h (P = NS) with ALG13ΔDUB and ALG13 C242R mutant proteins, respectively (Fig. 10F).
Reverse genetic screens to identify deubiquitinases that alter turnover of LMO2/LDB1 complex subunits. (A) Schema for shRNA screening for DUBs that perturb Halo-tagged protein stability. Jurkat cells stably expressing one Halo-tagged protein were screened with the criteria shown. (B) OTU DUBs were identified repeatedly in our screens and are listed here along with data indicating their presence in Jurkat cells and whether the shRNA had a growth effect and which Halo-tagged protein was more rapidly degraded. Yellow boxes denote that 2 of the 3 criteria shown in panel A were met; orange boxes denote that 3 of the 3 criteria were met. (C) Decay curve for Halo-LMO2 in Jurkat cells transduced with shRNAs shown. Error bars denote three independent analyses. (E) Immunoblot of Jurkat cells transduced with constructs shown. The empty vector control for Halo-LMO2 was EBFPII-Hygro. WT and mutant FLAG-ALG13 were expressed in EBFPII-blast vector. (D) Bar graph showing Halo-LMO2 t1/2. Error bars represent results from triplicate independent experiments. Comparisons to Halo-LMO2 (shown with the green bar) were performed by paired two-tailed t tests, with the P values shown above the bars. (F) Bar graph showing Halo-LMO2 t1/2 with coexpression of empty vector or FLAG-ALG13 WT or mutant proteins. Error bars are for three independent experiments.
DISCUSSION
In this study, we describe a novel technique to analyze the turnover of the components of the leukemogenic LMO2/LDB1 protein complex, employing Halo tagging and fluorescence-based pulse-chase analysis. The assay, which we termed HaloLife, is informative in that the turnover of tagged proteins is observed in live cells. Thus, proteins are observed in their natural milieu without pharmacological, nutritional, or mechanical disruption. This method has the added advantage of allowing the testing of the effects of various culture conditions and small-molecule therapeutics on protein turnover. The Halo tag is advantageous because it is relatively small and monomeric, representing approximately the mass of green fluorescent protein (GFP), which has been used in similar studies (45). Of course, as is the case in all epitope-tagging experiments, one must verify that the tag itself does not disrupt the behavior of the protein. In the case of the proteins presented here, each was localized to the nucleus (see Fig. S2 in the supplemental material) and retained its affinity for its physiological partners. Also, the proteins that had mutations that disrupted binding had the same effect on Halo-tagged versions as the untagged proteins themselves. The pulse-chase analysis showed that the Halo protein itself was very long-lived (t1/2 of >100 h). Each Halo-tagged protein had rapid turnover compared to Halo itself, such that the fusion proteins acted as “degrons” for the Halo protein. In light of the caveats noted, the t1/2 measured in the HaloLife assay can be viewed as an approximation of the true half-life of the native protein. To calculate half-life values, we chose to follow the decay of positive-testing cells over time since MFI values represent averages of the bulk population fluorescence. Our method may be improved by time lapse fluorescence microscopy (46). Nevertheless, our methods measured half-lives that matched those estimated from cycloheximide chase and quantitative immunoblotting (38) and provided an explanation for detected changes in steady-state abundance. Most importantly, the HaloLife assay has the compelling advantages of being performed in live cells, in their native cellular milieu, and at steady state without cellular disruption.
HaloLife analysis of LMO2 and its binding partners revealed a hierarchy of protein turnover, with LDB1 being the most stable protein. The observed half-life times in Jurkat cells were as follows (in increasing order): for Halo-LYL1, ∼1.8 h; for Halo-TAL1, ∼4.1 h; for Halo-LMO2, ∼6.4 h; for Halo-SSBP2, ∼5.1 h; for Halo-SSBP3, ∼6.8 h; and for Halo-LDB1, ∼20 to 24 h. Some half-lives such as that determined for Halo-LDB1 in the presence of bortezomib were extrapolated out to >20 h and suggest the stabilization of LDB1 beyond the cell cycle. Most remarkably, coexpression of LDB1 shifted the turnover of these Halo-tagged subunits such that each protein partner assumed a half-life of ∼20 h in the presence of excess LDB1, approximating the measured half-life of LDB1 itself. There was no reciprocal effect since none of the partner proteins prolonged the half-life of LDB1. All proteins tested were markedly stabilized by bortezomib, suggesting degradation by the ubiquitin proteasomal system. To be stabilized, each protein partner had to bind to LDB1 either directly or indirectly, as in the case of TAL1 and LYL1. Taken together, these findings suggest that the free subunits, i.e., those not bound to LDB1, are degraded more rapidly than those bound to LDB1. Furthermore, the prolonged half-life of LDB1 suggests that it is the core subunit in the assembly of the bHLH/LMO2/SSBP/LDB1 macromolecular complex, which we term the LDB1/LMO2 holocomplex. As LDB1 binds to its direct partners, SSBP proteins or LMO2, LDB1 impedes the turnover of other components of the complex so that stepwise assembly and slow turnover increase the steady-state abundance of the holocomplex. Accordingly, each subunit assumes a half-life similar to that of LDB1, suggesting that the whole complex may be degraded en masse. Two distinct lysines within LDB1, K134 and K365, have been implicated in LDB1 turnover. Both K134R and K365R mutations markedly prolonged LDB1 turnover in the HaloLife assay compared to wild-type LDB1, thereby confirming the role of these lysine residues in LDB1 stability. Neither lysine is within a domain mediating subunit binding (i.e., LDB1’s LCCD, residues 200 to 249, is responsible for SSBP binding and the LID is comprised of residues 300 to 330). Thus, these residues are unlikely to be occluded from ubiquitination by SSBP or LMO proteins. On the other hand, K134 is within the dimerization domain, so K134 could be masked by homodimerization. This raises the possibility of LDB1 homodimers being more stable than monomers. We discovered a slower-migrating LDB1 in the presence of N-ethylmaleimide that is consistent with a monoubiquitin conjugation to K134. If we assume that this residue is accessible only in unbound LDB1, then we predict that this monoubiquitinated LDB1 is monomeric. Although the stoichiometry of the LDB1 holocomplex has not been definitively solved, our prior mass spectrometry data do suggest stable LDB1 dimers in nuclear lysates. Interestingly, this theme of accessible lysines may be extended to the turnover of LMO2 and SSBP proteins as well. Our experiments with LMO2 implicated K74 and K78 in LDB1 binding. These residues may be sites of ubiquitination and may be exposed in free LMO2 subunits but sterically hindered in LMO2 bound to LDB1. Alternatively, K74 and K78 may be subject to other posttranslational modifications such as methylation or acetylation. K78 is particularly intriguing since it is unique to LMO2 and is adjacent to a hydrophobic pocket (L64 and L71) such that neutralization of the side-chain amine would favor LDB1 binding by accommodating I322. This contact interface is supported by a crystal structure of an LMO2-LID fusion protein (10). We copurified SSBP3 with FLAG-LDB1 and detected a di-Gly motif on K35 in mass spectrometry data (data not shown) which might have represented a remnant of trypsinized ubiquitin, although NEDD8 and ISG13 are other possible conjugates (47). Nevertheless, K35, K7, and other conserved lysines are within the LUFS domain of SSBP proteins and are expected to be masked by LDB1 binding whereas free SSBP subunits should have more-accessible lysine residues for modification.
In summary, free subunits of the LMO2/LDB1 complex are rapidly degraded in comparison to the slow degradation kinetics of the holocomplex. Complex assembly may proceed through binding and stabilization by masking key lysine residues in the free subunits. Recombinant full-length proteins and a structure of the holocomplex may be able usable to test this model (Fig. 11). On a more general note, our studies suggest that multisubunit protein complexes may have key core subunits with enhanced stability that can be conferred on binding subunits. To name a few examples, core subunits analogous to LDB1 exist for the T-cell receptor, BAF complex, Mediator complex, and TFIID protein complexes (48–52). It would be interesting to see whether lysine residues targeted for ubiquitination are masked in other macromolecular assemblies as well.
Model for LDB1 complex assembly and stabilization.
Prolonged turnover of nuclear factors and transcription factors has been suggested to be due to their association with chromatin. The subunits of the LDB1/LMO2 complex were localized to the nucleus, but we were unable to analyze whether they were chromatin bound. By comparison to Halo-CTCF, we estimated that Halo-LMO2 was present at 22,000 copies per cell or 20,900 copies per nucleus (confocal analysis showed 95% nuclear localization). Fine-resolution mapping of TAL1 sites in erythroid progenitors resulted in an estimation of 15,000 total sites in the genome, which suggests that LMO2 may be present in slight excess (53).
Note that the HaloLife assays were all performed in leukemic cells. The leukemia lines were of diverse lineages. Even so, one cannot rule out the possibility of a general defect in the turnover of LMO2 and LDB1 in all of these lines. Importantly, results of careful analysis of this protein complex turnover have major implications for regulating these major drivers of leukemia. Recent data from mouse genetics strongly support the idea of a role for Ldb1 in Lmo2-induced leukemia. The CD2-Lmo2 transgenic mouse model develops T-ALL with long latency but with complete penetrance (25). Conditional deletion of Ldb1 in this model abrogated T-ALL onset (65; U. P. Davé, personal observation). Thus, Ldb1 is a required Lmo2 partner in this murine model of T-ALL. This compelling result from mouse genetics and the primacy of LDB1 in a protein turnover hierarchy underscore the potential for targeting the LMO2/LDB1 interface in leukemias. When LMO2 is dissociated from LDB1, then free LMO2 and TAL1 are expected to undergo rapid degradation. Supporting this idea, the coexpression of LIM domain proteins that competed for the LID (LMO1, LMO2, LMO4, and LHX9) accelerated Halo-LMO2 turnover. ISL2, which has the least similarity to the LMO2 residues responsible for LID binding, did not accelerate turnover, underscoring the function of precise determinants of LID binding as a mechanism for LIM protein competition. We predict that a small molecule able to bind to the LID interface would also accelerate LMO2 turnover. Of course, such an inhibitor of LMO2 binding to LDB1 would affect normal hematopoietic stem cells as well. However, there could be a therapeutic index with higher LMO2/LDB1 holocomplex-expressing cells predicted to be more sensitive to such inhibition.
Previous work implicated RNF12 as a potential E3 enzyme responsible for LDB1 and LMO2 degradation (37, 54, 55). However, in our experiments, the steady-state abundances of LDB1 and other subunit proteins were unchanged with forced expression of RNF12 in Jurkat cells (data not shown). Thus, additional investigation is needed to characterize the degradation machinery of the LMO2 holocomplex, especially in its normal or leukemic cellular contexts, which could reveal E3 enzymes or DUBs that could be therapeutically targeted. DUB enzymes are particularly amenable to small-molecule inhibition since proteolytic mechanisms have been extensively studied. An shRNA knockdown screen performed using the HaloLife assay showed a very compelling candidate DUB, ALG13. There were other candidates identified in our screen such as OTUD7B, but ALG13 fulfilled our screening criteria and affected all subunits with no effect on Halo protein itself. Recently, with the development of proteolysis targeting chimeras (i.e., PROTACs), great interest developed in small molecules that can induce targeted degradation by recruitment of E3s to proteins of interest (56). Actually, one of these PROTACs is being analyzed in phase II clinical trials, with similar molecules on the horizon (57). In contrast, bortezomib is being tested in a randomized clinical trial in T-ALL as an addition to state-of-the-art multiagent chemotherapy. The results from our study show that bortezomib stabilizes LMO2 oncoprotein, which can potentially antagonize the effect of chemotherapies. However, the overall effects of bortezomib on T-ALL and patient survival are difficult to predict since bortezomib affects pathways other than LMO2 causing proteotoxic stress in leukemic cells (58). Our ongoing work on LMO2/LDB1 complex turnover should be highly revealing for both normal hematopoietic stem cell biology and for the development of novel leukemia therapies.
MATERIALS AND METHODS
Resources and reagents.The resources and reagents used in this study are listed in Table 1.
Reagents and resources
Experimental model and subject details. (i) Cell lines.HEK 293T, Jurkat, K562, U937, KOPT-K1, and LOUCY cells were acquired from the American Type Culture Collection (ATCC). HEK293T cells were cultured in Iscove’s modified Dulbecco’s medium (IMDM)–10% fetal bovine serum (FBS), and other lines were cultured in RPMI 1640–10% FBS, at 37°C in 5% CO2. Log-phase HEK 293T cells in 10-cm-diameter dishes containing 10 ml medium and 5 × 106 to 8 × 106 cells were transfected by a calcium phosphate–HEPES-buffered saline method with 1 pmol pH163 constructs and 2 pmol pMD-2.G (containing vesicular stomatitis virus G [VSV G]) for producing pseudotyped lentiviruses. At 12 to 18 h posttransfection, medium was aspirated and replaced with 6 ml fresh medium, which was harvested and replaced at 24 h and 48 h. Medium containing viral particles was aliquoted and frozen at –80°C, and viral titers were subsequently estimated by serial dilution infection of Jurkat cells. Various volumes of viral supernatant were mixed with 5 × 106 to 1 × 107 log phase Jurkat cells in a final volume of 10 ml within a T-25 flask (Eppendorf) and subsequently cultured for 72 h, at which time percentages of fluorescence-positive cells were first roughly determined using an EVOS FL inverted fluorescence microscope (Invitrogen) and then precisely determined using a CytoFLEX benchtop cytometer (Beckman). Microscopy and cytometry gating parameters were established using a parallel culture of noninfected cells as a reference. A multiplicity of infection (MOI) of 1 was associated with fluorescence positivity of 30% or less. Typical viral titers were 1 × 106 to 2 × 106 infectious particles per milliliter. Jurkat cells infected at an MOI of 1 to 2 were expanded into a 50-ml culture containing antibiotics to eliminate noninfected cells. Antibiotic regimen and dose varied depending on the selectable marker encoded by the virus in question and the cell line being transduced; antibiotic concentration kill curves were empirically established for naive cell lines. As an example, typical antibiotic concentrations for transduced Jurkat cells were puromycin at 2 μg/ml, hygromycin B at 200 μg/ml, G418 at 500 μg/ml, blasticidin at 10 μg/ml, or Zeocin at 50 μg/ml. After 4 to 10 days of antibiotic selection, cell populations were typically 100% fluorescence positive, at which point they were cryopreserved in liquid nitrogen using growth media supplemented with 10% dimethyl sulfoxide (DMSO), subjected to iterative rounds of transduction with additional viruses exactly as described above, or used directly for experiments.
(ii) Lentiviral expression vector system.Previously, we used multiplexed lentiviral infection with GFP-marked and red fluorescent protein (RFP)-marked viruses to create recombinant leukemia cell lines, in conjunction with fluorescence-assisted cell sorter (FACS) analysis (7). FACS analysis was laborious and expensive, while the use of GFP and RFP markers limited the number of coexpressed recombinant factors to two (LDB1 and LMO2). Moreover, we observed that initially homogenous FACS-sorted cell lines were able to inactivate transgene (GFP or RFP) expression over time, consistent with either transgene silencing or a competitive advantage/outgrowth of low-expressing clones (J. H. Layer and U. P. Davé, unpublished data). This phenomenon occurred variably among different cell lines/types. To circumvent these limitations for the present study, we designed a suite of novel lentiviral vectors. This modular vector family expresses additional fluorescence protein markers that are spectrally distinct, allowing multiplexed coinfection with five or more different viruses. Each vector also encodes a unique antibiotic resistance marker to allow positive selection of transduced cells. Antibiotic resistance in transduced cells eliminates the need for FACS analysis and disallows transgene silencing within transduced cell lines; all of which can be proven by confirmation of the antibiotic-enforced consistency of fluorescence marker expression, as monitored by flow cytometry.
(iii) Lentiviral vector construction.We modified a previously described second-generation lentiviral vector (59). First, artificial DNA fragments containing the encephalomyocarditis virus internal ribosomal entry site (IRES) sequence, enhanced green fluorescent protein (EGFP) cDNA, and puromycin resistance (PURO) cDNA were assembled in silico using publicly available DNA sequences, as follows. A 5′ EcoRI site preceded the IRES sequence, which was immediately followed by a SfiI site flanking the 5′ end of the EGFP coding sequence. The initiator methionine codon of EGFP was embedded in the SfiI site. The codon for the last amino acid of EGFP was immediately followed by an NheI site, which immediately preceded the 5′ end of an artificial cDNA-encoding human codon-optimized picornavirus 2A (P2A)-PURO resistance fusion gene. An XhoI site immediately followed the stop codon of the P2A-PURO cassette. This fragment was synthesized as a G Block by Integrated DNA Technologies (IDT, Coralville, IA.). Synthetic DNA was digested with EcoRI and XhoI and ligated to equivalently digested pBluescript SK(+) (Stratagene). Multiple clonal isolates were subjected to automated DNA sequencing with 5′ M13R and 3′ T7 promoter primers. A single clone perfectly matching the DNA sequence was digested preparatively with EcoRI and XhoI; a liberated insertion was isolated and ligated to equivalently digested pH110 (59). The resultant construct is referred to as pH163-EGFP-PURO. Th efunctionality of pH163 EGFP PURO was first tested for production of virus that could transduce Jurkat cells to EGFP positivity and puromycin resistance (see details below), and the vector backbone was subsequently used as a basis to create additional constructs encoding different combinations of fluorescence markers and antibiotic resistances, as follows. SfiI/NheI fragments corresponding to mCLOVER3, DsREDII, mAPPLE, mSCARLET, enhanced blue fluorescent protein 2 (EBFPII), mTagBFPII, enhanced yellow fluorescent protein (EYFP), mCITRINE, CERULEAN, mKATE1.3, SMurfBV+, firefly luciferase, or Streptococcus pyogenes Cas9 were designed in silico such that noncoding substitutions were made to eliminate any internal NotI, EcoRI, SfiI, NheI, or XhoI sites. Codons were also optimized for human-adaptive index determinations on a case-by-case basis, as necessary. mCLOVER3, mSCARLET, mTagBFPII, mKATE1.3, and SMurfBV+ fragments also encoded an amino terminal V5 epitope tag, useful for detection of the recombinant protein in cellular extracts via Western blotting. Synthetic G Block DNA was digested with SfiI/NheI and used to replace the equivalent EGFP fragment from H163 EGFP PURO. Insertion DNA was verified by automated DNA sequencing, and constructs were tested for functionality according to levels of viral production and transduction/expression within Jurkat cells of the respective fluorescent proteins, along with resistance to puromycin.
NheI/XhoI fragments corresponding to P2A-HYGRO, P2A-NEO, P2A-ZEO, and P2A-BLAST were designed in silico according to the considerations described above, and synthetic DNAs were used to replace the equivalent P2A-PURO cassette in H163-EGFP-PURO. Individual clonal constructs were validated/tested for the ability to produce virus functional for transduction of Jurkat cells to EGFP positivity and resistance to hygromycin B, G418, Zeocin, or blasticidin.
Individual clones conferring the appropriate fluorescent protein expression in combination with PURO selection, or antibiotic resistance in combination with EGFP expression, were used to isolate the functionally validated and relevant SfiI/NheI or NheI/XhoI fragment. The isolated functional DNA fragments were used to reconstitute the desired combination of fluorescent marker and antibiotic resistance in the H163 vector backbone, as depicted in Fig. S1 in the supplemental material.
(iv) cDNAs and tagged constructs.Subcloning of the 375-amino-acid (aa) human LDB1 cDNA was described previously (7); wild-type cDNA and mutant derivatives were arranged as either 5′ NotI/3′ EcoRI or 5′ BamHI/3′ EcoRI fragments. Vector-embedded epitope tags appended to LDB1 constructs were N-terminally located and were either tandem biotin acceptor domain (BAD)/FLAG (MAGGLNDIFEAQKIEWHEGGENLYFQGGDYKDDDDKGGAAASKVRS, FLAG peptide underlined) or HA×1 (MYPYDVPDYAGG). The 158-aa wild-type human LMO2 cDNA and mutant derivatives were synthesized as G Blocks with tandem 5′ NotI/BamHI and 3′ EcoRI sites and ligated into NotI/EcoRI-digested pBluescript II SK(+). The LMO2 cDNA encoded tandem C-terminal HA (GGMYPYDVPDYA) and SII (GGWSHPQFEK) tags. cDNAs encoding wild-type or mutant human 331-aa TAL1, 280-aa LYL1, 361-aa SSBP2, and 388-aa SSBP3 were all synthesized as G Blocks with 5′ NotI/BamHI and 3′ EcoRI sites and ligated into NotI/EcoRI-digested pBluescript II SK(+). Sequence encoding an N-terminal HA×1 tag ( MYPYDVPDYAGG) was located between the 5′ NotI and BamHI sites, and the BamHI site immediately preceded the natural initiator methionine codon. In order to create lentiviral vectors encoding subunits with BAD/FLAG, HA/SII, or HA×1 tags, clonally derived NotI/EcoRI fragments encoding BAD/FLAG-LDB1, LMO2-HA/SII, HA×1-TAL1, HA×1-LYL1, HA×1-SSBP2, or HA×1-SSBP3 were transferred from pBluescript II SK(+) vectors into likewise-digested H163 vectors. The N-terminal 312-aa Halo tag sequence was PCR amplified from His6HaloTag T7 Vector pH6HTN (Promega) as a 5′ SpeI, 3′ BamHI/EcoRI fragment and ligated into SpeI/EcoRI-digested pBluescript II SK(+); the resultant vector was named pHalo-tag-N. Tandem TGA stop codons were located between the BamHI and EcoRI sites. N-terminal Halo fusion constructs were created by ligating clonally derived BamHI/EcoRI fragments encoding LDB1, LMO2, TAL1, LYL1, SSBP2, or SSBP3 into equivalently digested pHalo-tag-N. In order to create lentiviral vectors encoding N-terminal Halo fusions, NotI/EcoRI fragments were recovered from these pHalo-tag-N vectors and ligated into likewise-digested H163 vectors in order to create H163-Halo-tag-N subunit vectors. All recombinant DNA manipulation and propagation utilized Escherichia coli XL1 Blue. All clonal insertions were verified in their entirety by automated DNA sequencing. All mutant derivatives used optimal human codons to encode amino acid substitutions. Maxipreps of lentiviral vector DNA were prepared for transfection/virus production by a modified alkaline lysis–lithium chloride-polyethylene glycol (PEG) precipitation protocol in conjunction with extensive phenol-chloroform extraction and ethanol precipitation. Additional details regarding constructs or protocols are available upon request.
(v) Whole-cell extract.Late-log-phase cultures of ∼7.5 × 107 cells were harvested by centrifugation at 800 × g for 10 min, and cell pellets were washed with phosphate-buffered saline (PBS; 2.7 mM KCl, 1.47 mM KH2PO4, 8.1 mM Na2HPO4, 137 mM NaCl) and resuspended in 500 to 1,000 μl extraction buffer (20 mM HEPES [pH 7.6], 300 mM NaCl, 20 mM imidazole, 0.1% Triton X-100, 10% glycerol, and protease inhibitor cocktail [Thermo/Pierce]). Cells were disrupted by mild sonication with the microtip of a Branson model 250 sonifier on the low-power setting, and the soluble extract was clarified by centrifugation at 14,000 × g for 15 min. Extract protein content was typically 5 to 10 μg/μl.
(vi) Immunoprecipitations.For immunoprecipitations (IP), 100 μl of soluble extract was supplemented with an additional 100 μl of extraction buffer also containing 5 μl anti-FLAG M2 resin (catalog number A2220; Sigma) or 5 μl of protein A/G resin (Santa Cruz) along with 1 to 2 μg of anti-LMO2 IgG and was then rocked at 4°C for 3 to 4 h. Immune complexes were isolated by centrifugation, washed 3 times with 200 μl of extraction buffer, and eluted by heating with 100 μl SDS sample buffer. Samples were stored at –80°C and briefly heated again at 75°C just prior to loading onto handcast discontinuous SDS-PAGE gels with a 4% acrylamide stacking gel and a 4%-to-15% linear gradient resolving gel (37.5%/1.0% [wt/vol] acrylamide-bisacrylamide), run at 15 V/cm for 90 to 105 min. Gels were transferred onto a 0.2-μm-pore-size polyvinylidene difluoride (PVDF) membrane (catalog number 10600022; GE) at 50 V for 2.5 h; filters were blocked in PBS–2% nonfat dry milk (NFDM; Marsh FoodClub) and incubated with antibodies in blocking buffer overnight at 4°C.
(vii) SDS-PAGE/Western blotting.Western blots were developed with enhanced chemiluminescence (ECL) detection (SuperSignal Pico West Plus, catalog number 1863099; Thermo/Pierce). All images were obtained within the linear signal detection range using a ChemiDoc Touch imaging system (Bio-Rad). Images were analyzed using ImageLab software version 5.2.1 (Bio-Rad) and exported to Adobe Photoshop and Illustrator for figure assembly.
(viii) HaloLife: live-cell pulse-chase.Cells (1.25 × 105) were collected from log-phase cultures by centrifugation at 1,200 × g for 1 min. The culture medium was removed, and cells were resuspended with 125 μl RPMI 1640 containing 10% FBS and HaloTag Ligand R110 (Promega Ca.) at a final concentration of 100 nM, per the manufacturer’s instructions. The resuspended cells were then incubated for 90 min at 37°C in 5% CO2. After 90 min, the cells were centrifuged at 12,000 × g for 1 min and washed with PBS containing 0.1% bovine serum albumin (BSA) a total of 3 times to remove excess HaloTag Ligand R110. Cells were resuspended in 600 μl RPMI 1640 containing 10% FBS, and four 150-μl aliquots were transferred to a 96-well round-bottom plate (TPP). A total of 10,000 events were then immediately analyzed from one of the four 150-μl aliquots using a CytoFLEX benchtop cytometer (Beckman). Data representing all subsequent chase time points were collected using this initial analysis as a reference. Between flow cytometry analyses, the 96-well plate containing the HaloTag Ligand R110-labeled cells were placed in an incubator at 37°C with 5% CO2 until the next collection point. Flow cytometry analyses were collected 3, 4, and 5 h after time zero (t0) for all cells, with the exception of those containing Halo-tagged LDB1 and LYL1 due to their significantly different observed half-lives. Flow cytometry events were recorded at 6, 12, and 24 h after t0 for cells containing Halo-tagged LDB1, and analyses were recorded 1, 2, and 3 h after the initial time point for cells containing Halo-tagged LYL1. Replicate experiments were done on consecutive days.
(ix) ImageStream.A total of 1.25 × 105 cells were collected from log-phase cultures by centrifugation at 1,200 × g for 1 min. The culture medium was removed, and cells were resuspended with 125 μl RPMI 1640 containing 10% FBS and HaloTag Ligand R110 (Promega Ca.) at a final concentration of 100 nM, per the manufacturer’s instructions. The resuspended cells were then incubated for 90 min at 37°C in 5% CO2. After 90 min, the cells were centrifuged at 12,000 × g for 1 min and washed with PBS containing 0.1% BSA a total of 3 times to remove excess HaloTag Ligand R110. The cells were then resuspended in 1 ml PBS and stained with SYTO 17 red fluorescent nucleic acid stain (Invitrogen) at a final concentration of 10 nM for 10 min, per the manufacturer’s instructions. The cells were washed once more and were then resuspended with 200 μl PBS before being analyzed using an ImageStream Mark II imaging flow cytometer (Millipore Sigma). Data analysis was done using nuclear localization analysis Wizard in IDEAS 6.2 software (Millipore).
(x) Confocal imaging.A total of 1.25 × 105 cells were collected from log-phase cultures by centrifugation at 1,200 × g for 1 min. The culture medium was removed, and cells were resuspended with 125 μl RPMI 1640 containing 10% FBS and HaloTag Ligand R110 (Promega Ca.) at a final concentration of 100 nM and were then incubated for 90 min at 37°C in 5% CO2. After 90 min, the HaloTag Ligand R110-labeled cells were centrifuged at 1,200 × g for 1 min and washed with PBS containing 0.1% BSA a total of 3 times to remove excess ligand. The washed cells were then resuspended in 1 ml of PBS and stained with SYTO 17 red fluorescent nucleic acid stain (Molecular Probes, Inc., OR) according to the manufacturer’s protocol. After the incubation period, the cells were centrifuged at 1,200 × g for 1 min and washed once with PBS. Once resuspended in 300 μl of PBS, cells were transferred to a 12-mm-diameter glass base dish and imaged with a Leica TCS SP8 confocal imaging system (Leica Microsystems Inc., IL) using an HC PL APO 40×/1.3 oil CS2 lens objective. Digital images were rendered, and signal intensities were analyzed using Imaris visualization and analysis software (Bitplane Inc., MA). Cellular localization of HaloTagged proteins was determined by calculating the ratio of mean HaloTag signal intensity within the nucleus versus that in the cytosol. The nuclear area was established using SYTO 17 red fluorescent nucleic acid stain, and the cytoplasmic region was determined using the diffuse EBFPII signal expressed by our lentiviral vectors.
(xi) Determining cellular abundances via in-gel fluorescence.A total of 2.7 × 108 cells were collected from log-phase cultures by centrifugation at 400 × g for 5 min. The culture medium was removed, and cells were resuspended with 5 ml RPMI 1640 containing 10% FBS and HaloTag Ligand R110 (Promega Ca.) at a final concentration of 100 nM and were then incubated for 90 min at 37°C in 5% CO2. After 90 min, the cells were centrifuged at 400 × g for 5 min and washed with RPMI 1640 containing 10% FBS a total of 3 times to remove excess HaloTag Ligand R110. Whole-cell extracts were made following the previously described protocol, but urea buffer (8 M) was used in place of the described extraction buffer. Cell equivalents of 1.35 × 106 cells of whole-cell lysates were then loaded onto an agarose gel and visualized on a ChemiDoc Touch imaging system (Bio-Rad) using optimum autoexposure settings for an Alexa Fluor 488 emission filter. Band intensities were quantified using Image Lab (Bio-Rad), and relative band intensities were calculated using the Halo-CTCF U2OS band as reference.
(xii) Determining cellular abundances via flow cytometry.A total of 1.0 × 106 cells were collected from log-phase cultures by centrifugation at 1,200 × g for 1 min. Culture medium was removed, and cells were resuspended in 1.5 ml of RPMI 1640 containing 10% FBS and HaloTag Ligand R110 (Promega Co.) at a final concentration of 100 nM. Resuspended cells were incubated for 90 min at 37°C in 5% CO2. After incubation, cells were centrifuged at 12,000 × g for 1 min and washed with PBS containing 0.1% BSA a total of 3 times to remove excess HaloTag Ligand R110. Cells were then resuspended in 200 μl of fresh RPMI 1640, transferred to a 96-well plate, and analyzed using a CytoFLEX benchtop cytometer (Beckman). Live cells were gated using forward scatter (FSC) and side scatter (SSC), HaloTag Ligand R110 was excited using a 488-nm-wavelength laser, and emission was read out using a FITC-A 525/40 fluorescence channel.
(xiii) mRNA sequencing.The transduced cells shown in Fig. 9A were washed and lysed and subjected to whole-RNA isolation. RNA was analyzed by Bioanalyzer. RNAs with values above 5 were used to build strand-specific cDNA libraries and subject to paired-end sequencing on a HiSeq 4000 system (Illumina). The RNA-seq data were quality controlled following the multiperspective guidelines (60). Quality control metrics were computed from QC3 software (61). No quality issues were observed. The alignment was performed by the use of TopHat 2 (62) on the GRCh38 human reference genome. Read counts per gene were generated using HTSeq (63). Differential expression analyses were performed using the MultiRankSeq R package (64). P values were adjusted the using Benjamini-Hochberg method.
Quantification and statistical analysis. (i) Pulse-chase FCS file analysis.All fluorescence correlation spectroscopy (FCS) files were analyzed using Flowjo 10.3 analysis software (Flowjo LLC, OR). To identify cells that were coexpressing EBFPII and/or mScarlet in conjunction with Halo-tagged proteins, nontransduced unstained Jurkat cells were used to establish a gating sequence. Their physical dimensions were grouped on an FSC-A/FSC-H plot to determine the total number of lymphocytes within the event population. A gate was then established on an FSC-A/SSC-A plot to select for live cells within the total lymphocyte population. The resulting population was then gated as a negative control for both fluorescence markers on a PB450-A (EBFPII)/FITC-A (HaloTag R110) plot. This gating sequence was then applied to all FCS files within the same experiment.
(ii) Half-life calculations.Log-linear regression curves were calculated from flow cytometry analysis data to determine Halo-tagged protein half-lives. PB450-A (EBFPII) and FITC-A (HaloTag R110 Ligand) doubly positive events were calculated as percentages of the parent population for all time points collected. Replicate data for each time point were averaged and then normalized to the initial time point. The natural log was calculated for each of the averages, and the resulting values were represented over time on a 2-dimensional scatterplot. A trend line was calculated, and the resulting slope was used to determine Halo-tagged protein half-lives.
(iii) Statistical analysis.The standard error of the mean (SEM) was calculated for individual time points in each Halo-tagged protein experiment using Microsoft Excel. SEM values were then applied to their corresponding time points within the log-linear regression curves used to determine Halo-tagged protein half-lives. Results from replicate experiments were used to calculate the standard deviation, which was then divided by the square root of the number of replicates to determine the SEM. The SEMs for Halo-tagged protein half-lives values were also calculated using the same formula. Half-life values were determined for at least 3 experiments as previously described and were then used to calculate the SEM.
(iv) Determining cellular abundances via in-gel fluorescence.Using Halo-CTCF U2OS cells as the reference, relative band intensities were then quantified as percentages of the reference using Image Lab (Bio-Rad). Cellular abundance was then determined by multiplying the relative band intensities by the number of molecules per cell of Halo-CTCF as previously reported by Cattoglio et al. (28).
(v) Determining cellular abundances via flow cytometry.Raw FCS files were analyzed using Flowjo 10.3 analysis software (Flowjo LLC, OR), and experimental replicates for individual cell populations were concatenated into single populations of 100,000 events prior to gating. Live cells were gated using forward scatter and side scatter, and mean fluorescence intensity values were measured for the FITC-A fluorescence channel. The mean fluorescence intensity value for Halo-CTCF-containing U2OS cells was used to normalize the other cell populations, and protein abundance was determined using standards and methods previously reported by Cattoglio et al. (28).
Data availability.Inquiries for further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding author.
ACKNOWLEDGMENTS
We thank Yuichiro Takagi, Sabine Wenzel, Fang Huang, Mark Goebl, Merv Yoder, Reuben Kapur, and Jörg Bungert for helpful discussions. We acknowledge the contributions of the members of the IUSCC Flow Cytometry Core, which was supported in part by NIDDK Cooperative Center of Excellence in Hematology (CCEH) grant U54 DK 106846, and RNA-seq studies were carried out in the Center for Medical Genomics at Indiana University School of Medicine, which is partially supported by the Indiana Genomic Initiative at Indiana University (INGEN); INGEN is supported in part by the Lilly Endowment, Inc. J.H.L. and U.P.D. both thank Steve Brandt, Mark Koury, and Ray Mernaugh for all of their guidance and help through the years. J.H.L. acknowledges and thanks P. A. Weil for consistently imparting the value and power of thoughtful experimental controls to every one of his many trainees.
This work was supported in part by Merit Review Award no. I01BX001799 from the United States Department of Veterans Affairs, Biomedical Laboratory Research and Development Service, and by R01CA207530 from the National Cancer Institute, awarded to U.P.D. Additional support was provided to U.P.D. by the Strategic Research Initiative of the Indiana University School of Medicine. J.H.L. was a recipient of the Biomedical Research Grant from Indiana University.
J.H.L. conceived and designed the entire study and all experiments therein; designed, prepared, and verified all recombinant DNA constructs; produced and verified all recombinant lentiviral stocks; produced, verified, and maintained all stable cell lines; prepared all RNA samples; performed all Western blotting and immunoprecipitations; analyzed the data; prepared all of the figures; and cowrote the manuscript. M.C. performed all confocal microscopy, implemented and optimized the HaloLife assay, performed all HaloLife assays, and analyzed all flow cytometry data. L.P. and D.U. assembled and distributed all of the lentiviral shRNA bacterial stocks. Y.G. performed bioinformatic analyses of RNA-seq data. U.P.D. interpreted data, contributed to experimental design, secured all funding, oversaw the entire study, and assembled and wrote the manuscript.
U.P.D. and J.H.L. have filed for a patent for the lentiviral vector system described. There are no other competing interests.
FOOTNOTES
- Received 23 December 2019.
- Returned for modification 22 January 2020.
- Accepted 14 March 2020.
- Accepted manuscript posted online 30 March 2020.
Supplemental material is available online only.
- Copyright © 2020 American Society for Microbiology.
