Anti-telomerase therapy provokes ALT and mitochondrial adaptive mechanisms in cancer

SUMMARY

To assess telomerase as a cancer therapeutic target and determine adaptive mechanisms to telomerase inhibition, we modeled telomerase reactivation and subsequent extinction in T-cell lymphomas arising in Atm^-/- mice engineered with an inducible telomerase reverse transcriptase allele. Telomerase reactivation in the setting of telomere dysfunction enabled full malignant progression with alleviation of telomere dysfunction-induced checkpoints. These cancers possessed copy number alterations targeting key loci in human T-cell lymphomagenesis. Upon telomerase extinction, tumor growth eventually slowed with re-instatement of telomere dysfunction-induced checkpoints, yet growth subsequently resumed as tumors acquired Alternative Lengthening of Telomeres (ALT) and aberrant transcriptional networks centering on mitochondrial biology and oxidative defense. ALT+ tumors acquired amplification/overexpression of PGC-1β, a master regulator of mitochondrial biogenesis and function, and they showed marked sensitivity to PGC-1β or SOD2 knock-down. Genetic modeling of telomerase extinction reveals vulnerabilities that motivate coincidental inhibition of mitochondrial maintenance and oxidative defense mechanisms to enhance anti-telomerase cancer therapy.

INTRODUCTION

Telomeres are nucleoprotein complexes at chromosome ends that function to maintain chromosomal integrity. Genetic models have defined their critical roles in cancer (Artandi and DePinho, 2010), aging and degenerative diseases (Sahin and DePinho, 2010). Telomeres are synthesized by telomerase consisting of a reverse transcriptase catalytic subunit (TERT) and an RNA template subunit (TERC) (Feng et al., 1995; Nakamura et al., 1997). As normal or premalignant cells divide, the end-replication problem of conventional DNA polymerases, coupled with low or absent telomerase activity, results in loss of telomere sequences and eventual telomere uncapping, which activates cellular checkpoints similar to those provoked by DNA double-stranded-breaks (DSBs) (Harley and Sherwood, 1997). Like classical DSBs, telomere dysfunction has been shown to induce p53 and associated cellular responses, such as senescence and/or apoptosis (Chin et al., 1999; d’Adda di Fagagna et al., 2003; Karlseder et al., 1999; Takai et al., 2003; van Steensel et al., 1998). Upon mutational inactivation of p53, continued cell cycling and survival of cells with telomere dysfunction provide a pro-carcinogenic mutator mechanism characterized by translocations and regional amplifications and deletions (Artandi et al., 2000; Chin et al., 1999; O’Hagan et al., 2002). At the same time, persistent telomere dysfunction and associated rampant chromosomal instability (even in p53 null cells) compromises cellular viability (Begus-Nahrmann et al., 2009; Chin et al., 1999) and constrains full malignant progression of such cancers (Artandi et al., 2000; Chang et al., 2003; Gonzalez-Suarez et al., 2000; Rudolph et al., 2001).

The occurrence of telomere erosion and importance of telomerase-mediated telomere maintenance in fully established cancers are evidenced by typically short telomeres relative to normal tissues and robust telomerase activity in most human cancers (Shay and Wright, 2006). This profile of shorter telomeres and telomerase activity in cancer has motivated the clinical development of telomerase inhibitors including a 13-mer antisense oligonucleotide in a number of cancer types (Agrawal et al., 2012; Shay and Wright, 2006). At present, uncertainty surrounds whether anti-telomerase therapy will be hampered by the potential lag time needed for telomere erosion-associated tumor cell killing, and/or whether re-entry into telomere-based crisis will engender genomic instability that may allow for emergence of adaptive responses and resistance mechanisms, such as Alternative Lengthening of Telomeres (ALT) mechanism which enables telomere maintenance via homologous recombination (Cesare and Reddel, 2010). Notably, in mTERC^-/- mice, p53 deficiency alleviates tumor suppression imparted by telomere dysfunction (Chin et al., 1999) and, in transformed cells with intact p53-dependent DNA damage checkpoint, there is activation of ALT (Chang et al., 2003).

The study of cancer pathogenesis and specific role of telomeres therein have been enabled by the use of genetically engineered mouse (GEM) models of cancer, which provide an in vivo assessment of the complex adaptive responses to targeted cancer treatments. Here, we exploited the experimental merits of mice to model and study more precisely telomere crisis, telomerase reactivation and telomerase extinction in cancer development, progression and treatment in the in vivo setting. To that end, we engineered an inducible TERT knock-in allele and studied telomere dynamics in mice mutant for Atm which develop T-cell lymphomas with high penetrance (Xu et al., 1996). The Atm mutant model was selected to study telomerase activation and its extinction as this model retains a robust p53-mediated checkpoint response upon telomere uncapping which strongly suppresses emergent tumors (Qi et al., 2003; Wong et al., 2003). Our study provides in vivo genetic evidence that telomerase reactivation facilitates the progression of spontaneous arising tumors experiencing telomere dysfunciton and, conversely, that telomerase extinction in established cancers activates ALT and novel adaptive mechanisms illuminating potential therapeutic combinations.

RESULTS

Telomerase reactivation promotes tumorigenesis by stabilizing telomeres and alleviating telomere dysfunction-induced checkpoints

We generated mice harboring an Atm null allele (Xu et al., 1996) and a 4-Hydroxytamoxifen (4-OHT)-inducible Telomerase Reverse Transcriptase-Estrogen Receptor (TERT-ER^T2) fusion knock-in allele (Jaskelioff et al., 2011). In the absence of 4-OHT, mice homozygous for TERT^ER (designated TERT^ER/ER) are telomerase activity deficient and sustain same cytogenetic and cellular phenotypes of conventional TERT or TERC knockout model (Jaskelioff et al., 2011). Upon 4-OHT treatment, TERT-ER protein activity can be restored to levels comparable to the native TERT protein (Jaskelioff et al., 2011). To generate lymphoma-prone mice with telomere dysfunction, Atm^+/- TERT^+/ER mice were intercrossed to produce first generation (G₁) Atm^-/- TERT^ER/ER and Atm^+/-TERT^ER/ER mice. G₁ Atm^+/-TERT^ER/ER mice were then serially intercrossed to successive generations in the absence of 4-OHT, culminating in G₃ and G₄ Atm^-/- TERT^ER/ER experimental cohorts (Figure S1A). For brevity, the Atm^-/- TERT^{+/+ or +/ER}, G₁ Atm^-/- TERT^ER/ER, G_3-4 Atm^-/-TERT^ER/ER, mice are designated G₀, G₁, and G_3-4, respectively.

G₀ and G₁ mice developed T-cell lymphomas at similar latencies and penetrance (Figure S1B), which is consistent with adequate telomere reserves to maintain capping in G₁ mice and avoid activated DNA damage checkpoints. In contrast, G_3-4 lymphomas emerge with a longer latency and lower penetrance relative to G₁ controls (Figure 1A), consistent with previous work (Maser et al., 2007; Qi et al., 2003; Wong et al., 2003). Next, to test the impact of telomerase reactivation on lymphoma kinetics in this model, continuous-release 4-OHT tablets were implanted subcutaneously at 18 weeks, an age when lymphomas in Atm^-/- mice with intact telomeres start to emerge. 4-OHT-treated G_3-4 mice developed overt lymphomas more rapidly and with higher penetrance than, not only vehicle-treated G_3-4 mice, but also G₁ mice (Figure 1A). In contrast, there were no differences in tumor latency and penetrance in the 4-OHT- and vehicle-treated G₁ mice (Figure 1A). On the histopathological level, vehicle-treated G_3-4 tumors were smaller in size and displayed less aggressive malignant features relative to 4-OHT-treated G_3-4 tumors as well as telomere intact G₀ and G₁ tumors (Figure 1B-1D; data not shown). Finally, 4-OHT treatment itself does not impact on lymphomagenesis as reflected by similar survival curves of G_3-4 Atm^-/- TERT^ER/ER mTR^-/- mice treated with either vehicle or 4-OHT (Figure 1A), supporting the view that acceleration of 4-OHT on G_3-4 Atm^-/- TERT^ER/ER tumors stems from telomerase reactivation and not from non-telomere activities of 4-OHT.

Figure 1. Telomerase reactivation in late generation TERT^ER/ER Atm^-/- mice promotes T-cell lymphomagenesis (related to Figures S1).

(A) Kaplan-Meier curves (Log-rank test) of T-cell thymic lymphoma-free survival for G₁ and G_3-4 TERT^ER/ER Atm^-/- (mTR^-/-) mice treated with 4-OHT or vehicle.

(B) Representative T-cell thymic lymphoma of G_3-4 TERT^ER/ER Atm^-/- mice treated with 4-OHT or vehicle. Scale bars represent 1 cm.

(C) Tumor sizes of G₁ and G_3-4 TERT^ER/ER Atm^-/- lymphomas treated with 4-OHT or vehicle. (t-test, and error bars indicate s.d.)

(D) Representative images of H&E-stained G_3-4 TERT^ER/ER Atm^-/- lymphomas treated with 4-OHT or vehicle. Scale bars represent 100 μm.

The capacity to regulate endogenous telomerase activity in vivo prompted serial analyses of malignant properties and cellular checkpoint responses as functional readouts of telomere status during lymphomagenesis and progression in G_3-4 mice. Quantitation of DNA damage foci by anti-γH2AX and -53BP1 staining showed markedly increased foci in vehicle-treated G_3-4 tumor cells and a dramatic decrease in such foci in 4-OHT-treated G_3-4 tumor cells (Figure 2A, 2B and S2). Correspondingly, vehicle-treated G_3-4 tumor cells exhibited decreased proliferation (Ki67) and increased apoptosis (Cleaved Caspase-3), senescence (SA-β-Gal) and p53 signaling (p53 phosphorylation and p21 expression), which are all alleviated by telomerase reactivation (Figure 2C-2F).

Figure 2. Telomerase reactivation alleviates telomere dysfunction and checkpoints in late generation TERT^ER/ER Atm^-/- lymphomas (related to Figures S2).

(A) Representative immunohistochemistry staining of γ-H2AX and 53BP1 in G_3-4 TERT^ER/ER Atm^-/- thymus and lymphomas treated with 4-OHT or vehicle.

(B) Quantification of (A) (n=5, t-test).

(C) Representative immunohistochemistry staining of cleaved Caspase3 and Ki67 in G_3-4 TERT^ER/ER Atm^-/- thymus and lymphomas treated with 4-OHT or vehicle.

(D) Quantification of (C) (n=5, t-test).

(E) Representative beta-gal staining of G_3-4 TERT^ER/ER Atm^-/- thymus and lymphomas treated with 4-OHT or vehicle (n=3).

(F) Immunoblotting of phosphorylated p53 (S15) and p21 in G_3-4 TERT^ER/ER Atm^-/- thymus and lymphomas treated with 4-OHT or vehicle.

Error bars indicate s.d and Scale bars represent 100 μm.

Telomerase reactivation in spontaneous tumors with unstable genomes promotes aggressive malignant properties

A higher percentage of G_3-4-4-OHT mice exhibited widespread tumor cell infiltration in spleen, kidney, liver, lung and bone marrow relative to vehicle-treated G_3-4 mice as well as 4-OHT- and vehicle-treated G₁ mice (Figure 3A and 3B). The more aggressive nature of 4-OHT-treated G_3-4 lymphomas was particularly evident from infiltration into brain (4/16) which was not observed in G₁ necropsies (n=7) (Figure 3A and 3B), demonstrating that the malignant progression of initiated tumors is enabled by somatic reactivation of telomerase. This acquisition of more aggressive tendencies may stem in part from increased genome instability during the formative stages of tumor development prior to telomerase activation. Indeed, telomere dysfunction-induced bridge-fusion-breakage events can generate non-reciprocal translocations as well as regional amplifications and deletions at sites of breakage which, under biological selection, can result in cancer-promoting copy number changes (Artandi et al., 2000; Maser et al., 2007; O’Hagan et al., 2002). To examine this possibility, we compared cytogenetic and genomic profiles of 4-OHT-treated G_3-4 tumors and 4-OHT-treated G₁ tumors, the latter without a period of telomere-driven crisis. Consistent with previous work (Maser et al., 2007), spectral karyotype (SKY) analysis showed a 3-fold increase in chromosomal rearrangements including non-reciprocal translocations in 4-OHT-treated G_3-4 tumors (n=3) relative to 4-OHT-treated G₁ tumors (n=3) (Figure 3C and 3D, 25.8 versus 8.8 total NRTs per metaphase, respectively; P<0.0001). The baseline of 8.8 NRTs per metaphase is consistent with previous work that ATM deficient tumors experience increased DNA double-stranded breaks, which promote chromosomal rearrangements as well (Xu et al., 1996).

Figure 3. Telomerase reactivation promotes invasiveness of late generation TERT^ER/ER Atm^-/- lymphomas (related to Figure S3 and Table S1).

(A) Representative images of infiltrated organs (lung, kidney, liver and brain) in G_3-4 TERT^ER/ER Atm^-/- mice treated with 4-OHT. Scale bars represent 100 μm.

(B) Quantification of infiltrating incidence of G₁ and G_3-4 TERT^ER/ER Atm^-/- mice treated with 4-OHT or vehicle.

(C) Representative spectral karyotype (SKY) images from metaphases of G₁ and G_3-4 TERT^ER/ER Atm^-/- lymphomas treated with 4-OHT.

(D) Quantification of non-reciprocal translocations (NRTs) detected by SKY in G₁ (n=3) and G_3-4 (n=3) TERT^ER/ER Atm^-/- lymphomas treated with 4-OHT. (P<0.0001, t-test)

(E) Recurrence plot of CNAs defined by array-CGH for 18 G_3-4-4-OHT lymphomas. The x-axis shows the physical location of each chromosome. The percentage of tumors harboring gains (dark red, log₂ ≥ 0.3), amplifications (bright red, log₂ ≥ 0.6), losses (green, log₂ ≤ -0.3), and deletions (dark green, log₂ ≤ -0.6) for each locus is depicted. Locations of physiologically relevant CNAs of TCR loci and some known cancer genes are indicated with asterisks.

Examination of copy number alterations (CNAs) in 18 G_3-4-4-OHT, 10 G₁-4-OHT and 5 G_3-4-VEH tumors (Figure 3E, S3A and S3B) revealed novel copy number gains on chromosomes 12, 17 and 18 and copy number losses on chromosome 14 in G_3-4-4-OHT tumors but not G₁-4-OHT tumors, consistent with increased genomic instability in late generation tumors (Figure 3E and S3A). Notably, G_3-4-4-OHT and G_3-4-VEH tumors show comparable levels of clonal CNAs, consistent with the view that wholesale genomic events driving lymphomagenesis occurs early in cancer development and that telomerase reactivation does not discernibly alter the scope of clonal genomic events in these cancers (Figure 3E and S3B). Along these lines, we assessed whether the observed CNAs in G_3-4-4-OHT mice might contribute to the lymphomagenesis by examining whether murine and human lymphomas sustain orthologous CNA events than might be expected by chance as conducted previously (Maser et al., 2007). Array-CGH profiles of G_3-4-4-OHT tumors identified 4928 genes resident within regions of genomic gain and 2297 genes resident within regions of genomic loss with the use of Segment Gain or Loss (SGOL) algorithm (Wiedemeyer et al., 2010) (Extended Experimental Procedures). Comparison of G_3-4-4-OHT tumors to human T-ALLs (Maser et al., 2007) showed that 565 of 4928 amplified genes (11.5%) and 300 of 2297 deleted genes (13%) are targeted for copy number alteration in both species (Table S1), which are significantly higher than those expected by chance (n=10000; P=3e-04 and 2e-04 for amplification and deletion, respectively) (Figure S3C). These cross-species shared copy number altered genes include several known tumor suppressors and oncogenes implicated in T-ALL biology such as Crebbp, Ikaros, Abl, Notch1, Myc and PTEN (Figure 3E). These data and previous work (Maser et al., 2007; O’Hagan et al., 2002) suggest that telomere dysfunction provides a mechanism that promotes structural genome alterations of cancer-relevant loci, which not only drive primary tumorigenesis but also confer additional malignant properties such as enhanced invasiveness.

Genetic extinction of telomerase inhibits cancer growth and leads to eventual tumor re-emergence

The TERT-ER system affords exploration of the impact of genetic extinction of telomerase activity in established cancers that, similar to human cancers, had evolved to first experience telomere dysfunction, and then subsequently acquire telomerase activity. In particular, this genetic model system enables assessment of tumor biological impact and potential adaptive mechanisms of anti-telomerase therapy following reacquisition of telomere dysfunction. Freshly harvested 4-OHT-treated G_3-4 tumor cells mice were passaged directly through SCID mice pre-implanted with vehicle or 4-OHT tablets (Figure S4A). Of 15 primary tumors lines (10 mice per each passage of treatment; 10⁷ cells per intraperitoneal injection), 11 lines generated xenograft tumors. In first and second passages (P₁ and P₂), these lines showed no differences in the two treatment arms in terms of penetrance and latency (Figure 4A; P₁ and P₂). However, at third passage (P₃), 9 of 11 lines showed lower penetrance and longer latency in the vehicle-treated arm relative to the 4-OHT-treated arm, -- a profile consistent with the view that multiple rounds of cell division are needed to achieve sufficient telomere erosion and re-entry into crisis. Six of these 9 lines showed complete loss of tumor formation in the vehicle-treated arm, whereas 3 lines yielded escaping tumors in several vehicle-treated recipient mice (Figure 4A; P₃). When these escaping P₃ tumors were passaged to P₄, they reacquired aggressive malignant properties approaching those of matched 4-OHT controls (Figure 4A; P₄), suggesting acquisition of resistant mechanisms to counteract telomerase deficiency. Moreover, 2 of the original 11 tumor lines did not show any attenuation of tumor growth in the vehicle-treated arm compared with the 4-OHT-treated arm (data not shown), suggesting early acquisition of such resistant mechanisms (see below).

Figure 4. Telomerase depletion in late generation 4-OHT-TERT^ER/ER Atm^-/- lymphomas leads to cell death and resistance (related to Figure S4).

(A) Kaplan-Meier curves (Log-rank test) of lymphoma-free survival of 4OHT- or vehicle-treated mice serially transplanted with TERT^ER/ER Atm^-/- lymphomas.

(B) Quantification of cleaved Caspase3 positive cells in serially transplanted lymphomas treated with 4OHT or vehicle (n=3-5, t-test).

(C) Immunoblotting of phosphorylated p53 (S15) and p21 in serially transplanted lymphomas treated with 4OHT or vehicle.

(D) Quantification of 53BP1 foci positive cells in serially transplanted lymphomas treated with 4OHT or vehicle (n=3-5, t-test).

(E) Quantification of γ-H2AX foci positive cells in serially transplanted lymphomas treated with 4OHT or vehicle (n=3-5, t-test).

Error bars indicate s.d. (*P<0.01relative to any other individual samples in the same group).

To assess how G_3-4-4-OHT tumors respond to telomerase extinction on molecular level, we audited p53 signaling, DNA damage foci and apoptosis in serial passages of the two treatment arms. In 4-OHT-treated P₁ to P₄ and vehicle-treated P₁ tumor cells, p53 signaling, DNA damage foci and apoptosis levels were similar to those of parental tumor cells (P₀) (Figure 4B-4E). In contrast, vehicle-treated P₂ tumor cells showed significantly higher levels of p53 signaling, DNA damage foci and apoptosis, and all of these processes were reduced in the resistant vehicle-treated P_3-4 tumor cells (Figure 4B-4E), raising the possibility that these resistant tumors may have stabilized their telomeres to alleviate these checkpoints. Intriguingly, while DNA damage foci were reduced in the resistant tumors compared with the vehicle-treated P₂ tumors, these foci remain higher than telomerase+ (4-OHT-treated) tumors (Figure 4D and 4E), consistent with residual genotoxic stress in the resistant tumors. However, despite ongoing DNA damage signaling, apoptosis was not observed, which is consistent with compromised p53 signaling in the resistant tumors (Figure S4B).

Emergence of ALT in tumors following extinction of telomerase

To elucidate adaptive mechanisms following telomerase extinction, we catalogued telomere lengths in serially passaged tumors in the two treatment arms. In passaged 4-OHT-treated P₁ to P₄ tumor cells, telomere lengths remained unchanged from those in parental tumor cells, while vehicle-treated P₁ and P₂ tumor cells showed progressive telomere shortening (Figure 5A). However, vehicle-treated P_3-4 tumor cells showed a sharp increase in average and maximal telomere lengths that exceeded those in matched 4-OHT controls (Figure 5A); however, these vehicle-treated P_3-4 tumor cells tumor cells showed more telomere signal-free ends by telomere-FISH (Figure 5B and 5C), -- a picture consistent with higher heterogeneity in the distribution of telomere lengths in resistant tumors. In addition, co-staining of promyelocytic leukemia (PML) bodies and telomeres showed an increase in ALT-associated PML bodies (APBs) (Figure 5D and 5E), and telomere-FISH showed more extra-chromosomal telomere fragments in the vehicle-treated P_3-4 tumors cells relative to the matched 4-OHT controls (Figure 5F and 5G). Finally, vehicle-treated P_3-4 tumor cells showed an increased telomere Sister Chromatid Exchange (tSCE) rate than matched 4-OHT controls (Figure 5H and 5I). These collective data provide clear evidence of ALT in vehicle-treated P_3-4 tumor cells which have resumed robust tumor growth in vivo. In addition, the two tumor lines, which did not show any attenuation of tumor growth in the vehicle-treated arm, also showed ALT features including APBs and extra-chromosomal telomere fragments (Figure S5A and S5B), consistent with the early acquisition of ALT.

Figure 5. TERT^ER/ER Atm^-/- lymphomas developed ALT in the absence of telomerase (related to Figure S5).

(A) Telomere lengths of serially transplanted lymphomas treated with 4OHT or vehicle measured by telomere-FISH coupled laser scanning cytometry (n=3-5).

(B) Telomere signal-free ends were measured by telomere-FISH in serially transplanted lymphomas treated with 4OHT or vehicle.

(C) Quantification of telomere signal-free ends in serially transplanted lymphomas treated with 4OHT or vehicle (n=3-5, t-test).

(D) Representative images of ALT-associated PML Body (APB) co-stained with PML and telomere.

(E) Quantification of ALT-associated PML Body (APB) numbers in resistant lymphomas and matched 4-OHT controls (n=5, t-test).

(F) Representative images of extra-chromosomal telomere fragments in resistant tumors. Circled area indicates metaphase spread not separated well.

(G) Quantification of extra-chromosomal telomere fragments in resistant tumors and matched 4-OHT controls (n=5, t-test).

(H) Representative images of telomere Sister Chromatid Exchange (tSCE) in resistant tumors.

(I) Quantification of telomere Sister Chromatid Exchange (tSCE) in resistant tumors and matched 4-OHT controls (n=5, t-test).

Error bars indicate s.d.

ALT tumors show up-regulation of a master regulator of mitochondrial biogenesis /function and oxidative defense

Serial loss of telomere function and genome destabilization in this genetically defined system provided an unprecedented opportunity to discern on a genome scale the associated adaptive molecular events in these ALT+ tumor cells via integrated transcriptomic and aCGH profile analyses. Cluster analysis of transcriptomes of 4-OHT-treated P₄ tumors (n=3), and parental 4-OHT-treated tumor revealed no significant differences (Figure 6A), consistent with genomic stability brought about by continuous telomerase activity. In contrast, vehicle-treated ALT+ P₄ tumor (n=5) expression profiles showed significant differences relative to the 4-OHT-treated telomerase-positive P₄ tumors – 891 up-regulated and 1345 down-regulated genes in ALT+ tumors with 1.5 fold change and P<0.01 (t-test; Figure 6A; Table S2). Pathway (IPA) analysis of the ALT-specific transcriptome showed strong representation of networks centering on mitochondrial biology and oxidative stress regulation (Figure S6A). Consistent with the known role of DNA recombination in ALT (Fan et al., 2009; Zhong et al., 2007), MRE11 and FANCA were also up-regulated in these ALT+ tumors (Figure S6B and S6C). With regard to mitochondria and oxidative pathways, Q-PCR verified aberrant expression of several key genes in these networks including up-regulation of a master regulator of mitochondrial biology and oxidative defense, PGC-1β, and its targets such as NRF2, SOD2 and Catalase, among others (Figure 6B).

Figure 6. ALT+ tumors overcome their higher levels of ROS and mitochondrial dysfunction by up-regulating PGC-1β and other oxidative defense genes (related to Figure S6, Table S2, and Table S3).

(A) Microarray heat map of differentially expressed genes in ALT+ tumors relative to telomerase+ lymphomas.

(B) RT-qPCR validation of oxidative defense genes up-regulated in ALT+ lymphomas (n=3, t-test).

(C) CNVs of chromosome 18 in 8 independent telomerase+ lymphomas and 5 independent ALT+ lymphomas.

(D) Mitochondria DNA contents of serially transplanted lymphomas and ALT+ lymphomas were measured by qPCR (n=3-5, t-test).

(E) RT-qPCR validation of mitochondrial biogenesis genes in serially transplanted lymphomas and ALT+ lymphomas (n=3, t-test).

(F) Oxygen Consumption Rates (OCRs) relative to mtDNA contents were measured in serially transplanted lymphomas and ALT+ lymphomas (n=3, t-test).

(G) ROS levels of serially transplanted lymphomas and ALT+ lymphomas were measured by DCFDA staining (n=3-5, t-test).

Error bars indicate s.d. (*P<0.001, **P<0.01 relative to any other individual samples in the same group).

As noted, telomere dysfunction-induced DNA double-strand breakage process and biological selection can lead to functionally relevant CNAs, prompting us to identify potential genetic events that may inform how ALT+ cells cope with telomere-based crisis. To that end, we performed array-based comparative genomic hybridization (aCGH) analysis of ALT+ versus telomerase+ P₄ tumors from the two treatment arms. While serially transplanted telomerase+ tumors showed minimal CNA differences relative to the parental tumor, ALT+ tumors acquired numerous CNAs (see below). Strikingly, integrated transcriptomic and copy number analysis revealed PGC-1β as the only gene in the mitochondrial function and oxidative stress regulation pathways showing both up-regulated expression and copy number gain in the ALT+ tumors relative to telomerase+ tumors (Table S3); these genomic data raised the possibility that PGC-1β might be a major driver of the adaptive response to telomere dysfunction.

To ask whether PGC-1β amplification/over-expression is a common event in our ALT+ tumors, we profiled 8 independent telomerase+ tumors and 5 independent ALT+ tumors, and found that 3 of 5 ALT+ tumors (60%) show amplification of a large region of chromosome 18, which includes PGC-1β gene, whereas none of the telomerase+ tumors (n=8) sustain such genomic alterations (Figure 6C). To reinforce the potential role of PGC-1β up-regulation as a possible adaptive response to ALT, we examined PGC-1β expression levels in ALT+ tumors without PGC-1β amplification, and established that PGC-1β expression is significantly higher in ALT+ tumors than their corresponding telomerase+ tumors (Figure S6D).

ALT+ tumors have higher levels of mitochondrial dysfunction and ROS relative to telomerase+ tumors

PGC-1α and PGC-1β are master regulators of mitochondrial biogenesis and function, play integral roles in the regulation of genes governing ROS defense, and are down-regulated in normal tissues experiencing telomere dysfunction as a result of activated p53-mediated repression (Sahin et al., 2011). This framework, coupled with above genomic observations, prompted us to hypothesize that telomerase extinction and ensuing telomere dysfunction repress the mitochondrial biogenesis and function in cancer cells, and the adaptive response to such telomere-based crises targets the p53-PGC axis in an effort to restore the mitochondrial function. To assess this possibility, we assessed PGC network expression, mitochondrial mass and mitochondrial function in passaged 4-OHT-treated and vehicle-treated P₁-P₄ tumor cells. Mitochondria DNA content were maintained in the telomerase+ arm but sharply reduced in the vehicle-treated P₂ tumor cells yet significantly restored in the ALT+ P_3-4 tumor cells (Figure 6D). Correspondingly, in contrast to the telomerase+ arm or early passage of vehicle-treated tumor cells, vehicle-treated P₂ tumor cells showed reduced expression of PGC-1β and its major targets NRF-1, ERRα, PPARα and TFAM which are critical for mitochondrial biogenesis and function (Figure 6E; note: PGC-1α is not expressed in these murine T-cell lymphomas). These expression patterns were significantly normalized in the ALT+ P_3-4 tumor cells (Figure 6E).

With regard to mitochondrial function, mitochondrial respiration assays (normalized for mtDNA content) showed that vehicle-treated P₂ tumor cells have dramatically impaired mitochondrial respiration relative to telomerase+ cells (Figure 6F and S6A). ALT+ P_3-4 tumor cells show substantial improvement in mitochondrial respiration, although function remains somewhat compromised relative to those of telomerase+ cells (Figure 6F and S6E). Finally, the capacity of PGC-1β to regulate oxidative defense genes (St-Pierre et al., 2006) and the integral role of mitochondria in ROS control (Murphy, 2009) prompted examination of intracellular ROS levels in ALT+and telomerase+ tumor cells. Using the fluorescent dye DCF-DA protocol, FACS analysis showed significantly increased ROS levels in vehicle-treated P₂ tumor cells relative to 4-OHT-treated P₀ to P₄ and vehicle-treated P₁ tumor cells (Figure 6G and S6F). In ALT+ tumor cells, ROS levels are lower than vehicle-treated P₂ tumor cells, albeit higher than telomerase+ tumor cell controls (Figure 6G and S6F). Thus, telomere dysfunction in cancer cells is commonly associated with repression of PGC1β and its network resulting in mitochondrial mass decline and the adaptive responses to telomere dysfunction appear to involve not only activation of ALT telomere maintenance mechanism but also up-regulation of genetic pathways promoting partial restoration of mitochondrial mass and function. Importantly, however, relative to telomerase+ cells, ALT+ tumor cells still have modestly elevated mitochondrial dysfunction and ROS which may reflect ongoing genotoxic stress relating to inefficiencies of telomere maintenance via ALT as reflected by moderately increased DNA damage foci above baseline in these tumors (see Figure 4D and 4E above).

ALT+ tumors show increased sensitivity to inhibition of PGC-1β or SOD2

Because relatively functional mitochondria and moderate levels of ROS are required for cancer cell survival (Weinberg et al., 2010), we asked whether shRNA-mediated knockdown of PGC-1β would have a negative or differential impact on telomerase+ and ALT+ tumor cells. Multiple independent shRNAs achieved PGC-1β knockdown in both telomerase+ and ALT+ tumor cells (Figure 7A). Consistent with the pivotal role of PGC co-activators in mitochondrial function and ROS regulation, both telomerase+ and ALT+ tumor cells showed significantly reduced mitochondrial numbers, compromised mitochondrial function, and increased ROS after PGC-1β knockdown (Figure 7B-7D). Notably, ALT+ tumor cells showed greater compromise in all three categories relative to telomerase+ tumor cells (Figure 7B-7D), raising the possibility that ALT+ tumor cells may show greater sensitivity to PGC-1β inhibition with increased mitochondrial dysfunction and ROS toxicity. Correspondingly, shRNA-mediated knockdown of PGC-1β showed a minimal impact on telomerase+ tumor growth and a profound impact on ALT+ cells (median survival from 15 weeks to 19 weeks versus 17 weeks to over 30 weeks; P=0.0065, Cox model). These findings are consistent with increased apoptosis in ALT+ tumors with PGC-1β knockdown (Figure S7A). In addition, since SOD2 is a major ROS antioxidant in mitochondria, regulated by PGC-1β and highly expressed in ALT+ cells (Figure 6B), we tested the impact of SOD2 knock-down in ALT+ and telomerase+ cells. As expected, SOD2 knockdown increased ROS levels in both telomerase+ and ALT+ cells, although ROS levels in ALT+ cells (both basal or after SOD2 knockdown) were substantially higher than those in telomerase+ cells (Figure 7F and 7G). On the biological level, while non-targeting shRNA-transduced telomerase+ and ALT+ cells showed similar cell survival rates (MTT assay), SOD2 knockdown in ALT+ cell decreased survival greater than in telomerase+ cells (Figure 7H; P=0.02), suggesting that ALT+ cells are more sensitive to increase of ROS than telomerase+ cells. The sensitivity of ALT+ cells to ROS was further confirmed by NAC rescue experiments on ALT+ cells with PGC-1β and SOD2 knockdown (Figure S7B and S7C). Together, these data suggest that targeting PGC or ROS detoxifying genes may enhance the therapeutic impact of anti-telomerase therapy, particularly in the context of ALT+ resistant cells.

Figure 7. Inhibition of PGC-1β or SOD2 differentially kills ALT+ tumors over telomerase+ tumors (related to Figure S7).

(A) PGC-1β expression levels in telomerase+ and ALT+ tumor cells before and after shRNA knocking down (n=3, t-test).

(B) Mitochondria DNA contents in telomerase+ and ALT+ tumor cells before and after PGC-1β knocking down (n=3, t-test).

(C) Basal oxygen consumption rates (OCR) in telomerase+ and ALT+ tumor cells before and after PGC-1β knocking down (n=3, t-test).

(D) ROS levels in telomerase+ and ALT+ tumor cells before and after PGC-1β knockdown (n=3, t-test).

(E) Kaplan-Meier curves (Cox model test, P=0.0065) of lymphoma-free survival of mice transplanted with telomerase+ and ALT+ tumor cells with PGC-1β shRNA (shPGC-1β-1+ shPGC-1β-2) or control shRNA (shLuc).

(F) SOD2 expression levels in telomerase+ and ALT+ tumor cells before and after shRNA knocking down (n=3, t-test).

(G) ROS levels in telomerase+ and ALT+ tumor cells before and after shRNA knockdown (n=3, t-test).

(H) Relative cells numbers of telomerase+ and ALT+ tumor cells with SOD2 shRNAs or control shRNA (shLuc) (n=3, Two way ANOVA, P=0.02).

(I) Model of regulation of apoptosis, senescence and mitochondrial function in telomerase+, ALT+ and telomere-dysfunctional cells.

Error bars indicate s.d. (*P<0.0001,**P<0.001, ***P<0.02).

DISCUSSION

One of the hallmarks of cancer cells is the shift from mitochondrial to glycolytic metabolism as the prime source of energy and anabolic support (Vander Heiden et al., 2009; Warburg, 1956). While this bioenergetic shift conveys decreased reliance on mitochondria as a major energy source for cancer cells, our genetic studies indicate a continued vital role in the maintenance of mitochondrial function in cancer. This view aligns with multiple lines of evidence suggesting that mitochondrial competence is indeed important for cancer cell viability (Jose et al., 2011; Koppenol et al., 2011) and efficient oncogene mediated transformation (Gao et al., 2009; Weinberg et al., 2010; Wise et al., 2008; Yuneva et al., 2007). More specifically, our study uncovers a critical role of robust mitochondrial function in cancer cells, particularly for those cancer cells with ALT-maintained telomeres.

Why might ALT+ cancer cells exhibit exquisite sensitivity to inhibition of PGC expression or anti-oxidant defense? We speculate that continued DNA damage signaling at ALT-maintained telomeres perhaps due to altered chromatin (Heaphy et al., 2011) or inefficient capping via this mechanism could lead to mitochondrial damage. Mitochondria are the major intracellular source of ROS in mammalian cells, and dysfunctional mitochondria produce more ROS, which at excessively high levels is known to be detrimental to cell survival (Lin and Beal, 2006; Murphy, 2009; St-Pierre et al., 2006). In this context, we propose that the pressure to maintain mitochondrial function reflects the need to maintain tolerable intracellular ROS levels. This oxidative defense mechanism may be particularly important for telomeres as ROS are particularly injurious towards the G-rich sequences of telomeres (Hall et al., 1996; Oexle and Zwirner, 1997; Retel et al., 1993), thereby causing increased telomere dysfunction in the absence of efficient telomerase-mediated repair. Correspondingly, our combined genomic and functional studies establish that ALT+ cells strive to maintain adequate mitochondrial and oxidative defense functions and show exquisite sensitivity to loss of such function. On the molecular level, this work also establishes that the intimate PGC-directed link between telomeres and mitochondria identified recently in normal tissues (Sahin et al., 2011) is operative in cancer cells (Figure 7G).

About 15% of human cancers maintain their telomeres in the absence of telomerase activity by the ALT mechanism. The existence of a telomerase-independent telomere maintenance mechanism was first identified in yeast with telomerase deficiency, and this mechanism was demonstrated to be dependent on RAD52, a gene involved in homologous recombination (Lundblad and Blackburn, 1993). Subsequent studies established that the ALT mechanism in human cancer cells requires certain DNA recombination proteins, including the MRN complex (MRE11, RAD50 and NBS1) (Jiang et al., 2005; Zhong et al., 2007), SMC5-SMC6 complex (Potts and Yu, 2007), flap endonuclease 1 (FEN1) (Saharia and Stewart, 2009), MUS81 (Zeng et al., 2009), Fanconi anemia group D2 (FANCD2) and Fanconi anemia group A (FANCA) (Fan et al., 2009). In our model, MRE11 and FANCA genes are indeed up-regulated in ALT+ cells relative to telomerase+ cells (Figure S4B and S4C). These observations provide genetic evidence in mammalian cells of the importance of these factors for DNA recombination-dependent ALT mechanism in this cancer model.

Our studies also highlight that, while ALT+ cells suppress apoptosis and senescence as efficiently as telomerase+ cells, these cells do not fully restore mitochondrial function which we speculate may relate to ongoing genotoxic stress partly associated with less efficient telomere capping by ALT-mediated telomere maintenance (Cesare et al., 2009) as well as p53-independent repression of PGC (Sahin et al., 2011). In ALT+ cells, the importance of this mitochondrial maintenance program is underscored by the up-regulated PGC network signaling on both the genomic (copy number) and transcriptional levels (Figure 7G). That the PGC network is a vital regulator in our ALT+ resistant tumors is further substantiated by our analysis of published transcriptomic data comparing human ALT+ with telomerase+ osteosarcomas (Lafferty-Whyte et al., 2009). Specifically, comparison of three key network genes (PGC-1α, NRF2 and SOD2; PGC-1β data is not available in this dataset) in 10 ALT+ and 8 telomerase+ osteosarcomas showed that five of ten ALT+ osteosarcomas have significantly higher levels of one or multiple of PGC-1α, NRF2 and SOD2 genes relative to the basal levels, which is more prominent than telomerase+ tumors (Figure S6G) (P=0.04), suggesting that up-regulation of mitochondrial function and ROS defense pathway genes is important in other ALT+ cancer types. Thus, our genetic modeling of telomerase extinction in cancer increases our understanding of how tumor cells might respond and adapt to telomerase inhibition and illuminates a clinical path hypothesis utilizing combination regimens targeting telomerase and PGC-mediated adaptive mechanisms in cancer.

EXPERIMENTAL PROCEDURES

Mice

Atm, Terc and TERT-ER mice described previously (Jaskelioff et al., 2011; Maser et al., 2007; Wong et al., 2003) were interbred and backcrossed to high grade of C57BL/6 (over 95%). The mating strategy to obtain experimental cohorts G₀, G₁ and G_3-4 is shown in Figure S1A. 4-OHT time release pellets (5 mg; Innovative Research of America) were inserted subcutaneously at age of 18 weeks to reach steady state blood levels of 1 ng ml^-1 4-OHT for 60 consecutive days.

Histology, tumor characterization and sample preparation

Antibodies used for IHC include anti-Ki67 (Dako), anti-53BP1 (Bethyl Labs), anti-γH2AX (Bethyl Labs) and anti-Cleaved Caspase 3 (Cell Signaling). For FACS analysis, cells were immunostained with CD4, CD8 and CD3 antibodies (eBioscience) and analyzed on a BD FACSCanto II (BD Biosciences).

Cytogenetic analysis

For metaphase preparation, metaphases were obtained from colcemid-treated cells incubated in 105 mM KCl hypotonic buffer for 15 min before fixation in 3:1 methanol–acetic acid. Spectral karyotyping (SKY) was done using the SkyPaint Kit and SkyView analytical software (Applied Spectral Imaging) according to manufacturer’s protocols. For telomere-FISH (fluorescent in situ hybridization), metaphase spreads were applied with telomere-specific T₂AG₃-FITC PNA (peptide nucleic acid) probes and centromere-specific Cent-Cy3 or Cent-pacific blue PNA probes, and counter-stained with DAPI (for microscopy) or TOTO3 (for laser scanning). Laser scanning cytometry quantification was performed with an iCys Research Imaging Cytometer (Compucyte) as described earlier (Jaskelioff et al., 2011; Sahin et al., 2011).

Telomere Sister Chromatid Exchange assay

Telomere Co-FISH staining was performed as previously described (Potts and Yu, 2007). Details are described in Extended Experimental Procedures.

RT-qPCR, mitochondria DNA content measurement and Western blot

RT-qPCR primers are described in Table S4. Mitochondrial DNA content was measured by the relative levels of CoxI vs. beta globin by qPCR. CoxI and beta globin primers are described in Table S4. Antibodies used for western blotting are anti-phospho p53 (ser15, Cell Signaling), anti-p21 (Santa Cruz Biotechnology), anti-Vinculin (Sigma) and anti-Mre11 (BD Biosciences).

shRNA knockdown

Lentivirial based shRNA constructs targeting mouse PGC-1β and SOD2 were ordered from The Dana-Farber/Harvard Cancer Center DNA Resource Core. Verified shRNA sequences were cloned into inducible lentiviral constructs PLKO-TRC-901 (IPTG inducible; Broad Institute) and PLKO-Tet (Doxycycline inducible; Novartis).

Pathway analysis

891 up-regulated and 1345 down-regulated genes with 1.5 fold change and P<0.01 in ALT+ tumors relative to telomerase+ tumors were applied with Knowledge-based Pathway (IPA) analysis. The significantly represented pathways were compared with the pathways obtained from the similar analysis on 10 ALT+ and 8 telomerase+ osteosarcomas (Lafferty-Whyte et al., 2009). Overlapped pathways are listed.

Array-CGH profiling and analyses

Array-CGH, SGOL score analysis, homologous mapping and permutation analysis are described in Extended Experimental Procedures.

Mitochondrial oxygen consumption measurements in live cells

Oxygen consumption rates (OCRs) were measured using the Seahorse XF24 instrument (Seahorse Biosciences). Basal mitochondrial respiration were measured at four time points, and respiration non-linked to mitochondrial ATP synthesis were measured at four time points after adding 1 μM oligomycin. Non-mitochondrial OCRs were obtained by adding 5 μM antimycin A, and uncoupled respiration was obtained by adding 1 μM FCCP.

ROS measurement

For determination of ROS levels, tumor cells were stained with 5 μM CM-H₂DFCDA (Invitrogen) at 37 °C for 30 minutes, followed by FACS analysis on a BD FACSCanto II (BD Biosciences).

COX model test

Effects of shPGC-1β knockdown on Tel+ and ALT+ cells were tested by Cox analysis. Details are described in Extended Experimental Procedures.