Giulio Spinozzi

VISPA: a computational pipeline for the identification and analysis of genomic vector integration sites

The analysis of the genomic distribution of viral vector genomic integration sites is a key step in hematopoietic stem cell-based gene therapy applications, allowing to assess both the safety and the efficacy of the treatment and to study the basic aspects of hematopoiesis and stem cell biology. Identifying vector integration sites requires ad-hoc bioinformatics tools with stringent requirements in terms of computational efficiency, flexibility, and usability. We developed VISPA (Vector Integration Site Parallel Analysis), a pipeline for automated integration site identification and annotation based on a distributed environment with a simple Galaxy web interface. VISPA was successfully used for the bioinformatics analysis of the follow-up of two lentiviral vector-based hematopoietic stem-cell gene therapy clinical trials. Our pipeline provides a reliable and efficient tool to assess the safety and efficacy of integrating vectors in clinical settings.

VISPA2: a scalable pipeline for high-throughput identification and annotation of vector integration sites

BackgroundBioinformatics tools designed to identify lentiviral or retroviral vector insertion sites in the genome of host cells are used to address the safety and long-term efficacy of hematopoietic stem cell gene therapy applications and to study the clonal dynamics of hematopoietic reconstitution. The increasing number of gene therapy clinical trials combined with the increasing amount of Next Generation Sequencing data, aimed at identifying integration sites, require both highly accurate and efficient computational software able to correctly process “big data” in a reasonable computational time.ResultsHere we present VISPA2 (Vector Integration Site Parallel Analysis, version 2), the latest optimized computational pipeline for integration site identification and analysis with the following features: (1) the sequence analysis for the integration site processing is fully compliant with paired-end reads and includes a sequence quality filter before and after the alignment on the target genome; (2) an heuristic algorithm to reduce false positive integration sites at nucleotide level to reduce the impact of Polymerase Chain Reaction or trimming/alignment artifacts; (3) a classification and annotation module for integration sites; (4) a user friendly web interface as researcher front-end to perform integration site analyses without computational skills; (5) the time speedup of all steps through parallelization (Hadoop free).ConclusionsWe tested VISPA2 performances using simulated and real datasets of lentiviral vector integration sites, previously obtained from patients enrolled in a hematopoietic stem cell gene therapy clinical trial and compared the results with other preexisting tools for integration site analysis. On the computational side, VISPA2 showed a > 6-fold speedup and improved precision and recall metrics (1 and 0.97 respectively) compared to previously developed computational pipelines. These performances indicate that VISPA2 is a fast, reliable and user-friendly tool for integration site analysis, which allows gene therapy integration data to be handled in a cost and time effective fashion. Moreover, the web access of VISPA2 (http://openserver.itb.cnr.it/vispa/) ensures accessibility and ease of usage to researches of a complex analytical tool. We released the source code of VISPA2 in a public repository (https://bitbucket.org/andreacalabria/vispa2).

537. New Graph-Based Algorithm for Comprehensive Identification and Tracking Retroviral Integration Sites

Vector integration sites (IS) in hematopoietic stem cell (HSC) gene therapy (GT) applications are stable genetic marks, distinctive for each independent cell clone and its progeny. The characterization of IS allows to identify each cell clone and individually track its fate in different tissues or cell lineages and during time, and is required for assessing the safety and efficacy of the treatment. Bioinformatics pipelines for IS detection used in GT identify the sequence reads mapping in the same genomic position of the reference genome as a single IS but discard those ambiguously mapped in multiple genomic regions. The loss of such significant portion of patients’ IS may hide potential malignant events thus reducing the reliability of IS studies. We developed a novel tool that is able to accurately identify IS in any genomic region even if composed by repetitive genomic sequences. Our approach exploits an initial genome free analysis of sequencing reads by creating an undirected graph in which nodes are the input sequences and edges represent valid alignments (over a specific identity threshold) between pairs of nodes. Through the analysis and decomposition of the graph, the method identifies indivisible subgraphs of sequences (clusters), each of them corresponding to an IS. Once extracted the consensus sequence of the clusters and aligned on the reference genome, we collect the alignment results and the annotation labels from RepeatMasker. By combining the set of genomic coordinates and the annotation labels, the method retraces the initial sequence graph, statistically validates the clusters through permutation test and produces the final list of IS. We tested the reliability of our tool on 3 IS datasets generated from simulated sequencing reads with incremental rate of nucleotide variations (0%, 0.25% and 0.5%) and real data from a cell line with known IS and we compared out tool to VISPA and UClust, used for GT studies. In the simulated datasets our tool demonstrated precision and recall ranging 0.85-0.97 and 0.88-0.99 respectively, producing the aggregate F-score ranging 0.86-0.98 which resulted higher than VISPA and UClust. In the experimental case of sequences from LAM-PCR products, our tool and VISPA were able to identify all the 6 known ISs for >98% of the reads produced, while UClust identified only 5 out 6 ISs. We then used our tool to reanalyze the sequencing reads of our GT clinical trial for Metachromatic Leukodystrophy (MLD) completing the hidden portion of IS. The overall number of ISs, sequencing reads and estimated actively re-populating HSCs was increased by an average fold ~1.5 with respect the previously published data obtained through VISPA whereas the diversity index of the population did not change and no aberrant clones in repeats occurred. Our tool addresses and solves important open issues in retroviral IS identification and clonal tracking, allowing the generation of a comprehensive repertoire of IS.

212. Lentiviral Insertional Mutagenesis Helps to Uncover the Mechanisms of Resistance to AZD9291 and CO-1686 in EGFR-Mutant Lung Adenocarcinoma

The discovery of cancer-driver mutations, accounting for the growth and spreading of cancer cells, led to the development of anti-cancer targeted therapies, which hit in a specific manner cell pathways directly involved in tumor progression. This new class of therapeutic agents has been shown to be more effective and less toxic than conventional chemotherapy in advanced forms of cancer. However, the inevitable development of acquired resistance has limited their success. An example of this concept is the case of Epidermal Growth Factor Receptor (EGFR)-mutant lung cancer. The discovery of EGFR mutations that confer sensitivity to the Tyrosine Kinase Inhibitors (TKI) erlotinib and gefitinib have underlined the importance of defining molecular subgroups to design more efficacious targeted therapies. Unfortunately, on average ~1 year after starting treatment, resistance to these agents, caused by EGFR secondary mutation T790M, occurs at high frequency. This observation suggests that strategies to delay or prevent the emergence of this resistance mechanism would prolong the lives of many lung cancer patients with EGFR mutations. AZD9291 and CO-1686 are two novel third-generation EGFR TKIs designed to irreversibly and specifically target both the initial activating EGFR mutations and the resistance T790M. Phase I/II studies show compelling clinical activity of these compounds. Nevertheless the observed progression-free survival is about 12 months and the reasons of the relapse are under evaluation. We took advantage of a lentiviral vector (LV) -based insertional mutagenesis platform, developed by our lab, to screen genes that confer resistance to CO-1686. To this aim T790M+ (PC9BRc1) and T790M- (PC9) human lung adenocarcinoma cells were transduced with a genotoxic LV, harboring the Spleen Focus Forming Virus enhancer/promoter in the Long Terminal Repeats (LV-SF-LTR) or a non-genotoxic LV with self-inactivating LTR. After 2 weeks, transduced cells were divided in three groups receiving Erlotinib, CO-1686 or DMSO as control. In our rationale, the integration of the genotoxic LV in the cellular genome can potentially deregulate the expression of neighboring genes that contribute to confer resistance to these TKIs. Therefore, exposure to the treatment would result in the selection and expansion of the cell clones harboring those specific traceable mutations. Drug-resistant colonies were obtained after 4 weeks of erlotinib and 6 weeks of CO-1686 treatment. While results on PC9BRc1 are still pending, sequencing analysis performed on 100,000 LV integration sites retrieved by PCR-based technologies on PC9 drug-resistant colonies identified SOS1 as the principal LV-induced gene deregulation responsible for the resistance to both Erlotinib and CO-1686. A similar experiment aimed at identifying the AZD9291-induced resistance in lung cancer cells is on-going. The identification of biomarkers of resistance will allow the development of new therapies to overcome resistance, improving the life expectancy of lung adenocarcinoma patients.

529. Lentiviral Vectors with a Reduced Splicing Interference Potential Have a Significantly Improved Safety Profile In Vivo

Genotoxicity assays based on systemic vector injection into newborn tumor-prone Cdkn2a−/− and Cdkn2a+/− mice has shown that self-inactivating (SIN) lentiviral vector (LV) harboring strong or moderate enhancer/promoters in internal position caused acceleration in hematopoietic tumor onset compared to control mice. Integration site (IS) analysis in vector-induced tumors showed that oncogene activation or tumor suppressor inactivation occurs by mechanisms of aberrant splicing and/or enhancer-mediated overexpression of cellular genes. Although oncogene activation may be reduced by the use of SIN design, moderate cellular promoters and insulator sequences, how to reduce genotoxic splicing-capture events and aberrant transcript formation triggered by vector integration is still unclear. Here, we specifically designed SINLVs harboring sequences complementary to microRNAs (mirT sequence) which are active in hematopoietic cells (mir223 and mir142-3p) within the SIN LTR (mirsT-LTR.LV) or in the vector backbone and outside the gene expression cassette (mirT-LV). In our rationale, the mirT sequences when incorporated in an aberrantly generated mRNA would be selectively degraded through the miRNA pathway. Thus, by taking advantage of our in vivo models, we assessed the genotoxicity of these LVs with mirT sequences. Systemic injection of mirsT-LTR.LV (N=34) and mirT-LV (N=39) in Cdkn2a−/− mice did not cause any significant acceleration in hematopoietic tumor onset compared to un-injected mice (N=37) or mice injected with a SINLV that does not harbor mirT sequences (N=24). Similar results have been obtained after injection of the same vectors in Cdkn2a+/− mice (N=29 mirsT-LTR.LV, N=25 mirT-LV, N=40 un-injected and N=15 injected control mice). To gain additional information on the safety profile of these vectors, we performed IS analysis (N>10,000) in tumor-derived DNA. By this analysis, we previously found that Map3k8 activation by LV insertions was the major mechanism of genotoxicity when prototypical SINLVs were injected into Cdkn2a−/− mice. Now, we found that mice treated with mirsT-LTR.LV and mirT-LV did not show any Map3k8 activating insertions, suggesting that the new vectors are efficient in preventing its activation and confirming their superior safety profile. Furthermore, as expected, Pten was the most frequently targeted gene in tumors derived from Cdkn2a−/− mice injected with the LVs harboring mirT sequences. Pten insertions mainly targeted exons, suggesting the potential inactivation of its transcription unit. Finally, we found that Sfi1 was the major Common Insertion Site (CIS) in Cdkn2a+/− mice injected with LVs harboring mirT sequences. This CIS gene however appears to be the product of an intrinsic bias of LV integration, rather than the result of a selection process. Overall, our studies showed that these new advanced design LVs have a significantly improved safety profile and could represent the vector design of choice in future gene therapy applications.

674. Insertional Mutagenesis to Identify Mechanisms of Cetuximab Resistance in Colorectal Cancer

Anti-cancer drugs designed to target specific molecular pathways have shown an excellent therapeutic potential but also very poor long-term durability of tumor responses, mainly due to the outbreak of resistant clones among the residual neoplastic cell population. For that reason, understanding the molecular mechanisms underlying the onset of anti-cancer drug resistance (ACDR) is one of the major goals of clinical research. ACDR has been widely studied by DNA/RNA sequencing of primary human samples and several culprits identified. We have previously developed an approach based on lentiviral vector (LV)-induced insertional mutagenesis that allowed to identify the genes involved in lapatinib and erlotinib resistance on HER2+ human breast cancer cell lines and EGFR+ pancreatic cell line respectively. Here we took advantage of this platform to investigate ACDR genes in colorectal cancer (CRC). Cetuximab, anti-EGFR monoclonal antibody, is used as first line therapy in metastatic CRC, which results in prolonged survival of treated patients. However, nearly all patients relapse due to ACDR. We thus selected CRC cells sensitive to cetuximab deriving either from five microsatellite stable cell lines or from eight Patient Derived Xenografts (PDX), primary human CRC cells implanted subcutaneously into immunodeficient mice (NSG). To induce insertional mutagenesis we generated a luciferase-expressing LV harboring the SFFV enhancer/promoter in the long terminal repeats able to perturb the expression of genes nearby the integration site. As control, we used a non-genotoxic SIN-LV. We set up a collagenase IV-based disaggregation protocol that allows single-cell suspension and a serum-free culture condition to maintain the stemness of in vitro cultured cells. This protocol allowed to efficiently disaggregate and expand CRC cells in vitro as well as reach a LV copy number per cell ranging from 0.25 to 5.6. Luciferase gene expression was stable and allowed live-animal monitoring for up to 30 weeks after transplant. CRC-0069 and -0077 PDXs and NCI-H508 and HDC82 cell lines were transduced ex vivo and kept in vitro and/or transplanted in NSG mice. After in vitro or in vivo expansion of the transduced CRCs cetuximab treatment was applied. After an initial shrinking of the tumor mass in mice we observed ACDR in 3 out of 10 mice transplanted with NCI-H508 cells transduced with SFFV-LV and in none of the controls. Genomic DNA from resistant cells is being used for insertion site (IS) analysis to identify common IS, ACDR gene candidates. IS obtained from SIN-LV groups will be used to filter LV integration biases, whereas IS from SFFV-LV transduced cells but not treated with cetuximab will be used to filter mutations that provide a proliferative advantage unrelated to cetuximab treatment. We will validate the most promising candidates by LV-mediated overexpression and knockdown techniques. This approach could pave the way to perform insertional mutagenesis-based forward genetics studies on primary human samples.

2. Identification and Ranking of Different Chromatin Insulators to Block Vector-Driven Enhancer-Mediated Insertional Mutagenesis In Vivo

Aimed at refining the safety profile of self-inactivating (SIN) lentiviral vectors (LV) for gene therapy applications we investigated the impact of chromatin insulators (CI) on vector-mediated genotoxicity. Specifically, we studied four recently identified CI whose function is mediated by CCCTC-binding factor (CTCF), the best characterized insulator protein in vertebrates, and cloned these CI in the LTRs of a SIN. LV with a strong enhancer/promoter in internal position (CI. SIN. LV).We took advantage of two sensitive in vivo genotoxicity assays based on the systemic injection of LVs in newborn tumor-prone Cdkn2a−/− and Cdkn2a+/− mice, that allow to measure vector-induced genotoxicity as accelerated tumor onset proportional to the genotoxic potential of the tested LV.CI. SIN. LVs displayed slightly not statistically significant improvement in the median survival time vs. the uninsulated SIN. LV counterpart (ranging from 193.5 to 214 days vs. 186 days, respectively) in Cdkn2a−/− mice.In Cdkn2a+/− mice, two insulated vectors significantly improved the median survival time, which resulted non-statistically different from Mock mice (450 and 511 vs. 505.5 days respectively), while the other two CI. SIN. LVs studied resulted to be still slightly genotoxic (median survival time: 412 and 429.5 days). To gain more insights on the safety profile of these LVs we retrieved and analyzed the vector integration sites (IS) (n>14000 IS) and identified common integration sites (CIS) in the murine tumors generated in our experimental framework and in both murine models.In Cdkn2a−/− mice, uninsulated SIN. LV-induced tumors harbored activating integrations targeting Map3k8 oncogene, while tumors obtained in mice treated with two out of four different insulated LVs significantly reduced the frequency of tumors with Map3k8-activating insertions. The reduced targeting frequency of Map3k8 was accompanied by a skewing of integrations inactivating Pten, Rasa1 or other tumor-suppressors, an escape genotoxicity mechanism on which insulators cannot act.In Cdkn2a−/+ mice we identified different predominant CIS genes targeted by the different insulated vectors. These data show that heterozygous Cdkn2a−/+ mice allow discriminating between more subtle shades of genotoxicity of the different vector designs and are therefore instrumental to understand the different molecular mechanisms of insertional mutagenesis and ways to avoid them. Interestingly by comparing the results from both in vivo assays we observed that one CI displayed superior safety profile in terms of significant improvement in the median survival time and/or in terms of reduced oncogenic CIS identified.In summary we validated new human-origin insulator elements able to block SIN. LV genotoxicity in vivo. Overall, these data highlight the importance of stringent in vivo genotoxicity testing of improved vector versions and support the use of CI for future gene therapy applications.

530. Development of New Lentiviral Vectors With a Reduced Splicing Interference Potential and a Safer In Vivo Genotoxic Profile

The excellent therapeutic potential of self-inactivating (SIN) lentiviral vectors (LV) has been demonstrated in pre-clinical studies and clinical trials. However, weaker mechanisms of insertional mutagenesis could endanger their clinical applications. Systemic vector injection into newborn tumor-prone Cdkn2a-/- and Cdkn2a+/-mice, conducted in our previous work, demonstrated that SINLVs harboring strong or moderate enhancer/promoters in internal position caused acceleration in hematopoietic tumor onset with respect to control mice. Integration sites analyses of vector-induced tumor showed that oncogene activations or tumor suppressor inactivation by LV integrations occur by combining mechanisms of transcript truncation, induction of aberrant splicing and/or enhancer-mediated overexpression of cellular transcription units. Although oncogene activation may be reduced by the use of self-inactivating design, moderate cellular promoters and insulator sequences how to reduce genotoxic splicing-capture events and aberrant transcript formation triggered by vector integration is still unclear.From this and a previous study, we identified the LV sequences most frequently involved in chimeric transcript formation. In our rationale, these LV sequences could be tagged by sequences complementary to microRNAs (mirT sequence) active in hematopoietic cells in order to allow selective degradation, through the miRNA pathway, of vector-mediated aberrantly spliced transcripts. Hence, we specifically designed SIN LVs harboring mirT sequences recognized by mir223 and mir142-3p (that are expressed in hematopoietic lineages) within the SIN LTR (mirsT-LTR LV) or in the vector backbone and outside the gene expression cassette (mirT LV). We then assessed the genotoxicity of the SIN LVs harboring mirT sequences by taking advantage of our in vivo models. Interestingly, injection of mirsT-LTR LV (N=73) and mirT LV (N=73) in Cdkn2a-/- mice did not caused any significant acceleration in hematopoietic tumor onset compared to control un-injected mice (N=40). Similar results have been obtained after injection in Cdkn2a+/- mice (N=28 for mirsT-LTR LV, N=26 for mirT LV and N=34 un-injected mice). We are currently performing integration site analyses in Cdkn2a-/- and Cdkn2a+/- treated mice to dissect if and how the integrated mirsT-LTR LV and mirT LV proviral genome interacts with the surrounding cellular genome.Overall, these studies show that this new advanced design lentiviral vectors completely abrogated residual vector genotoxicity in highly sensitive mouse models and could represent the vector design of choice in future gene therapy applications.

23. Uncovering Mechanisms of Resistance to Cetuximab by Insertional Mutagenesis in Heterotopically-Engrafted Human Colorectal Cancers

The development of high-throughput technologies has made possible the identification of cancer-driver mutations, which account for the growth and spreading of cancer cells. The discovery of these specific cancer biomarkers led to the development of anti-cancer targeted therapies, which hit in a specific manner cell pathways directly involved in tumor progression. This new class of therapeutic agents, which comprehend small molecules and antibodies, has been shown to be more effective and less toxic than conventional chemotherapy in advanced forms of cancer. However, the inevitable development of acquired resistance, due to the acquisition of multiple mutations or activation of compensatory pathways, has limited their success. A paradigm for this concept are the anti-Epidermal Growth Factor Receptor (EGFR) monoclonal antibodies, cetuximab and panitumumab, which are Federal Drug Administration (FDA)-approved agent for the treatment of EGFR-expressing metastatic colorectal cancer (mCRC). Unfortunately, despite the massive initial reduction, response is transient and tumors become refractory within 12-18 months.We take advantage of a lentiviral vector (LV)- based insertional mutagenesis platform to induce cetuximab-resistance in Patient-Derived Xenografts (PDXs) of mCRC, with the final goal of uncovering the molecular mechanism of the resistance. Insertional mutagenesis has been successfully used in our lab to explore novel genes involved in the resistance to lapatinib in two breast cancer cell lines and to erlotinib in a pancreatic cancer cell line. Given these encouraging results we now aim to apply this technology to human cancer specimens to better reflect the real clinical response. Established PDXs from cetuximab-sensitive liver metastatic lesions of CRC have been excised, chemically digested to single cell suspension and consequently transduced with a genotoxic LV, harboring the hyperactive enhancer/promoter of Spleen Focus Forming Virus in the Long Terminal Repeats (LV-SF-LTR) sequences or a non-genotoxic LV, containing self inactivating LTR sequences. Transduced tumor cells have been re-implanted in NSG mice to reconstitute the tumors. Once engrafted, mice have been divided in two groups receiving either cetuximab or the drug-vehicle. LV-SF-LTR transduction of a large amount of cells with a high vector copy number will allow us to randomly hit genes responsible for the resistance to the tumor-specific targeted therapy and deregulate their expression. Therefore, exposure to the treatment will not rest tumor growth in these tumors. LAM-PCR and deep sequencing analysis performed on drug-sensitive and drug-resistant tumors will be used to map the integration sites of the LVs and consequently to identify LV-induced gene deregulations responsible for the pharmacological resistance.The identification of the biomarkers accounting for the drug-resistance is a real challenge that will allow screening for rational drug combinations to reverse resistance, improving the life expectancy of CRC patients.

535. Increasing Accuracy and Precision of Vector Integration Site Identification of Sequencing Reads With a New Bioinformatics Framework

In hematopoietic stem cell (HSC) gene therapy (GT) applications patients are transplanted with autologuos HSCs that have been ex-vivo genetically modified with integration competent vectors to express a therapeutic transgene. Specific PCR techniques coupled to next generation sequencing and bioinformatics analysis allow the high throughput retrieval, sequencing and mapping of proviral/genomic DNA junctions present in the blood and bone marrow derived cell populations sampled at different time points after therapy. The increase in sequences available for IS mapping is accompanied by an increase in false positives derived by sequencing errors or sequencing read parsing and mapping on the reference genome. In particular, by analyzing IS datasets form vector marked human and mouse tumor cells, clones with defined integration sites and GT patients, we observed that when multiple sequences arising from the same IS are aligned on the reference genome >10% mapped near (+/- 4 bases) the true insertion site. Without correction, these misaligned sequences not only result in an overestimation of the overall number of IS but in some cases also in the generation of false common insertion sites, worrisome hallmarks of insertional mutagenesis. To mitigate this issue we and others, based on empirical observations, merge sequencing reads mapping within +/- 3 bp into a single IS. Although this adjustment reduces the impact of the “wobbling” around the true ISs, a dedicated method and model is still missing.To further increase the accuracy of genomic positioning of sequencing reads we developed a new bioinformatics framework as post-processing plugin for pipelines that correctly partitions sequencing reads in a given genomic position by considering the relative abundance and distribution of each sequence cluster using local modes and Gaussian scores through an adaptive approach that varies the parameters of the Gaussian curve and proposes different solutions. To chose the best solution, the algorithm first evaluates each solution by exploiting 100 simulations of the input reads and then selects the resulting best solution using the Kolmogorov-Smirnov test. The simulation step is designed to test the mappability of the IS genomic interval and to quantify the impact of the observed nucleotide variations of the reads with respect to the reference genome (PCR artifacts or real genomic differences) that may lead to different mapping results that justify a larger span of the mapped reads surrounding the putative IS. The algorithm returns the list of IS and relative number of reads with the p-value of the best solution.We performed 3 ad-hoc in vitro experiments on a cell clone with 6 known IS in which we measured the precision of IS placement obtaining an average of 100% with our new method whereas <30% using our previous method based on a rigid sliding window approach of 4 bp. We applied our new approach to our clinical trial datasets obtaining improvements in IS genomic placement and overestimation with a reduction of potential false IS of 3% without changing the biological results.