Article Text

Original research
Real-world pan-cancer landscape of frameshift mutations and their role in predicting responses to immune checkpoint inhibitors in cancers with low tumor mutational burden
  1. Vaia Florou1,
  2. Charalampos S Floudas2,
  3. Asaf Maoz3,
  4. Abdul Rafeh Naqash4,
  5. Carter Norton1,
  6. Aik Choon Tan5,
  7. Ethan S Sokol6,
  8. Garrett Frampton6,
  9. Heloisa P Soares1,
  10. Sonam Puri1,
  11. Umang Swami1,
  12. Breelyn Wilky7,
  13. Peter Hosein8,
  14. Jonathan Trent8,
  15. Gilberto de Lima Lopes8,
  16. Wungki Park9 and
  17. Ignacio Garrido-Laguna1
  1. 1Medicine, University of Utah Health, Huntsman Cancer Institute, Salt Lake City, Utah, USA
  2. 2Center for Immuno-Oncology, National Cancer Institute, Bethesda, Maryland, USA
  3. 3Medical Oncology, Dana-Farber Cancer Institute, Boston, Massachusetts, USA
  4. 4Medical Oncology/TSET Phase 1 Program, The University of Oklahoma Stephenson Cancer Center, Oklahoma City, Oklahoma, USA
  5. 5Oncological Sciences and Biomedical Informatics, University of Utah Health, Huntsman Cancer Institute, Salt Lake City, Utah, USA
  6. 6Foundation Medicine Inc, Cambridge, Massachusetts, USA
  7. 7Medicine, University of Colorado Denver Health Sciences Center, Aurora, Colorado, USA
  8. 8Medicine, Sylvester Comprehensive Cancer Center, Miami, Florida, USA
  9. 9Medicine, Memorial Sloan Kettering Cancer Center, New York, New York, USA
  1. Correspondence to Dr Vaia Florou; vaia.florou{at}


Background Pembrolizumab is FDA approved for tumors with tumor mutational burden (TMB) of ≥10 mutations/megabase (mut/Mb). However, the response to immune checkpoint inhibitors (ICI) varies significantly among cancer histologies. We describe the landscape of frameshift mutations (FSs) and evaluated their role as a predictive biomarker to ICI in a clinical cohort of patients.

Methods Comprehensive genomic profiling was performed on a cohort of solid tumor samples examining at least 324 genes. The clinical cohort included patients with metastatic solid malignancies who received ICI monotherapy and had tumor sequencing. Progression-free survival (PFS), overall survival, and objective response rates (ORR) were compared between the groups.

Results We analyzed 246,252 microsatellite stable (MSS) and 4561 samples with microsatellite instability across solid tumors. Histologies were divided into groups according to TMB and FS. MSS distribution: TMB-L (<10 mut/Mb)/FS-A (absent FS) (N=111,065, 45%), TMB-H (≥10 mut/Mb)/FS-A (N=15,313, 6%), TMB-L/FS-P (present ≥1 FS) (N=98,389, 40%) and TMB-H/FS-P (N=21,485, 9%). FSs were predominantly identified in the p53 pathway. In the clinical cohort, 212 patients were included. Groups: TMB-L/FS-A (N=80, 38%), TMB-H/FS-A (N=36, 17%), TMB-L/FS-P (N=57, 27%), TMB-H/FS-P (N=39, 18%). FSs were associated with a higher ORR to ICI, 23.8% vs 12.8% (p=0.02). TMB-L/FS-P had superior median PFS (5.1 months) vs TMB-L/FS-A (3.6 months, p<0.01). The 12-month PFS probability was 34% for TMB-L/FS-P vs 17.1% for TMB-L/FS-A.

Conclusions FSs are found in 47% of patients with MSS/TMB-L solid tumors in a pan-cancer cohort. FS may complement TMB in predicting immunotherapy responses, particularly for tumors with low TMB.

  • tumor biomarkers
  • immune checkpoint inhibitors

Data availability statement

Data are available on reasonable request. Data are available on reasonable request from the corresponding author.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • A tumor mutational burden (TMB) cut-off of 10mut/Mb is used to select patients with advanced solid malignancies for immune checkpoint inhibitor (ICI) treatment. However, patients with select tumors with TMB below this cut-off can still respond to ICI. Biomarkers to identify which patients with tumors harboring low TMB will respond to ICI are needed. Frameshift mutations have been shown to enhance immunogenicity.


  • Here, we describe the landscape of frameshift mutations using a large real-world genomic cohort of solid tumors. These mutations are frequently found across a variety of solid tumors and have a role in predicting ICI responses particularly in tumors with low TMB.


  • Prospective studies are warranted to validate our findings, which if validated may expand the indication of ICI in patients with otherwise limited treatment options.


Immune checkpoint inhibitors (ICIs) include antibodies targeting programmed cell death 1 (PD1) or the ligand PD1 ligand (PDL1) and the cytotoxic T lymphocyte antigen 4 (CTLA4). They have transformed the treatment landscape of numerous solid and hematological malignancies in recent years. Despite the successful application of ICI in many cancers, most patients do not derive clinical benefit from ICI.1

Identifying biomarkers of response to ICI has been challenging despite multiple past and ongoing efforts. For solid tumors, there are currently two tissue agnostic biomarkers for which ICI has been approved: mismatch repair deficiency (MMR) resulting in microsatellite instability (MSI)2 and a tumor mutational burden of 10 or more mutations per megabase (TMB-H).3

Pembrolizumab was the first tissue/site agnostic approval by the FDA in 2017 for all tumors that are MSI, based on a cohort of 149 patients enrolled across 5 clinical trials (KEYNOTE-016 (NCT01876511), KEYNOTE-164 (NCT02460198), KEYNOTE-012 (NCT01848834), KEYNOTE-028 (NCT02054806), and KEYNOTE-158 (NCT02628067)).4 The objective response rate (ORR) was 39.6% (95% CI 31.7% to 47.9%), and 78% of patients had responses lasting for 6 months or more.4

In 2020, the Food and Drug Administration (FDA) approved pembrolizumab in TMB-H solid tumors based on the KEYNOTE-158 multicohort clinical trial.3 The tissue TMB-H was predefined as at least 10 mutations per megabase. Among the 805 patients who were evaluable for TMB in this study, 13% had tTMB-H status, and the ORR was 29% (95% CI 21% to 39%) vs 6% in patients with TMB<10 mutations per megabase (95% CI 5 to 8).3

TMB represents the number of mutations per megabase of DNA within the tumor cells. High TMB has been shown to confer more immunogenicity in terms of neoantigens and is known to correlate with favorable outcomes to ICI.5 6 Despite pembrolizumab’s approval across tumor types which are TMB-H, its clinical activity may not be consistent across all cancer types. In the KEYNOTE-158 trial, the majority of patients had either small cell lung (32%), cervical (15%), endometrial (14%), vulvar (14%), or anal (13%) cancer.3 7 While the ORR of all patients was 29%, patients with certain types of cancers exhibited trends toward significantly less benefit, that is, only 1/14 patients with anal cancer TMB-High responded vs 8/75 with non-TMB-High and 0/1 patients with mesothelioma TMB-High responded vs 9/84 with non-TMB-High.4 Despite the low number of patients, these findings raise questions about whether a universal TMB cut-off is appropriate for all tumor types and why certain tumors with low TMB still benefit from ICI. Consequently, a more qualitative approach to the TMB could potentially overcome the limitations of an absolute numerical threshold for all. Indel mutations causing a frameshift and mutations affecting the RNA splicing can generate neojunctions and neoantigens that lead to enhanced immunogenicity.8 In a pan-cancer analysis from 5777 solid tumors obtained from The Cancer Genome Atlas (TCGA), renal carcinomas were found to have the highest proportion and number of indel mutations, and overall the number of frameshift neoantigens was higher in the tumor types for which ICI are approved compared with those that have not.9 In the same study, indels altering the reading frame were found to generate three times more predicted neoantigens as non-synonymous single-nucleotide variation mutations, supporting the hypothesis that frameshift neoantigens contribute to antitumor immunity.9

Using the TCGA dataset, we have also previously reported a positive correlation between frameshift burden and activated T-cell infiltration10 in non-small cell lung cancer (NSCLC). Among a group of 122 patients with metastatic NSCLC treated with ICI, our study revealed that patients whose tumors carried frameshift mutations experienced improved progression-free survival (PFS) (6.2 vs 2.7 months, p=0.01) as well as improved overall response rates (26% vs 12%, p=0.04) compared with patients whose tumors lacked frameshift mutations.10 These findings suggest that frameshift mutations may serve as a predictive marker for responses to ICI.

In this work, we aimed to explore the role of frameshift mutations in enhancing TMB as a predictive biomarker of ICI in solid tumors beyond lung cancer. We examined this in a retrospective cohort of patients with metastatic solid tumors treated with ICI monotherapy. We also leveraged a real-world (RW) genomic dataset to comprehensively characterize the landscape of frameshift alterations in a pan-cancer cohort.

Materials and methods

Clinical cohort

PFS was defined as the period from ICI therapy initiation date to the date of disease progression or death, and overall survival (OS) was defined as the period from the ICI initiation date to the date of death. We retrospectively analyzed the 12-week ORR, complete response, partial response, stable disease, progressive disease in patients with available imaging using the Response Evaluation Criteria in Solid Tumors version 1.1.

RW genomic cohort

The genomic cohort consisted of data from formalin-fixed paraffin-embedded tumor samples submitted to a commercial CLIA-certified laboratory for molecular profiling (Foundation Medicine) from a national cohort of patients who underwent testing as part of the course of routine clinical care. The cancer-type diagnosis was rendered by the ordering physician.

Comprehensive genomic profiling was performed on hybridization-captured, adaptor ligation-based libraries for at least 324 cancer-related genes plus select introns from at least 28 genes frequently rearranged in cancer, examining all classes of alterations (base substitutions, short insertions/deletions (indels), rearrangements, and copy-number alterations)11 12 (FoundationOne and FoundationOneCDx). A total of 250,813 patient samples with calculated MSI and TMB values across 20 tumor types were available for analysis. TMB was calculated using somatic non-driver alterations on at least 0.8 Mb of exonic sequence.13 Microsatellite status was determined based on sequencing of homopolymer repeat regions. Frameshift burden was calculated as a sum of all frameshift alterations (out of frame short indels) in each sample; analyses included both pathogenic variants and variants of uncertain significance.

Analysis of tumor genomic pathway alterations

We examined the percentage of altered genes involved in the ten oncogenic signaling pathways as analyzed by the TCGA Pan-Cancer Analysis.14 These oncogenic pathways are: RTK/RAS pathway, NRF2/KEAP1 pathway, PI3K pathway, WNT pathway, MYC pathway, P53 pathway, Cell Cycle pathway, HIPPO pathway, TGFβ pathway and NOTCH pathway. For each oncogenic signaling pathway, we computed the percentage of the samples altered in each histology. Similarly, we also investigated the ten DNA damage response (DDR) pathways as analyzed by the TCGA Pan-Cancer Analysis. These DDR pathways are: base excision repair, nucleotide excision repair, MMR, Fanconi anemia, homology-dependent recombination, non-homologous end joining, direct repair (DR), translesion synthesis (TLS), nucleotide pools and DDR others.15 For each DDR pathway, we computed the percentage of the samples altered in each histology.

Statistical methods

χ2/Fisher’s exact test (as appropriate) was used to determine associations between the categorical covariates of interest. Unadjusted Kaplan-Meier survival curves with log-rank testing were generated to compare the PFS and OS. HRs and CIs in univariate and multivariate analysis were derived using Cox proportional hazard model. A p<0.05 was considered significant. The analysis was conducted using IBM SPSS (V.22.0, SPSS). Plots were generated in R.16


Cohort summary

We identified 212 patients across 2 institutions who were diagnosed with metastatic solid tumors, received ICI as monotherapy, and whose tumor underwent genomic sequencing by commercially available vendors. The median age of patients was 63.5 years (range 31–87), and 60% were males (table 1). We identified 10 tumor types, and the most common were genitourinary cancers (n=51, 24%), followed by gastrointestinal malignancies (n=43, 20.2%) and melanoma (n=37, 17.4%) (see table 1 for distribution of tumor types). The median number of prior therapies was 1 (range 0–8), and most patients received ICI monotherapy (table 1). Of the 212 tumors, 196 were microsatellite stable (MSS) (92.5%) (table 1).

Table 1

Baseline demographics and clinical characteristics of the clinical cohort

We have previously described the clinical and immunological implications of frameshift mutations in patients with lung cancer,10 and so we excluded patients with this cancer histology from the present clinical cohort.

To determine how TMB and FS are distributed across a variety of cancers, we analyzed genomic data from all solid tumor-type samples (N=250, 813) from a RW genomic cohort. Most tumors were MSS (n=246,252, 98.1%) and TMB-L (TMB<10 mut/Mb) (n=209,567, 83.1%) (table 2). The tumor type distribution is described in online supplemental table 1. For the subsequent genomic analyses on the distribution of TMB and FS, we excluded tumors with MSI (MSI-H, N=4516). Twenty different solid tumors were included and the most common histology overall was NSCLC (n=53,990, 21.9%), followed by colorectal cancer (CRC; n=32,055, 13.0%) and breast cancer (n=28,078, 11.4%) (online supplemental table 1).

Supplemental material

Table 2

Distribution of tumors in the RW- genomic cohort based on SS and TMB

Clinical outcomes

We divided the 212 clinical patients into four groups based on tumor FS (FS-A: frameshift absent, FS-P: frameshift present) and the tumor TMB (TMB-H: TMB equal or more than 10 mutations/megabase, TMB-L: TMB less than 10 mut/Mb). Patients were grouped as following: TMB-L/FS-A (n=80, 38%), TMB-L/FS-P (n=57, 27%), TMB-H/FS-A (n=36, 17%), TMB-H/FS-P (n=39, 18%). The respective median PFS for these groups was 3.6 vs 5.1 vs 11.6 vs 12.0 months (p<0.01) (figure 1), and the median OS was 13.3 vs 11.9 vs 30.3 vs 24.5 months (p=0.057). The 12-month PFS probability was 34% for TMB-L/FS-P vs 17.1% for TMB-L/FS-A. The 12-month OS probability was 48% for TMB-L/FS-P vs 53% for TMB-L/FS-A.

Figure 1

Progression free survival (PFS) based on tumor TMB and frameshift indel (p<0.01). TMB, tumor mutational burden.

Overall, the median PFS was 10.6 and 3.6 months (HR 0.51 CI 0.36 to 0.73; p<0.001) for patients with TMB-H and TMB-L, respectively. The median OS for TMB-H and TMB-L was 30.2 vs 12.2 months (HR 0.54 CI 0.35 to 0.84; p<0.01), respectively.

The median OS for FS-A versus FS-P was 20.4 vs 14.4 months, respectively (HR 0.98, 95% CI 0.67 to 1.39; p>0.1), and the median PFS was 4.0 vs 7.9 months (HR 0.63, 95% CI 0.46 to 0.87, p=0.005) (figure 2), respectively. The ORR at 12 weeks was 23.8% in FS-P vs 12.8% in FS-A (p=0.02) (figure 3).

Figure 2

Tumors with FS are associated with better PFS in the clinical cohort (p<0.01). FS, frameshift mutation; PFS, progression-free survival.

Figure 3

Objective response rate at 12 weeks (p=0.02), CR, complete response; ORR, objective response rate; PD, progressive disease; PR, partial response; SD, stable disease.

Among all patients, 30% (N=65) received ICI as first-line therapy and had improved clinical outcomes compared with those who received ICI beyond first line; median PFS was 9.21 vs 3.45 months (p=0.0003), and median OS 14.01 vs 8.28 months (p=0.0001), respectively. The ORR was numerically higher in the patients receiving first line ICI 24.6% vs 15.0% in the rest, although not statistically significant (p=0.135).

Of note, the distribution of patient baseline characteristics in regard to race, age, number of prior treatments, and ECOG was not statistically different between each of the four genomic groups, with the exception of microsatellite status (p<0.01). All patients (n=11) with MSI tumors belonged in the TMB-H/FS-P group.

The landscape of predictive biomarkers for ICI across tumor histologies within an RW-genomic cohort

In contrast to MSI, which was found only in 1.82% of the tumors, TMB-H was nine times more frequent (16.4%, n=41 246) (table 2). Almost all MSI-H tumors were also TMB-H (online supplemental figure 1). The tumor type with the highest incidence of MSI-H was endometrial carcinoma (16.28%) (online supplemental figure 2), (online supplemental table 2), followed by gastric cancer (4.69%), CRC (4.69%), and prostate cancer (2.6%). The remaining tumor types had MSI-H tumors in less than 2% of the cases. Only 4 out of 8044 melanoma samples were MSI-H (online supplemental table 2).

Among the MSS tumors, the median TMB was lower than 10 mut/Mb in all tumor types except melanoma (23.7 mut/Mb). The tumor types with the next highest median TMB were bladder cancer (9.8 mut/Mb), NSCLC (9.7 mut/Mb), small cell lung cancer (9.2 mut/Mb), and head and neck cancers (8.8 mut/Mb) (online supplemental figure 3A). Pancreatic and thyroid cancer had the lowest median TMB in the cohort (2.6 and 2.4 mut/Mb, respectively).

For the MSI-H tumors, glioblastoma had the highest median TMB (67.5 mut/Mb), followed by CRC (53.6 mut/Mb), head and neck cancers (49.4 mut/Mb), melanoma (45.3 mut/Mb), kidney cancer (45.4 mut/Mb) (online supplemental figure 3B). On the other hand, MSI-H thyroid cancer had the lowest median TMB (16.8 mut/Mb) (online supplemental figure 3B).

The burden of frameshift mutations (FIB) was calculated for each histology according to the MSI-H status (online supplemental figure 4). For the MSS tumors, the histology with the highest mean FIB was kidney cancer, followed by bladder and CRC (online supplemental figure 4A). The MSS histologies with the lowest FIB were melanoma, cervical cancer, and soft tissue sarcoma. Among MSI tumors, colorectal, kidney, and gastric cancer had the highest FIB, with soft tissue sarcoma, thyroid, and melanoma having the lowest. The mean FIB for the entire MSS cohort was 5.91 vs 9.93 in the MSI-H cohort (online supplemental figure 4B).

Joint distribution of TMB and FS indels across all tumors in the RW-genomic cohort

Similar to the clinical cohort, we examined the distribution of TMB and FS indels (figure 4, online supplemental table 3) in the RW-genomic cohort. Because the MSI-H tumors represented less than 2% of the cohort, we limited this analysis to MSS tumors. The samples were divided into groups based on the TMB and presence or absence of FS indels: TMB-L/FS-A, TMB-L/FS-P, TMB-H/FS-A, and TMB-H/FS-P. The tumors with the highest proportion in the TMB-L/FS-P group were CRC (60.97%), kidney cancer (59.56%), breast cancer (48.22%), and endometrial cancer (45.44%). The tumors with the lowest representation in the TMB-L/FS-P group were melanoma (12.91%), cervical cancer (23.85%), thyroid cancer (27%), and soft tissue sarcoma (27.89%).

Figure 4

Fsindel landscape across all tumor histologies. BIL, biliary cancer; BRCA, breast cancer; BLCA, bladder cancer; CESC, cervical cancer; CUP, cancer of unknown primary; CRC, colorectal cancer; ESCA, esophageal cancer; GLM, glioblastoma; HNSC, head and neck cancers; KICA, kidney cancer; MEL, melanoma; NSCLC, non-small cell lung cancer; OVCA, ovarian cancer; PRAD, prostate cancer; PANC, pancreatic adenocarcinoma; STS_LMS, Soft Tissue Sarcoma; SCL, small cell lung cancer; THCA, thyroid cancer; UCEC, endometrial cancer.

Distribution of FS mutations in genes involved in oncogenic and DNA damage repair pathways in MSS tumors of the RW-genomic cohort

The p53 pathway had the highest frequency of frameshift mutations (10%, 24,015 of 246,252 samples) (figure 5A). The tumor subtypes with the highest fraction of frameshift mutations in this pathway were (in descending order): ovarian cancer (13%), small cell lung cancer (13%), breast cancer (12%), pancreatic cancer (12%), esophageal cancer (12%), NSCLC (10%), CRC (10%), head and neck cancers (10%), prostate cancer (9%), endometrial cancer (9%), gastric cancer (9%), biliary cancer (9%), cancer of unknown primary (9%), bladder cancer (8%), soft tissue sarcoma (7%), thyroid cancer (6%), glioblastoma (5%), kidney cancer (4%), melanoma (3%), cervical cancer (2%). The second most frequently involved pathway was the PI3K, with endometrial cancer having the highest frameshift frequency (18%). In the cell cycle pathway, small cell lung and bladder cancers were the tumor subtypes with the most frequent frameshift mutations (22% and 15%, respectively). The highest frequency of frameshift mutations for all pathways was in the CRC (44%) WNT pathway.

Figure 5

(A) Percentage of FS in MSS tumors based on pathways. (B) Percentage of all mutations in MSS tumors based on Pathways. BER, base excision repair; FA, fanconi anemia; HR, homologous repair, MMR, mismatch repair; NER, nucleotide excision repair; NHEJ, non-homologous End Joining; others, translesion synthesis (REV3L), damage sensor (ATM, ATR, CHEK2), direct repair (ALKBH3, MGMT).

Among the DNA damage repair (DDR) pathways, the cumulative group of three pathways (damage sensor, DR, and TLS) had the highest frequency of frameshift mutations. The tumor subtypes with the highest frequency of frameshift in this pathway were (in descending order): glioblastoma (22%), endometrial cancer (22%), small cell lung cancer (17%), breast cancer (16%), pancreatic cancer (15%), ovarian cancer (15%), NSCLC (14%), prostate cancer (14%), esophageal cancer (14%), soft tissue sarcoma (13%), CRC (13%), cancer of unknown primary (13%), gastric cancer (12%), head and neck cancers (11%), biliary cancers (11%), bladder cancer (10%), thyroid cancer (9%), kidney cancer (8%), melanoma (6%) and cervical cancer (5%).

For all mutations, the most frequently involved pathways were (figure 5B): RTK-RAS (77%, 189,110 of 246,252 samples), p53 (70%, 173,599 of 246,252 samples), Cell cycle (60%, 147,575 of 246,252 samples), PI3K (42%, 103,439 of 246,252 samples), and the cumulative DDR pathway. The tumor types with the highest frequency of alterations in the RTK-RAS pathway were (in descending order): melanoma (100%), pancreatic cancer (100%), CRC (96%), NSCLC (93%), glioblastoma (87%), thyroid cancer (86%), bladder cancer (81%), esophageal cancer (79%), cancer of unknown primary (71%), biliary cancers (68%), gastric cancer (65%), breast cancer (65%), endometrial cancer (62%), ovarian cancer (50%), cervical cancer (44%), head and neck cancers (42%), soft tissue sarcoma (39%), small cell lung cancer (32%), kidney cancer (25%), prostate cancer (20%).


In this study, we report the landscape of frameshift mutations and the interplay with TMB in a pan-cancer cohort of 250,813 samples. We also analyzed the ICI responses and outcomes in 212 patients within a solid tumor cohort. We correlated these outcomes with the presence of frameshift mutations and TMB. We found that the presence of frameshift mutations was associated with significantly higher ORRs than tumors without frameshift mutations (23.8% vs 12.8%, respectively). In addition, for tumors with low TMB, we showed that the median PFS was improved in patients with tumors harboring frameshift mutations compared with those that did not. While we did not find any difference in median OS, we hypothesize that this could be related to the mixed tumor subtypes represented in our clinical cohort and the vastly different tumor biology of each disease, and possibly the impact of treatment beyond progression to ICI.

The highest rate of frameshift mutations in our MSS genomic tumor cohort was noted in kidney, bladder, and CRCs. In an analysis that included 5777 tumors from TCGA reported by Turajlic et al,9 renal cell carcinomas were also found to have the highest proportion and number of indel mutations. The finding of CRC as the tumor with the third highest incidence of frameshift mutations is important. In a recent neoadjuvant trial of dual ICI with ipilimumab and nivolumab in early stage colon cancer, 27% of patients with MSS tumors showed pathological responses,17 suggesting ICI efficacy in certain MSS CRCs. Similarly, activity was also shown in heavily pretreated patients with metastatic MSS CRC, in a phase 1a/1b study of botensilimab, an anti-CTLA4 antibody, in combination with balstilimab, an anti-PD1 antibody.18 The ORR was 22% (95% CI 12% to 35%) with a disease control rate of 73% (95% CI 60% to 84%) and a 12-month OS rate of 61% (95% CI 42% to 75%).18

Within the low TMB group, CRC also had the highest incidence of frameshift mutations (63.5%). Interestingly, melanoma had the highest median TMB but had one of the lowest rates of frameshift mutations in both the MSS and MSI cohorts, possibly because of the UV-mediated mutagenesis in melanoma which predominantly results in missense mutations rather than frameshifts.

Among the tumors from the TCGA dataset,14 the RTK-RAS pathway was the signaling pathway with the highest frequency of all genomic alterations. Similarly, this pathway also had the highest frequency of alterations across all tumor types in our RW-genomic cohort. However, the RTK-RAS was not frequently affected by frameshift mutations, even in tumor subtypes where RAS is an oncogenic driver, such as pancreatic and CRC. In contrast, frameshift mutations were more commonly noted within the p53 pathway, which involves the tumor suppressors TP53 and CDKN2A and the MDM2/4 oncogenes, possibly related to its role in genomic integrity. Interestingly, there were no frameshift mutations affecting Myc pathway across all histologies.

Our study is limited by the retrospective nature of the clinical data presented and the inclusion of several primary sites/histologies in the analysis. The available clinical information was from a limited number of patients which were selected based on the availability of clinical and genomic data. The unknown impact on survival of subsequent treatments in the clinical cohort is also a limitation. Further, the treatment status of the patients within the RW-genomic cohort was unknown but likely reflects a mix of treated and untreated patients. Although not well established yet, chemotherapy may affect the tumor mutational status, which could confound our results. In addition, we investigated the incidence of frameshift mutations in a selected set of genes, although a test commonly used in clinical practice. The genomic panel used in this study covered 0.8–1.1 Mb of the exome to calculate TMB. Lastly, for the tumor genomic pathway alteration analysis, we used summative genomic data, which may overestimate the frequency of each mutation.

More biomarkers are urgently needed to identify patients who may benefit from immunotherapy. TMB is an imperfect marker, and it is known that patients with certain tumors with low TMB still derive benefits from ICI. While the etiology is unclear, this study aimed to look at TMB as a predictive biomarker with a more qualitative approach by investigating the role of specific genomic alterations. Whereas certain mutations, such as frameshift, are more immunogenic than others, all are weighted equally in the TMB calculation, and our data suggest that this distinction may be important in predicting responses to ICI. The incidence of frameshift mutations across different solid cancers with low TMB and MSS status varies from 12% to 60%. If the benefit of ICI is confirmed prospectively in tumors with low TMB and frameshift mutations, it can potentially expand ICI indication for a substantial number of patients who may otherwise have limited treatment options.

Data availability statement

Data are available on reasonable request. Data are available on reasonable request from the corresponding author.

Ethics statements

Patient consent for publication

Ethics approval

This was a retrospective cohort of patients treated at Huntsman Cancer Institute/ University of Utah and University of Miami Sylvester Comprehensive Cancer Center. Both institutional databases were queried with Institutional Review Board (University of Utah, IRB_00127278) approval to identify patients with metastatic solid tumors who received ICI monotherapy between 2015 and 2020 and whose tumors underwent TMB testing. In addition, billing codes and tumor registry queries were used to identify patients. Due to the retrospective nature of the study, informed consent of the patients was not required.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter @VaiaFlorou, @chfloudas, @@asaf_maoz, @thenasheffect, @sonampuri, @CentralParkWMD

  • Contributors Acquisition of data: VF, WP, ESS and GF. Analysis and interpretation of data: VF, CSF, AM, ARN, CN, ACT and IG-L. Writing of the manuscript: VF, CSF, AM and IG-L. Review and revision of the manuscript: VF, CSF, AM, ARN, CN, ACT, ESS, GF, HPS, SP, US, BW, PH, JT, GdLL, WP and IG-L. Conception and design: VF, CSF, WP and IG-L. Development of methodology: VF, CSF and IG-L. VF is responsible for the overall content as guarantor.

  • Funding Supported by National Institutes of Health Cancer Center Support Grant No. P30CA042014 and the Huntsman Cancer Institute (01-01903-2000-28497).

  • Competing interests VF consultant or advisory role Deciphera, Incyte. AM has patent applications related to the use of cytokines to enhance macrophage immunotherapy. ESS and GF are employees of Foundation Medicine and Shareholders in Roche. HPS consultant or advisory role Ipsen, AstraZeneca, ITM, Novartis, TerSera. SP consultant or advisory role IntegrityCE, Dava Oncology. US consultant or advisory role Astellas, Exelixis, Seattle Genetics, Imvax, Sanofi, and AstraZeneca. WP consultant or advisory role Astellas. IG-L consultant or advisory role SOTIO, Kanaph, Jazz, OncXer.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.