Defining molecular risk in ALK+ NSCLC

Anaplastic lymphoma kinase (ALK)-positive non-small-cell lung cancers (NSCLC) have the best prognosis among metastatic pulmonary malignancies, with a median patient survival currently exceeding 5 years. While this is definitely a major therapeutic success for thoracic oncology, it may not be entirely attributable to rapid drug development and the strenuous clinical efforts. At the genetic level, ALK+ disease is also unique, distinguished by the lowest tumor mutational burden (mean below 3 mutations/Mbp), the lowest frequency of TP53 mutations (20–25%) and very few other co-mutations compared to other NSCLC. The relative simplicity and stability of the genetic landscape not only contribute to the relatively favourable clinical course, but also make study of the effects from individual molecular features easier. EML4-ALK fusion variant 3 (E6;A20) and TP53 mutations were recently identified as main molecular determinants of adverse outcome: they occur in about 30–40% and 20–25% of newly-diagnosed cases, respectively, have possibly synergistic effects and are independently associated with more aggressive disease, shorter progression-free survival under treatment with ALK inhibitors and worse overall survival. Secondary detection of TP53 mutations at disease progression in previously negative patients defines another subset (about 20%) with similarly poor outcome, while detection of ALK resistance mutations guides next-line therapy. As our biological understanding deepens, additional molecular risk factors will be identified and refine our concepts further. The translation of clinical risk at the molecular level and the ability to predict early events are of key importance for individualized patient management and preclinical modeling in order to advance therapeutic options.

The question about molecular risk in anaplastic lymphoma kinase (ALK)-positive non-small cell lung cancer (NSCLC) is mainly posed by the recent therapeutic advances: prior to the availability of ALK tyrosine kinase inhibitors (TKI) and other targeted therapies, metastatic NSCLC was a rapidly lethal disease with a median overall patient survival (OS) below one and a half years [1]. In contrast, under sequential treatment with ALK TKI the median life expectancy of ALK + lung cancer patients currently exceeds 5 years [2]. This impressive extension of patient life-span creates both the opportunity and the necessity to characterize early events, as their mechanistic

Research Perspective
understanding, prediction and tailored management will be crucial for further therapeutic progress.
Based on several retrospective analyses, it is well known that clinical parameters, e.g., advanced age [3], male sex [3], current smoking [2] and worse performance status [2], can predict worse survival of ALK + NSCLC patients. These associations are plausible considering the naturally limited life expectancy of older individuals, the longer survival for women of any age [4], and the experience with other NSCLC, including EGFR + lung adenocarcinoma, in which a positive smoking history and worse clinical condition at baseline are also associated with inferior outcome [5][6][7]. However, unfortunately, at the same time, predictive and prognostic implications of clinical variables are of limited utility, since clinical factors are neither a good source of mechanistic insights, nor informative for causal therapies that would improve the course of disease in individual patients.
Therefore, the translation of clinical profiles associated with higher risk into molecular features is an important, but challenging task in ALK + NSCLC. Special obstacles include the rarity and genetic heterogeneity of the disease due to multiple ALK fusion variants [8], which are further potentiated by its complex management, including highly variable sequences of TKI and local ablative treatments [9], long patient survival of several years [2], limited availability of tissue from small biopsies, and variable ability of next-generation sequencing (NGS) assays to detect gene fusions in tissue or circulating tumor DNA (ctDNA) [10].
Recently, however, several studies combining state-of-the-art molecular profiling with detailed clinical annotation of large patient cohorts with long clinical follow-up identified two key molecular risk factors in ALK + NSCLC: EML4-ALK fusion variant 3 (E6;A20) [11][12][13][14][15][16] and the presence of TP53 mutations [17][18][19]. Among newly diagnosed patients, these genetic events occur independent from each other in about 30-40% and 20-25% of cases, respectively, have synergistic effects and are both associated with shorter progression-free survival (PFS) after treatment with first-and second-generation ALK inhibitors and with worse OS (Table 1) [11,16,18]. In addition, detection of TP53 mutations in tissue or liquid rebiopsies at the time of disease progression in previously TP53 negative patients, identifies another approximately 20% of cases with a poor outcome, comparable to that with primarily TP53 mutated tumors (TP53 status conversion in approximately 25% of cases x initially wildtype TP53 result in approximately 75-80% of cases) [20].
Thus, the biology of ALK + NSCLC displays some basic similarities with that of the other major oncogenedriven lung cancer subtype, namely EGFR + NSCLC, in which the oncogene variant (e.g. exon 19 indels vs. other alterations [6]) and the presence of TP53 mutations [21] influence benefit from TKI and patient survival, as well [22]. However, there are important differences. First, while in EGFR + NSCLC the oncogene variants, such as exon 19 indels, L858R, "rare" point mutations and exon 20 insertions, cause a largely similar oncogenic drive [23][24][25], which nevertheless translates into a different clinical course only after institution of EGFR-directed therapies due to differential TKI sensitivity [26], the unfavourable EML4-ALK V3 variant in ALK + NSCLC has a different biology per se. There is evidence that the shorter V3 oncoprotein is more stable [11,27,28], causes stronger ALK phosphorylation [11] and promotes cell motility and metastasis more efficiently [13,16], resulting in a higher a priori clinical risk [29]. These data are supported by clinical observations: a higher incidence of metastatic disease [13] and as a higher number of metastatic sites in stage IV patients [14] at diagnosis, i.e. a more adverse course before and independent of the type of treatment ( Figure 1A and Table 1) [16,29]. Interestingly, in keeping with the finding of earlier and broader metastatic dissemination, the presence of V3 is also associated with a worse performance status of newly diagnosed ALK + NSCLC patients ( Figure 1B), which has already been recognized as a clinical risk factor in ALK + disease [2], but is itself neither biologically insightful nor druggable. This observation further underlines the importance of defining disease risk at the molecular level in order to facilitate therapeutic advances. Of note, preclinical and limited clinical data suggest that besides EML4-ALK V3, other "short" EML4-ALK variants, such as V5 (E2;A20) [11], and non-EML4-ALK fusions [30] are also associated with worse outcome, while the longer EML4-ALK V2 (E20;A20) appears to be favourable [31]. The stronger kinase effects and weaker suppression of the V3 oncoprotein by first-and secondgeneration ALK inhibitors [11] may also facilitate earlier TKI escape through the development of ALK resistance mutations, which are another salient and clinically relevant characteristic of V3-driven disease [12]. Their occurrence depends not only on the type of ALK fusion, but also on TKI sequencing and is important for the choice of nextline therapy [9,32,33]. A recently published analysis in a subset (for example 37% or 112/303 cases with detection of V1, V2, V3 in tissue) of patients from the ALEX trial suggested a superior outcome with alectinib compared to crizotinib for patients with all three main EML4-ALK fusion variants V1, V2, V3, but also a trend that the benefit from alectinib, i.e. the response rate (p = 0.10) and the PFS (p = 0.11), might be lower in case of non-V1 EML4-ALK variants, as detected in tissue samples [34]. Since these data are still immature (based on the data cut-off of December 2017), and the patient subgroups are small (n = 8-25, i.e. smaller than in previous retrospective analyses), these results have to be interpreted with some caution.
unique features of ALK + tumors distinguishing them from all other NSCLC [18,38]. Since TMB of oncogene-driven lung cancer is higher in the presence of TP53 mutations ( Figure 2C), the two parameters appear to be linked, possibly through DNA damage facilitated by TP53 loss due to genetic instability [39]. Such an accumulation of genetic abnormalities is crucial for the development of TKI failure [40][41][42], which is associated not only with the presence of TP53 mutations [21], but also with a higher TMB in EGFR + NSCLC [43]. For example, specific co-occuring genetic alterations acquired with disease progression (e.g. in CTNNB1 and PIK3CA), were shown to cooperatively promote tumor metastasis, while other evolutionary paths impair the apoptotic response and cause resistance to EGFR inhibitors [44,45]. According to very recent data on HER2-amplified gastroesophageal adenocarcinoma, increased genomic complexity can reduce efficacy of targeted therapies even when more than one target is hit at the same time [46,47]. Future studies are warranted to analyze whether this observation can be conceptualized: is the degree of intratumoral heterogeneity a predictor of response to targeted drugs across cancer types independent of the druggable target itself [48]? Conversely, the lower baseline TMB and TP53 mutation rate of ALK + compared to EGFR + NSCLC (Figure 2A, 2B) suggest a more "benign" biology. Consistent with this notion, ALK + NSCLC patients treated with even just one TKI have a longer OS than TKI-treated EGFR + NSCLC patients ( Figure 3, Table 2), and survival has generally been longer in trials of ALK inhibitors compared to trials of EGFR inhibitors (for example median OS was > 45 months in the Profile 1014 trial of the first-generation ALK inhibitor crizotinib [49] vs. 28 months in the LUX-Lung-3 study of the second-generation EGFR inhibitor afatinib [50]). This lower genetic complexity of ALK-driven NSCLC compared to other lung tumors may explain why therapeutic progress has been much faster and survival gains much larger for ALK + compared to other lung cancer patients. Interestingly, the worse outcome of smoker ALK + and EGFR + NSCLC patients [2,6] appears to correlate with a higher TMB [51], which illustrates again how "traditional" clinical risk factors can be redefined at the molecular level in order to promote deeper understanding of basic pathogenetic processes. The scarcity of additional genetic alterations is presumably an important reason why TP53 mutations have a major effect on the clinical phenotype of ALK + NSCLC: they are associated with increased disease aggressiveness and metastatic dissemination synergistically with EML4-ALK V3, and they are linked with shorter PFS under TKI and shorter OS independently from EML4-ALK V3, so that double positive V3 + TP53 + patients have a very high risk of death with a median OS of around 2 years in our series (Table 1) [18]. In contrast, a predictive and prognostic impact of TP53 mutations has been hard to discern in case of EGFR/ALK-negative NSCLC [52], in which the much higher number of genetic alterations at baseline [53] presumably obscures the effect of TP53 status and dilutes the consequences of genetic instability [39]. A analogous difference becomes apparent if solid cancers [54,55] are collectively considered against hematologic malignancies, for example acute myeloid leukemia, multiple myeloma, chronic lymphocytic leukemia and mantle-cell lymphoma: in the latter TMB is generally lower (median <3 mutations/ Mbp) [56] and TP53 abnormalities are less frequent (generally <10-15%), but more important for clinical course and crucial for patient management [57][58][59][60]. In the model disease of precision medicine, chronic myeloid leukemia, TP53 abnormalities and other cytogenetic aberrations or co-mutations are also associated with clonal evolution, TKI failure and poor outcome [61][62][63]. Transgenic mouse models of oncogene-, for example Kras-or Egfr-driven NSCLC, demonstrate this principle nicely through a paucity of concomitant genetic alterations [64], but a dramatic phenotypic change upon TP53 loss with metastases and earlier death [65]. Presumably along the same lines, TP53 alterations impair outcome of ALK + NSCLC patients more in case of TKI treatment compared to chemotherapy [18], which is itself genotoxic [66].
Thus, although scarce, concomitant genetic alterations appear to be another important determinant of tumor biology and patient outcome in ALK-driven NSCLC, beside the oncogenic driver, i.e. the ALK fusion itself. For example, KRAS amplifications promote resistance to ALK inhibitors by activating RAS-MAPK signaling [67], which is amenable to SHP2 inhibition [68]. A similar picture emerges for EGFR + NSCLC, in which however the spectrum of oncogene alterations and comutations is much broader [43,69]. The independent and possibly synergistic effects of both the ALK fusion variant and TP53 mutations (as well as other, yet to be identified molecular features) on the clinical course of ALK + NSCLC patients [18] mean that considerable biological and clinical variability is to be expected, if a study would take only one of these molecular factors into account. For effective guidance of patient management that is based on the molecular properties of the tumor, broad profiling approaches will be required, which could for example utilize combined targeted RNA and DNA NGS [22].
trials. Moreover, in vitro and animal studies of the tumorpromoting effects from different ALK fusion variants and TP53 mutations in ALK + disease will be instrumental for deeper mechanistic insights towards uncovering of therapeutic susceptibilities and new drug development.
In summary, ALK + NSCLC is currently spearheading the advent of "precision medicine" in thoracic oncology [11-14, 16, 18, 19, 27, 28, 30]. Distinguished by the lowest genetic complexity and the longest patient survival among NSCLC, ALK + disease is serving as a model for the illustration of basic biological principles and for the development of novel therapeutic strategies, which will probably prove useful in other lung cancer subtypes and tumor entities as well. As our understanding of its pathogenesis deepens, additional molecular features critical for patient outcome will be identified and used to further refine our concepts. Elaboration of a baseline molecular risk stratification-complemented by profiling and targeting of treatment resistance-is a crucial step towards tailored, more effective patient management and better preclinical modeling that will foster therapeutic progress.

ACKNOWLEDGMENTS
We would like to thank Petra Kettenring of our Clinical Trial Unit for assistance with the collection of patient data and samples, and we thank all technical assistants of the Center for Molecular Pathology (CMP) for expert handling of tissue and technical support. tyrosine kinase inhibitor (TKI)-treated ALK + (n = 74 with >1 TKI, n = 109 with just 1 TKI) and EGFR + (n = 344) NSCLC patients, along with a random sample of n = 40 EGFR/ALK-wildtype NSCLC patients that received chemotherapy in the Thoraxklinik at Heidelberg University Hospital [22]. Basic clinical and treatment characteristics of ALK + and EGFR + patients are shown in Table 2. Median overall survival was 65 months for ALK + patients that received >1 TKI, 40 months for ALK + patients that received just 1 TKI, 25 months for TKItreated EGFR + patients and 10 months for wildtype patients; p < 0.001 across groups and p = 0.043 for the comparison between the EGFR + and ALK/1-TKI subgroups with a log-rank test; WT: wildtype; CHT: chemotherapy.