Page 165 - Haematologica May 2020

Resistance to Notch1 neutralizing antibody
consistent with marginal variations in genes involved in lipid metabolism in these models (Online Supplementary Figure S13). However, acute treatment of PDTALL19 mice with OMP52M51 was also associated with reduced NOTCH1 FL and ICD levels (Online Supplementary Figure S12), indicating that these changes were related to NOTCH1 blockade rather than to resistance. Finally, given the marked changes in lipid metabolism seen in the PDTALL19 model, we investigated other possible pheno- typic changes associated with resistance. By flow cytom- etry analysis, we found that OMP52M51-resistant cells exhibited a smaller size compared with control cells and there were also significant changes in surface expression of the T-cell markers CD3 and CD7 and the pan-leuco- cyte marker CD11a (Online Supplementary Figure S14).
Assessment of the genetic background of OMP52M51 resistance in the PDTALL8 model
In the case of PDTALL8 model, transcriptional data sug- gested that mechanisms underlying the stable resistance to OMP52M51 treatment could involve an “on target” mutation of the Notch pathway selected during treat- ment and serial transplantation experiments, leading to the loss of sensitivity to OMP52M51.
To investigate this hypothesis, we performed single nucleotide polymorphism (SNP) arrays and whole exome sequencing (WES) of paired control-resistant mice (see Online Supplementary Table S5 for WES metrics details), allowing the identification of variants that could be not detected by Sanger sequencing due to a relatively low variant frequency. Cytoscan arrays failed to identify copy number variations associated with resistance to OMP52M51 in PDTALL8 cells (Online Supplementary Figure S15). However, bioinformatics analysis of WES revealed that control mice displayed a higher tumor mutational load than OMP52M51-resistant samples, both considering the total number of variants (34,641 variants in controls compared to 12,206 in OMP52M51-resistant samples) and shared confident non-synonymous calls (440 vs. 54; Figure 6A-B and Online Supplementary Table S6). This difference could be explained by the “tumor clonal selection” model. Speculatively, the OMP52M51 antibody could act as a selective agent, favouring out- growth of a subpopulation of cells from the initial tumour. Interestingly, WES analysis highlighted the pres- ence of two NOTCH1 activating variants mapping to the heterodimerization domain (HD),16 i.e. p.Q1584H and p.L1585P, present only in the OMP52M51-resistant mice (Figure 6C). We validated these mutations by Sanger sequencing and extended analysis to additional samples from the same experiment (four to five mice/group; Online Supplementary Figure S16). Sanger sequencing con- firmed that both mutations were present in OMP52M51 resistant mice and were lacking in controls (Figure 6D).
To investigate whether these mutations occurred at low level in parental cells, we performed targeted sequencing analysis. All treated xenograft (three repli- cates/group) presented p.L1585P and p.Q1584H variants in cis (Online Supplementary Figure S17). On the other hand, at a depth of 300X the p.L1585P and p.Q1584H variants were not identified in control samples and there- fore, if present prior to treatment, must have been present at a frequency of less than 5%, which was the estimated sensitivity of the next-generation sequencing (NGS) assay.
Finally, according to the literature, these mutations destabilize the structure of the HD domain17 and there- fore could affect binding of OMP52M51. This was indeed shown by flow cytometry experiments showing that the binding of OMP52M51 is lower in PDTALL8 resistant compared with control cells (Figure 6E).
Discussion
In the last 10 years, personalized treatment of cancer has improved substantially thanks to the identification of specific genetic alterations and consequent development of target therapies against oncogenic drivers. In this land- scape, the resistance of cancer cells to pharmacological treatment remains the major challenge to face in order to increase the efficacy of new drugs.18 Although we are aware of some intrinsic limitations of xenografts, such as the lack of the immune system, the systemic T-ALL mod- els used in this study are suitable to study effects of drugs directly targeting tumors cells, and we used them to investigate the molecular mechanisms of resistance to Notch targeted therapy. Our results demonstrate that the resistance appeared following long-term administration of OMP52M51 antibody in each of the three PDX models tested, though with different timing. In fact, PDTALL8, the PDX with a late onset of resistance and loss of Notch signaling inhibition, was characterized by a stable resist- ance. On the contrary, in the PDTALL19 model, resist- ance appeared earlier, Notch signaling was inhibited and, importantly, it was an unstable trait. Analysis of the slope of the percentage of T-ALL cells in blood of these mice during serial drawings suggested that adaptation to OMP52M51 in this model consisted mainly of a delayed but constant growth of the PDX without the develop- ment of true resistance during the first round of treat- ment. However, upon repeated rounds of treatment with OMP52M51 stable resistance eventually occurred, sug- gesting a two-stage form of resistance in this model. PDTALL11 disclosed an intermediate behavior both regarding the time of development of resistance, the sta- bility of resistance and Notch signaling inhibition, likely due to a mixture of different mechanisms. Speculatively, an additional round of treatment might lead to the selec- tion of a completely resistant clone also in the PDTALL11 model, although this was not investigated here.
Intriguingly, previous studies elegantly addressed the issue of clonality of T-ALL xenografts and correlated the genetic complexity of T-ALL cells with the speed of leukemia development in xenograft models.19,20 Specifically, in the case of delayed T-ALL growth, leukemia cells were in part genetically diverse, the result- ing xenograft leukemia arising from different but branched subclones present in the original sample. Although not investigated in our study, it could be specu- lated that the clonal architecture of the PDX might have an influence on the time of development and the type of resistance to anti-NOTCH1 therapy.
In the PDX tested, the resistance was never associated with the loss of PTEN, AKT activation or mutations in FBW7 gene, which represent some of the previously described mechanisms of resistance to Notch inhibition by GSI.7,13 The strikingly different phenotypes of the PDTALL8 and PDTALL19 models underscored two novel mechanisms of resistance. In the case of PDTALL19, we
haematologica | 2020; 105(5)
1325