Main

APL is characterized by a specific t(15;17) translocation that encodes a fusion of the promyelocytic leukemia (PML) and retinoic acid receptor-α (RARA) proteins. PML-RARA is a transcriptional repressor with both gain-of-function and dominant-negative properties, resulting in transcriptional repression of retinoic acid and non–retinoic acid target genes1,2 and culminating in differentiation arrest. Gene silencing involves enhanced recruitment of nuclear receptor co-repressors, the polycomb complex or Daxx onto PML-RARA or its obligatory retinoid X receptor-α (RXRA) partner, resulting in changes in chromatin organization and DNA methylation3,4,5,6,7,8,9.

APL is a unique model in cancer biology in that two therapeutic agents, retinoic acid and arsenic trioxide, both target PML-RARA10,11. In vivo, both drugs induce leukemia differentiation to varying extents, making APL the only cancer that can be treated through differentiation therapy12. High plasma concentrations of retinoic acid are essential for prolonged clinical responses in patients3,13. At the molecular level, retinoic acid binds PML-RARA and turns it into a transcriptional activator, but also triggers its degradation3,14,15. PML-RARA degradation depends on PML-RARA cleavage by differentiation-activated proteases14,16 and on recruitment of proteasomes by the retinoic acid–bound RARA moiety of PML-RARA15,17. Arsenic activates kinases targeting PML-RARA, or its RXRA partner, and induces PML-RARA degradation through specific SUMOylation and subsequent ubiquitination of its PML moiety5,18,19,20,21,22. The respective contributions of drug-induced differentiation, apoptosis, transcriptional activation and PML-RARA degradation to APL eradication have been a matter of debate. The molecular pathogenesis of APL and the bases for its response to therapy have been studied in PML-RARA transgenic mice, which recapitulate responses to therapy in individuals with APL23,24,25,26. In particular, the combination of retinoic acid and arsenic triggers prolonged remission in mouse models and humans. However, although the two drugs synergize to effect APL eradication in vivo, this does not seem to result from enhanced differentiation23,24,27,28,29.

Retinoic acid–induced APL differentiation and PML-RARA transactivation are enhanced by cAMP signaling, and many retinoic acid–resistant APLs differentiate after exposure to retinoic acid and cAMP1,30,31. At the molecular level, cAMP activates protein kinase A (PKA), which dissociates RARA from the SMRT co-repressor, allowing transcriptional activation by retinoic acid metabolites through RXRA1,32. RARA phosphorylation by cAMP-activated PKA on Ser369, within the ligand binding domain, also allows recruitment of the cdk7–cyclin H subcomplex of transcription factor IIH and phosphorylation-induced activation of the ligand-independent AF1 transactivating domain33. The importance of these events for the APL response to therapy is unknown.

Recent evidence has indicated that not all cancer cells are identical: only a minority have the ability to regenerate new tumors and, hence, control transplantability and metastasis development34. PML-RARA promotes immortalization of myeloid progenitors ex vivo and in vivo3,5,35, but the effect of retinoic acid or arsenic on APL LICs has never been analyzed in vivo. Here we show that PML-RARA degradation is tightly regulated by a PKA phosphorylation site in PML-RARA and that PML-RARA degradation triggers LIC loss, which—rather than differentiation—is responsible for eradicating APL in mice.

Note: Supplementary information is available on the Nature Medicine website.

Results

Differentiation by retinoic acid spares clonogenic cells

Transduction of PML-RARA DNA into mouse hematopoietic progenitor cells allows them to be replated indefinitely in methylcellulose, with only 1% of the transformed cells being clonogenic and yielding promyelocytic colonies5. Growth in methylcellulose containing retinoic acid led to 8% of the original number of colonies in untreated cultures as well as to terminal granulocytic differentiation of those colonies that developed (Fig. 1a). Unexpectedly, when retinoic acid–treated cultures, or even individual colonies, were replated in methylcellulose without retinoic acid, many dense promyelocytic colonies developed that could be serially replated, qualitatively similar to the untreated parental cells (Fig. 1a). Retinoic acid treatment nevertheless resulted in quantitative differences in the frequency of clonogenic cells, which dropped from 1% to 0.3% (Fig. 1b). In untreated starting cultures, PML-RARA protein levels were comparable to those found in individuals with APL or transgenic mice (data not shown).

Figure 1: Retinoic acid (RA) triggers APL cell differentiation and loss of clonogenic precursors ex vivo and in vivo.
figure 1

(a) Ex vivo methylcellulose-cultured, PML-RARA–transformed primary hematopoietic progenitors (top) were cultured for 1 week with 10−6 M RA (bottom right) and subsequently without RA (bottom left). Numbers of colonies (mean ± s.d. of five independent experiments) and representative flow cytometry and MGG stains of cells recovered at various steps are shown. Scale bar, 10 μm. 5FU, 5-fluorouracil. (b) Cell composition of untreated and RA-treated cultures. (c–e) RA decreases LIC abundance in PML-RARA APL in vivo. (c) Left, bone marrow MGG staining of a normal mouse and an APL mouse untreated or treated with 10 mg of RA (RA 10) after 7 d; right, flow cytometry of the corresponding marrow cells using two myeloid differentiation markers, CD11b (Mac) and GR-1. (d) Spleen weight (mean ± s.d. of four experiments) and white blood cell counts (WBC; mean ± s.d. of seven experiments) after 1 week of treatment with 10-mg RA pellets. (e) Survival of mice inoculated with increasing amounts of APL cells (mean ± s.d. of two independent experiments, see also Supplementary Fig. 2b). (f) Bone marrow cells from RA-treated and untreated mice were injected into secondary syngeneic recipients. Abundance of PML-RARA DNA as determined by quantitative real-time PCR, and time to death of secondary recipients (mean ± s.d. of three independent experiments) on day 2 with 40 mg of RA (RA 40) or day 7 with 10 mg of RA. NS, not significant.

Full size image

Human APLs expressing the promyelocytic leukemia zinc finger (PLZF)-RARA fusion protein are resistant to retinoic acid therapy3. Unexpectedly, hematopoietic progenitor cells transformed ex vivo with PLZF-RARA behaved exactly as PML-RARA–transformed cells with respect to retinoic acid–induced differentiation, gene activation and growth inhibition of clonogenic cells (Supplementary Fig. 1 online). Such complete differentiation of PLZF-RARA–transformed cells suggests that differentiation is unlikely to be the sole determinant for clinical response of APL to retinoic acid. Together, these ex vivo experiments establish that retinoic acid triggers two distinct and uncoupled events: induction of terminal differentiation of promyelocytes and transient growth arrest of clonogenic progenitors (Fig. 1b).

Retinoic acid decreases APL LIC frequency in vivo

Ex vivo transformation may not fully reflect APL pathogenesis. We thus turned to an APL transplantation model derived from mice expressing the Tg(PML-RARA)935Kog human transgene (referred to here as MRP8–PML-RARA transgenic mice)23,31. To allow for rapid and reproducible APL development, we injected 107 APL blasts intravenously into syngeneic recipients and then treated well-established APLs with retinoic acid for up to 7 d before analyzing therapy response. Retinoic acid treatment led to the rapid in vivo differentiation of APL cells, followed by their replacement by normal marrow (Fig. 1c–f)23,24,25,26. Retinoic acid also normalized spleen weight and blood counts (Fig. 1d), although relapses (assessed by luciferase imaging; Supplementary Fig. 2a online) always occurred weeks later23,24. When we inoculated this morphologically normal retinoic acid–treated day 7 marrow into secondary recipients and monitored their survival, we found that they developed APL, demonstrating the presence of remaining LICs in the inoculate. Yet comparison with marrow inoculates from untreated mice indicated that LIC abundance was greatly reduced (Fig. 1e, f). Because this reflects in part our inoculation of a smaller number of APL cells (assessed by PML-RARA DNA), we treated mice for 2 d with high doses of retinoic acid (40 mg) and compared time to death in recipients inoculated with similar numbers of treated or untreated APL cells. A highly significant survival difference was observed, indicating that only 1% of LICs remain after a 2-d, 40-mg treatment with retinoic acid (Fig. 1e, f and Supplementary Fig. 2b). We cannot formally exclude the possibility that retinoic acid induces transient cell-cycle arrest of LICs, delaying their regrowth in secondary recipients, but the normal size of secondary colonies derived from retinoic acid–treated methylcellulose ex vivo strongly argues against this hypothesis.

Low-dose retinoic acid triggers differentiation but not LIC loss

Apart from PML-RARA mutations, low plasma retinoic acid concentrations, linked to retinoic acid–induced catabolism, often account for resistance to retinoic acid3,13. At doses of 10 mg, plasma retinoic acid levels reached 10−7 M after 3 d, and 9-cis-retinoic acid, the main retinoic acid metabolite that activates RXRs, reached 2 × 10−8 M. Treatment with 1.5 mg of retinoic acid triggered full differentiation after 3 d, with plasma retinoic acid at 1.5 × 10−8 M (Fig. 2a). However, on day 7, promyelocytes were still present in the marrow, spleen size never normalized and normal hematopoiesis was not restored (Fig. 2a, b). Transplantation of marrow from mice treated with 1.5 mg retinoic acid into secondary recipients did not significantly affect time to death (Fig. 2c). Thus, immediate differentiation can be uncoupled from the decrease in LIC frequency, and the latter is likely to be responsible for retinoic acid–induced clinical remission.

Figure 2: Full differentiation without APL clearance in PLZF-RARA APL or PML-RARA APL treated with low-dose retinoic acid (RA).
figure 2

(a) MGG-stained bone marrow (left) and flow cytometry (right) of PML-RARA APL cells from mice treated with high (RA 10) or low (RA 1.5) doses of RA for the indicated times. Arrows indicate remaining APL blasts. Scale bar, 10 μm. (b) Spleen weight and PML-RARA DNA content of mice after 7 d of treatment, as indicated (mean ± s.d. of three experiments). (c) Time to death in secondary recipients inoculated with bone marrow cells treated or not treated with RA for 7 d (mean ± s.d. of three independent experiments). (d) Bone marrow staining (left) and flow cytometry analysis (right) of PLZF-RARA mouse APLs treated for 12 days with 10 mg of RA. Terminal granulocytic differentiation was observed. Arrow indicates the absence of normal hematopoiesis restoration. (e,f) Spleen weight (e) and white blood cell counts (WBC; f) of untreated or RA-treated PLZF-RARA APL after 7 d (mean ± s.d. of three experiments). (g) Abundance of ZBTB16-RARA DNA by quantitative PCR after 12 d of RA treatment. (h) Time to death of secondary recipients injected with 3 × 106 bone marrow cells collected from primary recipients after 10 d of treatment with or without RA (mean ± s.d. of four independent experiments).

Full size image

Retinoic acid differentiates PLZF-RARA APL without LIC loss

Mice transgenic for both PLZF-RARA and RARA-PLZF (referred to here as PLZF-RARA–RARA-PLZF transgenic mice) develop an APL-like disease36. We established a transplantation model of these APL cells in nude mice and, unexpectedly, found that retinoic acid induced terminal granulocytic differentiation (Fig. 2d). However, in contrast to our observations in PML-RARA APLs, restoration of normal hematopoiesis in PLZF-RARA–RARA-PLZF mice was undetectable by flow cytometry (Fig. 2d), May-Grünwald-Giemsa (MGG) staining or ZBTB16-RARA (PLZF is now known as ZBTB16) DNA clearance (Fig. 2). Spleen weight remained elevated, even 2 weeks after treatment, and blood counts increased markedly, reminiscent of the retinoic acid syndrome3 (Fig. 2e, f).

Transplantation of retinoic acid–treated (granulocytic) or untreated (blastic) marrow into secondary recipients consistently resulted in the same kinetics of leukemia development (Fig. 2g, h). Retinoic acid treatment prolonged survival of primary recipients37 (Supplementary Fig. 3a online). Notably, in a human subject with t(11;17) APL, retinoic acid similarly led to a significant decrease in marrow blast counts, which was immediately reversed after withdrawal of retinoic acid (Supplementary Fig. 3b). This situation exemplifies the complete dissociation between terminal retinoic acid–induced differentiation and the eradication of LIC or APL regression, indicating that retinoic acid–induced differentiation is not the basis for APL remission.

LIC loss by retinoic acid and arsenic requires proteolysis

Arsenic triggers some differentiation and APL regression in mice23,24,31 (Fig. 3a), although it does not seem to be as efficient in humans, possibly because of a different delivery route (intraperitoneal versus intravenous, respectively) and/or pharmacokinetics. In contrast to retinoic acid alone, the retinoic acid–arsenic association triggers rapid and complete remission in most mouse models of APL and most individuals with APL23,24,27,28. Follow-up studies in mice and with human subjects from two independent clinical trials suggests that they are actually cured, even in the absence of chemotherapy12,23 (Z. Chen (Shanghai Institute of Hematology) and E. Estey (Washington University), personal communications). In mice, retinoic acid and arsenic did not synergize for differentiation on day 3 (Fig. 3a), but they synergized markedly for both disease regression and LIC loss on day 3 or 6 (Fig. 3a–d). The retinoic acid–arsenic synergy for LIC eradication on day 3 was shown with a similar abundance of PML-RARA–positive cells in the bone marrow of primary recipients treated with retinoic acid alone or retinoic acid and arsenic, implying that this combination specifically targets APL LICs.

Figure 3: LIC eradication by retinoic acid (RA) and arsenic depends on active proteolysis.
figure 3

(a–e) RA and arsenic synergize for LIC eradication. (a) MGG staining (left) and flow cytometry (right) of cells from mice left untreated or treated with 10 mg of RA or arsenic for 3 d. Normal non-myeloid hematopoiesis was restored in marrow treated with RA only (arrow) but not with RA and arsenic. Scale bar, 10 μm. (b) Luciferase imaging of APL mice treated for 3 d (left) and PML-RARA DNA quantification of the corresponding marrow (right). Scale bar, 1 cm. (c) Survival of the secondary transplants of day 3 marrow (left) and corresponding follow-up of luciferase activity (right). Representative of three experiments. (d) Day 6 spleen weight (mean ± s.d. of three independent experiments) and survival of secondary day 6 marrow recipients (representative of three independent experiments). The three secondary recipients of RA- and arsenic-treated marrow were negative for PML-RARA DNA after a year. (e) Effects of RA and/or arsenic on PML-RARA protein levels in mouse APL cells after 12 h of ex vivo treatment. Actin served as a control. (f–h) The RA-arsenic synergy for LIC clearance depends on active proteolysis. (f) Bone marrow MGG staining (top) and flow cytometry (bottom) after 6 d of treatments. Arrows indicate lack of normal hematopoiesis recovery when bortezomib was combined with RA-arsenic. (g) Ex vivo PML-RARA degradation in mouse APL cells by RA and arsenic (both 10−7 M) is reversed by the proteasome inhibitor MG132 (PML-RARA is detected with an antibody to RARA; actin served as a control). (h) PML-RARA DNA in the marrow (top) and survival of secondary transplants that received marrow cells from APL mice treated for 3 d (bottom) (representative of two independent experiments).

Full size image

As expected from their nonoverlapping pathways11,15,18,20,21, retinoic acid- and arsenic-triggered PML-RARA degradation were synergistic and proteasome dependent in ex vivo–treated mouse APL cells (Fig. 3). To address the role of PML-RARA degradation in LIC clearance, we treated mice with retinoic acid and arsenic in the presence or absence of the proteasome inhibitor bortezomib (Fig. 3f). Although bortezomib induced some differentiation and APL regression on its own, when combined with retinoic acid and arsenic, it consistently delayed APL clearance, blocked restoration of normal hematopoiesis on day 6 and sharply antagonized loss of LICs on day 3 (Fig. 3f–h). Bortezomib also significantly blocked recovery of normal hematopoiesis by retinoic acid alone (Supplementary Fig. 4 online). Bortezomib may also modulate other pathways in this complex biological system, but these results strongly suggest that PML-RARA degradation is a major molecular determinant of LIC clearance.

Retinoic acid–cAMP clears LICs in sensitive APLs

When we combined cAMP with 1.5 mg of retinoic acid, we noted a marked in vivo synergy for disease regression and restoration of normal hematopoiesis, but not for immediate differentiation, on day 3 (Fig. 4a–e). H89, an inhibitor of PKA, substantially reversed this effect of cAMP (Fig. 4d). When combined with 1.5 mg of retinoic acid, particularly with arsenic, phosphodiesterase inhibitors (PDEIs) consistently decreased the frequency of LICs in secondary transplantation experiments (Fig. 4e) without changing APL cell abundance in the inoculates (data not shown). Thus, cAMP synergizes with low-dose retinoic acid for LIC eradication, but not immediate differentiation.

Figure 4: Activation of cAMP signaling synergizes with low-dose retinoic acid (RA) to induce LIC loss in wild-type PML-RARA but not in RA-resistant PML-RARAL902P APL.
figure 4

(a) Day 7 spleen weight (mean ± s.d. of six different experiments). (b) Luciferase quantification of PML-RARA APL mice treated with 10 or 1.5 mg of RA, with or without cAMP, for 1 to 6 d. Asterisk indicates a mouse that died of leukemia progression. Representative of three independent experiments. (c) Flow cytometry on day 3 after treatment, showing similar differentiation for all treatments. (d) Bone marrow MGG staining of mice treated as indicated. PKA inhibitor H89 blocked APL recovery of normal hematopoiesis after treatment with RA-cAMP on day 7. Arrows indicate remaining promyelocytes. Scale bar, 10 μm. (e) Survival of secondary transplant recipients from primary APL mice treated as indicated. In this experiment, cAMP signaling was triggered by a PDEI. Six secondary mice, four inoculated with primary day 3 marrows and two with day 6 marrow, were all negative for PML-RARA DNA after 1 year, as assessed by PCR. Representative of three independent experiments. (f–i) RA-cAMP differentiates RA-resistant PML-RARAL902P APL31 but does not degrade PML-RARA or induce LIC loss. (f) PML-RARAL902P does not bind RA. LBD, ligand binding domain. (g) Left, MGG staining of control and RA-cAMP–treated mice on day 7; right, corresponding spleen weights (mean ± s.d. of six independent experiments). (h) PML-RARA DNA content after 7 d of the indicated treatments (mean ± s.d. of three experiments) and survival of secondary recipients of marrow treated for 7 d (mean ± s.d. of five independent experiments). (i) Analysis of PML-RARA protein after ex vivo treatment with 10−6 M RA and PDEI, alone or together, for 12 h. Actin served as a control.

Full size image

The retinoic acid–cAMP combination differentiates retinoic acid–resistant APLs expressing PML-RARAL902P (Tg(PML-RARAL902P)4048Kog) which is defective for retinoic acid binding1,31(Fig. 4f, g), presumably through RXRA activation after in vivo isomerization of retinoic acid into 9-cis-retinoic acid1,8,9,32. Despite efficient differentiation, retinoic acid–cAMP did not decrease LIC abundance, nor did it significantly degrade PML-RARAL902P (Fig. 4f–i). Such uncoupling between efficient differentiation and absence of APL clearance suggests that differentiation reflects transcriptional activation, whereas changes in LIC frequency mirror PML-RARA catabolism.

Phosphorylation of PML-RARA is essential for LIC loss

cAMP-activated PKA phosphorylates RARA on Ser369 (corresponding to PML-RARA Ser873; ref. 33). After exposure to retinoic acid, PML-RARA Ser873 becomes phosphorylated (C.R.-E., unpublished data, and Fig. 5). We generated MRP8–PML-RARAS873A transgenic mice to address the role of this residue in retinoic acid response and retinoic acid–cAMP synergy (Fig. 5). One founder and its offspring eventually yielded two independent APLs with features identical to those previously obtained with PML-RARA5,8,23,31. In a transplantation model, initial retinoic acid–induced differentiation was maintained on day 3, but the mutant APLs all showed a markedly impaired response to retinoic acid on day 7, which most likely reflects ongoing LIC proliferation (Fig. 5). Indeed, the marrow remained leukemic, normal hematopoiesis never reappeared and blood counts rose sharply (Fig. 5d, e). As assessed by PCR, the marrow remained entirely leukemic, whatever the treatment combination, including retinoic acid–arsenic, even when treatments were prolonged after 7 d (Fig. 5f and data not shown). Arsenic-triggered differentiation was maintained, and these APLs regressed with cAMP, consistent with the existence of PML-RARA–independent cAMP growth-suppressive pathways32. Synergy between cAMP and retinoic acid or arsenic for differentiation31 was maintained on day 6, possibly reflecting release of SMRT independently of RARA phosphorylation32. However, LIC clearance by the various drug associations was substantially diminished (Fig. 5g), establishing that Ser873 phosphorylation is crucial for retinoic acid–induced LIC loss.

Figure 5: Loss of the Ser873 PKA phosphorylation site in PML-RARA results in retinoic acid (RA)-resistant APLs.
figure 5

(a) Mutated version of PML-RARA used to generate transgenic mice. (b) Spleen weight after 6 d of treatment with RA, cAMP and/or arsenic (mean ± s.d. of four independent experiments). (c) Flow cytometry of day 3 cells from corresponding mice, showing RA-induced differentiation. (d) MGG staining on day 6. Scale bar, 10 μm. (e) White blood cell counts (WBC) showing RA-induced leukocytosis, as in PLZF-RARA APLs. (f) PML-RARA DNA abundance on day 10. (g) Survival (mean ± s.d. of three experiments) of secondary recipients after 6 d of treatment. (h) RA-induced PML-RARA phosphorylation in retrovirally transduced mouse progenitor cells expressing PML-RARA (left) or PML-RARAS873A (right). Detection was carried out with an antibody33 recognizing Ser369-phosphorylated RARA (P-(S873); top); output of affinity column was blotted with antibody to RARA (bottom).

Full size image

Ser873 modulates PML-RARA degradation, not gene activation

In principle, retinoic acid resistance of these mutant leukemias could reflect altered retinoic acid–activated transcription and/or PML-RARA degradation. We first analyzed target genes activated in RARA-transduced Rara−/−; Rarb−/−; Rarg−/− mouse embryonic fibroblasts6 and found that the wild-type or phosphorylation mutant similarly activated expression of Rarb and Cyp26a1 in dose-response experiments (Fig. 6a). Rarb and Cyp26a1 (RARA and PML-RARA targets) and Tgm2 (a PML-RARA–specific target) were also normally induced by retinoic acid in PML-RARA or PML-RARAS873A APL cells (Supplementary Table 1 online). Thus, the absence of LIC loss does not reflect defective retinoic acid–dependent transcriptional activation by the mutant receptor, although we cannot rule out the possibility that the latter fails to regulate an unidentified subset of crucial PML-RARA target genes.

Figure 6: PKA phosphorylation site in RARA and PML-RARA is dispensable for retinoic acid (RA)-induced activation of target genes but desensitizes PML-RARA to RA-induced degradation.
figure 6

(a) Activation of Rarb and Cyp26a1 gene expression in RARA- or RARAS369A-transduced Rara−/−; Rarb−/−; Rarg−/− mouse embryonic fibroblasts after 6 h with the indicated RA concentrations. (b) PML-RARA degradation in mouse APLs triggered by 15 h of exposure to the indicated compounds ex vivo. Top, PML-RARAS873A is less sensitive to RA-induced degradation than wild-type PML-RARA (arrows; see Supplementary Fig. 5 for a lower RA concentration). Middle, synergistic effects of 10−7 M RA and a PDEI in normal APLs. Bottom, PDEI-arsenic combination results in additive PML-RARA or PML-RARAS873A degradation. (c) A new model of the mechanism of RA, arsenic and cAMP action in APL.

Full size image

In contrast, we consistently observed that PML-RARAS873A was less sensitive to retinoic acid–induced proteolysis (Fig. 6b). PDEIs induced baseline degradation (additive with that triggered by arsenic) for both PML-RARA and its S873A mutant (Fig. 6b). However, PDEIs strongly synergized with retinoic acid to enhance PML-RARA degradation in a Ser873-dependent manner, as previously suggested38 (Fig. 6b and Supplementary Fig. 5 online). Thus, PML-RARA phosphorylation at Ser873 regulates retinoic acid–triggered degradation of the fusion protein, most likely by enhancing proteasome docking onto ligand-activated RARA15,17.

Discussion

It was previously thought that the therapeutic benefit of retinoic acid in APL relied on the induction of terminal cell differentiation through retinoic acid–triggered reactivation of crucial targets repressed by PML-RARA3. This model was called into question by the fact that arsenic, despite being an excellent APL therapy27,39, only modestly induces differentiation in vivo, does not do so ex vivo and does not activate most PML-RARA targets23,29. Here, we establish that APL drugs trigger two separable biological events: differentiation and LIC clearance, with only the latter being associated with remission of mouse APL (Fig. 6c).

Ex vivo, retinoic acid triggers transient growth arrest of most LICs and terminal granulocytic differentiation of colonies that develop (Fig. 1a, b). Such heterogeneity in LIC response may reflect hierarchical differences in the originally transduced progenitors. A study of human stem cells transduced by PML-RARA found that retinoic acid actually increases LIC abundance, without any differentiation35. However, APL is not a stem cell disease40, and PML-RARA overexpression in stem cells may not accurately recapitulate oncogene-triggered changes of fate for other hematopoietic progenitors. Elucidation of the actual cellular basis for LIC clearance, whether it is loss of self-renewal or apoptosis, awaits purification of these cells. However, the fact that agents degrading PML-RARA (retinoic acid, arsenic and cAMP) all cooperate for LIC loss and APL eradication23,24,27,28,31 strongly suggests that even minute amounts of PML-RARA allow LIC self-renewal in vivo. How could PML-RARA degradation control LIC fate? Given the overlap between retinoic acid and arsenic targets in APL cells only39,41, a subset of target genes may be regulated through PML-RARA degradation. Alternatively, PML-RARA may behave as a nuclear anchor that sequesters master proteins, as shown for Daxx and PML18,42. Daxx sequestration is dependent on PML Lys160 SUMOylation18, and, notably, the latter is essential for PML-RARA–mediated transformation5. Although, as in Bcr-Abl–transformed cells, degradation of PML by arsenic may contribute to abrogation of LIC self-renewal43, arsenic is unlikely to act solely through the normal PML proteins, as it is ineffective in PLZF-RARA APLs24 (data not shown).

In several settings (1.5 mg of retinoic acid, bortezomib and APLs expressing PML-RARAL902P or PML-RARAS873A), retinoic acid–dependent transcriptional activation and differentiation are maintained, whereas PML-RARA degradation and LIC clearance are impeded1,31 (Figs. 2, 3, 4, 5). These observations could suggest that LIC proliferation is dependent on the absolute level of PML-RARA, whereas differentiation of the bulk of APL cells primarily relies on transcriptional activation of PML-RARA target genes by retinoic acid. The molecular basis for the different behavior of these two cell populations remains to be understood. Accordingly, higher retinoic acid concentrations are required for complete PML-RARA catabolism and LIC clearance than for in vivo differentiation and gene activation1,15,31 (Figs. 1 and 6). In this respect, liposomal retinoic acid, which maintains higher and more prolonged plasma levels than do 10-mg retinoic acid pellets (10−6 M versus 10−7 M, respectively), cures most APL mice and even some humans with APL as a single agent25,44. The suboptimal retinoic acid and arsenic doses used in our study may therefore have facilitated the synergies between retinoic acid and arsenic and between retinoic acid and cAMP. Thus, RARA agonists resistant to Cyp26a1 catabolism should be better drugs than retinoic acid and, conversely, a low-dose retinoic acid induction regimen, while initially retaining the ability to fully differentiate APL cells45, would be expected to yield poor long-term outcome. That retinoic acid fully differentiates PLZF-RARA APLs, but fails to clear LICs, was unexpected but consistent with some clinical observations3 (Supplementary Fig. 3b). Our experiments therefore do not support the proposal that retinoic acid resistance reflects impaired PLZF-RARA target gene activation and differentiation, both dependent on a repression domain in PLZF3. Because retinoic acid degrades PLZF-RARA24,46, retinoic acid resistance may reflect either a high basal level of PLZF-RARA protein expression or the influence of the reciprocal RARA-PLZF fusion, which may control progenitor fate through PLZF target genes36,47.

Retinoic acid resistance of PML-RARAS873A APL implies that basal, cAMP-induced or retinoic acid–induced phosphorylation of Ser873 is crucial for the response of APL to retinoic acid (Fig. 5), although PML-RARA phosphorylation does not account for all of the effects of cAMP on APL cells. For example, cAMP-induced release of SMRT from RARA, which is independent of Ser369 (ref. 32), could contribute to PDEI-induced degradation of PML-RARAS873A, as well as to enhanced differentiation. In any case, we establish the in vivo relevance of retinoic acid–cAMP cross-talk1,30,31,32,38 and show that it targets APL LICs, at least in part, through PML-RARA phosphorylation and accelerated catabolism (as defective AF1 phosphorylation33 could also contribute to retinoic acid resistance). We thus identify the first biological system, to our knowledge, in which phosphorylation controls nuclear receptor signaling in vivo and identify cAMP signaling as another candidate for targeted APL therapy. Given that the RARs have been implicated in stem cell maintenance48,49 and that the retinoic acid–cAMP association blocks clonogenic growth of acute myeloid leukemias that do not express PML-RARA32, cAMP could enhance retinoic acid sensitivity in other malignancies, such as acute myeloid leukemia or neuroblastoma50.

Our results provide insight into the very high rates of complete clinical remission induced by frontline combined retinoic acid–arsenic treatments in individuals with APL27,28 and suggest the further investigation of PDEIs, which were indeed beneficial for an index case31. At a fundamental level, our studies show that differentiation is not the primary basis for the therapeutic efficacy of retinoic acid or arsenic in clearing APL. In cancers where acquisition of stem cell features is conferred by oncogene expression, oncoprotein degradation could be a new general therapeutic strategy.

Methods

Cell culture and viruses.

We obtained retroviruses by transient transfection of Plat-E cells with various pMSCV vectors. We infected lineage-depleted mouse hematopoietic cells collected from 5-fluorouracil–treated mice with these retroviruses5. After spinoculation, we cultured transduced cells in methylcellulose medium (Stem Cell Technologies) supplemented with 100 ng ml−1 stem cell factor and 10 ng ml−1 each of interleukin (IL)-3, IL-6 and granulocyte-macrophage colony-stimulating factor (AbCys), with G418 for selection. After 1 week, we recovered neomycin-selected cells from methylcellulose and either analyzed them or replated them at 10,000 cells per well in the presence or absence of 1 μM retinoic acid (Sigma-Aldrich).

Immortalized mouse embryonic fibroblasts from which the three RARs were excised6,32 were transduced with RARA derivatives. We cultured cells in DMEM containing 10% FBS and treated them overnight with retinoic acid before gene expression analysis. We collected APL or bone marrow cells from femurs and tibiae of leukemic mice by flushing marrow with RPMI through a 26-gauge needle. We then cultured APL cells in complete RPMI medium supplemented with IL-3, IL-6 and stem cell factor, as well as retinoic acid, arsenic, MG132 and/or the phosphodiesterase inhibitor piclamilast8.

In vivo mouse studies.

PML-RARA and PML-RARAL902P APLs were previously described5,23,31. A similar transplantation model was established in nude mice, using APLs from PLZF-RARA–RARA-PLZF mice36. For secondary transplantations, we injected 3 × 106 bone marrow cells intravenously into three syngeneic recipient mice per primary recipient. We used two mice for each treatment modality and time point (days 3 and 6). We repeated mouse experiments two to six times. We administered retinoic acid by subcutaneous implantation of release pellets containing 10 or 1.5 mg of retinoic acid (Innovative Research of America). We administered cAMP by subcutaneous implantation of Alzet pumps31. We administered arsenic and piclamilast (200 mg d−1) by daily intraperitoneal injections18. We administered bortezomib (10 μg d−1) and H89 (60 mg d−1) using Alzet pumps. We measured plasma retinoic acid concentrations as previously described31. In vivo imaging was done in a Xenogen IVIS100 facility. Animal handling was done according to the guidelines of institutional animal care committees, using protocols approved by the Comité Régional d'Ethique Expérimentation Animale n°4. We paraffin-embedded and stained tissues as previously described31. We flushed, cytospinned and colored bone marrow cells with May Grunwald Giemsa (MGG) as previously described31.

PML-RARAS873A transgenic mice were generated on an FVB background using the human MRP8 expression cassette5,8. Seven founders were obtained that had Mendelian transgene transmission and presented with PML-RARA expression similar to that of our previous PML-RARA transgenic mice5,8. Several lines developed mild splenomegaly, and one line developed APLs in the F0 founder and one F1. Resequencing of the transgene in these APLs did not reveal any additional genetic changes in PML-RARAS873A.

Real-time PCR.

We isolated total RNA from the bone marrow using an RNeasy kit (Qiagen). We synthesized first-strand cDNA with SuperScript II reverse transcriptase (Invitrogen). We did quantitative real-time PCR with a LightCycler system (Roche). We purchased probes and primers for TaqMan assays for Rarb, Cyp26a1 and Tgm2 from Applied Biosystems. We used the Ywhaz gene as an endogenous control to calibrate the amount of mRNA target in different samples. We quantified PML-RARA and ZBTB16-RARA DNAs by real-time PCR using a 7500 Fast Real-Time PCR system (Applied Biosystems) with either TaqMan (for PML-RARA) or SYBR Green (for ZBTB16-RARA). The gene encoding 18S RNA was used as an internal control. The primers used were: PML, 5′-GCTGCCGTGCGCACCGAT-3′ and 5′-GGGGCTAGGCGGTCCATC-3′; ZBTB16, 5′-TACAGGAAGCTGCACAGTGG-3′ and 5′-TTGTAGATGCGGGGTAGAGG-3′.

Protein analyses.

We resolved cell lysates by 7% SDS-PAGE and transferred them onto nitrocellulose membranes. We carried out detection with the chemiluminescent substrate SuperSignal West Pico (Pierce Biotechnology), using rabbit polyclonal antibodies to RARA (RP115) and with actin serving as a loading control. For the detection of phosphorylated PML-RARA, we first applied cell extracts to PhosphoProtein affinity purification columns (Qiagen). After being washed, column eluates containing protein peaks were concentrated, resolved by SDS-PAGE and analyzed by immunoblotting using antibodies to RARA phosphorylated at the PKA site33, or RP115 as a control for the total amount of PML-RARA (Fig. 5h).

For flow cytometry analysis, we first blocked cellular Fc receptors with normal rat IgG. We then carried out immunophenotypic analysis using fluorochrome-conjugated monoclonal antibodies to CD11b and Gr-1 (clone RB6-8-C5, eBioscience). Staining was done at 4 °C for 20 min. We washed the cells twice and resuspended them in HBSS with 2% FBS and 0.5 μg ml−1 propidium iodide. We gated dead cells out by high propidium iodide staining and forward light scatter.

Statistics.

Two-sided t tests were used to validate the significance of the observed differences, which were considered statistically significant when P < 0.05.