Table 1.
Potential biomarkers and targets identified by the model. As predicted by the proposed model the IL-8 and Lactate can serve as prominent distinguishing factor between the resistant and non-resistant subtypes. However, OPN, LIF, and IL-2 can be appropriate target nodes for ICI resistance (non-response) emergent in the fibro-dominated and fibrotic only scenarios respectively.
Fig 1.
Workflow: We start with construction of the essential cellular molecular network of the HNSCC TME from the existing literature.
For this purpose, we begin with the relevant cell types in a HNSCC TME obtained from the literature. Further, to incorporate a wide range of possible interactions from a given cell type, we considered different functional cell states within each cell type and the relevant molecular species mediating the cell-cell interactions. Using the well-established modeling rules for different types of interactions, we simulate the TME system with different combinations of selected parameters. The steady state populations can be grouped into five distinct compositional regions (in CAF-Killer T-tumor cell population space) resembling the clinically observed TME subtypes. We chose the relevant parameter balances within each compositional regions to unpack the different mechanisms that explains the emergence and diversity of a given TME compositional group. Further, we simulate the model with anti-PD1 for growth-proliferation parameters pertaining to each of the five compositional groups and important parameter balances constructed from the remaining parameter sets. Finally, we perform cytokine profiling and computational knockout studies to identify potential biomarkers and targets.
Fig 2.
Explaining the phenotypic heterogeneity of the HNSCC TME.
(a) The nodes are either the cell states or the molecular species, whereas the edges represent diverse forms of interactions. The acronyms C_0, C_PDL1 + , and C_PDL1- refer to stem, PDL1+ (programed death ligand1), and PDL1- tumor cells, respectively. T_K + , T_K-, T_Help, T_Reg, and T_Ex stands for PD1+ (programmed death 1), PD1- killer T cells, Helper T cells, Regulatory T cells, and Exhausted T cells, respectively. M_1 and M_2 refer to macrophages of M1 and M2 phase, respectively. Further, F_WT and CAF correspond to wild-type and invasive cancer-associated fibroblasts, respectively. The acronyms IL-2, IL-8, IL-10 LIF, IFNG, IRF8, OPN, ICAM1, and Lac denote Interleukin-2, Interleukin-8, Interleukin-10, Leukemia Inhibitory Factor, Interferon Gamma, Interferon Regulatory Factor-8, Osteopontin, Intercellular Adhesion Molecule-1, and Lactate, respectively. Based on the accessibility of the tumor cells to the killer T cells, each of the three tumor cell states C_0, C + , and C- are further divided in to two categories Killer T-exposed and killer T non-exposed. S5 Text details all the abbreviations used in the network model. The killer T non-exposed tumor cell states are protected by the resident CAF population. (Refer to S1 Fig for detailed representation of the model) (b) We simulated the dynamics for 10,000 combinations of the parameters (that have a direct influence on the proliferation, conversion, and death of the CAF, Killer T, and tumor cells) based on the reconstructed network system to obtain the possible steady-state groups. We found five distinct groups: desert, immune desert, fibro desert, immune-dominated, and fibro-dominated. (c) The associated mapping of these five groups in the hyper-parameter space. (d-h) Bar charts for the median population of tumor cells, killer T cells, and CAF population across different groups.
Fig 3.
(a-b) Time profiles for killer and exhausted T cells for different T cell exhaustion rates. A high T-cell exhaustion program that can drive an immune system with a high proliferation rate to the immune-desert phenotype. (c) Phase space for killer T cells versus the regulatory T cells for different regulatory T cell promoting roles of CAF. Demonstrates the possibility of an immune-cold scenario with abundant regulatory T cells due to high CAF-Treg interaction. (d)-(e) Post-ICI trajectories of killer T cells, tumor cells, and exhausted T cells for different balances between the anti-PD1 binding rate and T cell exhaustion rate. Unlike other low-proliferation induced immune-desert phenotypes (represented in dotted lines), the ICI intervention can improve the exhaustion-driven immune desert scenario. (f) Lactate reduction can further reduce the total post-ICI tumor cell count.
Table 2.
Comparing the model conclusions with the appropriate existing literature.
Fig 4.
Immune/non-fibrotic, favorable case:
(a-b) Shows despite the absence of CAF, the pre-ICI PDL1- tumor cells critically depend on the balance between the oncogenic (via Texh) and cytotoxic activities of T cells. (c) The intervention of ICI drives toward an aggressive cytotoxic T-cell-driven immune response. (d) Indicates the pivotal role of residual fixed resource supply rate (nutrients, blood flow, oxygen supply) in governing the possibility of recurrence despite a favorable prognosis.
Fig 5.
Immune-dominated-- High immune accessibility, moderate CAF: For a given CAF-tumor interaction rate, lower T-cell cytotoxicity leads to higher PDL1- to PDL1+ tumor cell population.
(a) For high levels of cytotoxicity, the tumor cell population remains insensitive to the proliferation rate () of the killer T cells in the immune dominated scenario. (b) CAF plays a dual role in governing the helper T cell population. CAF reduces the helper T cells via the regulatory T cells. However, a high tumor-promoting role of CAF increases the resident PDL1- -tumor cells that, in turn, increases the helper T cells via antigen sensing mechanism. (c) A high exhaustion rate can potentially increase the share of the PDL1- tumor cell population owing to the abundance of the exhausted tumor cells and their tumor-promoting effects. (d) The post-ICI scenario significantly reduces the tumor cells and CAF. The application of anti-PD1 reduces the resident CAF population and drastically reduces the tumor cell population. The population refer to the median population in the immune-dominated group (S6 Fig). Furthe the line bars indicate the 25% and 75% quartiles of the population within the immune-dominated scenario. (e) The post-ICI tumor cell population is proportional to the tumor-promoting role (via paracrine interaction) of the remaining CAF population.
Fig 6.
Fibro dominated, a modified immune desert: (a) depicts the conceptual framework of the proposed hypothesis on the CAF-pockets protected from the immune response.
(b) demonstrate the fact that the immune accessibility index governs the post-ICI proportion of immune-accessible and immune-inaccessible tumor cells. (c) Phase-space for LIF and total tumor cell population. The change in LIF levels due to the ICI therapy reduces with respect to decreasing accessibility. (d) Contrary to the immune-dominated TME, the post-ICI CAF population remains almost constant at low immune accessibility due to a lack of change in the post-ICI LIF levels compared to its pre-ICI counterpart.
Fig 7.
One-time IL-2 spikes— reprogramming immune-desert:
(a) A spike in the initial IL-2 levels (beyond a threshold) can drive the system trajectories towards an immune-non-desert steady state. (b) Although some trajectories return to the immune-desert arrangement, the trajectory settles in a non-zero killer T cell population beyond a threshold IL-2 injection. (c) Time profiles for killer T cells for different external IL-2 levels.
Fig 8.
OPN and LIF knockout, from fibro-dominated to fibro-desert:
(a-b) Demonstrate the effect of OPN reduction on the proportion of the inaccessible to accessible tumor cells in two scenarios: with and without LIF. As shown in both cases, below a certain threshold of OPN concentration, the balance between the accessible and inaccessible tumor cells improves significantly. An additional knockout of LIF extends the OPN-knockout-driven reduction in tumor cells towards complete removal. (c) Suggests that OPN reduction modulates the immune accessibility index of the TME. Further, a LIF + OPN reduction drives the TME towards full accessibility.
Fig 9.
IL-8: Potential biomarker for identifying TME subtype:
(a-b) Show that with high immune accessibility, the IL-8 level is low and is further reduced following a neo-adjuvant ICI therapy, whereas the scenario with low immune accessibility (characteristic of a fibro-dominated TME) can be identified with a slight increase in the post-ICI IL-8 level. Although the IL-8 trend is similar for immune-dominated and immune/non-fibrotic TME, the absolute levels differ significantly between the two subtypes. (c) The removal of both OPN and LIF reprograms the TME towards a immune/non-fibrotic subtype. Therefore, the IL-8 level post-OPN and LIF knockout followed by ICI also drops drastically compared to its pre-ICI and pre-removal counterparts.
Fig 10.
Lactate removal improves overall immune response:
(a) Lower levels of residual lactate concentration inside the TME enhance the cytotoxicity of the killer T cells in the pre-ICI setting. (b) The negative impact of lactate on the immune response continues to the post-ICI scenario, wherein a high lactate level impedes the PD1-killer T cell activity, leading to a high tumor cell population. (c) For high-killer T cell cytotoxicity, lactate can also serve as a distinguishing factor between immune-accessible and inaccessible HNSCC TMEs. In highly immune-accessible TMEs, the application of anti-PD1 renders a significant decrease in lactate compared to its pre-ICI counterpart. In contrast, the low immune-accessible TME correlates with higher post-ICI lactate levels.