The landscape of TIICs in osteosarcoma
CIBERSORT algorithm was used to screen out samples with CIBERSORT output P value less than 0.05 in GSE16091, GSE21257 and GSE39055 data sets for research. Ultimately, 95 patients with osteosarcoma were screened out (tabel1). Percentage bar chart was drawn to show the proportion of 22 TIICs in each sample (Figure1A). The results revealed that the proportion of TIICs in osteosarcoma were B cells naïve (5.14%), B cells memory (0.88%), Plasma cells(1.77%), T cellsCD8(6.43%), T cells CD4 naive(0.90%), T cells CD4 memory resting (0.74%), T cells CD4 memory activated (1.02%), T cells follicular helper (4.79%), Tregs (2.40%), T cells gamma delta (6.02%), NK cells resting (0.38%), NK cells activated (2.26%), Monocytes (1.53%), Macrophages M0 (40.54%), Macrophages M1 (3.66%), Macrophages M2(13.34%), Dendritic cells resting (1.40%), Dendritic cells activated (0.79%), Mast cells resting (2.88%), Mast cells activated (2.43%), Eosinophils (0.05%), Neutrophils (0.62%), respectively. Figure1B indicated the correlation coefficient between 22 TIICs. Red and blue colors indicate positive and negative correlation, respectively. Color intensity corresponds to the degree of correlation. We can found that eosinophils and memory B cells have the strongest positive correlation (r=0.61), CD8+T cells had the strongest negative correlation with M0 Macrophages and gdT cells (r=-0.54).
Predictive value of TIICs in osteosarcoma
To investigate the prognostic value of 22 TIICs, Kaplan-Meier curves were drawn and the results were evaluated by log-rank test. In Figure2, we can find that high abundance of naive B cells (p=0.047) and Monocytes (p=0.03) in osteosarcoma is associated with poor prognosis, while other TIICs were not significantly associated with patients’ outcomes (p>0.05).
Establishment and confirmation of an immune risk score model
Considering that the effect of TIICs on the prognosis of patients is not independent, we constructed an immune risk score model to evaluate the prognosis of patients individually. Firstly, we used the forward stepwise regression method to screen out a minimal set of TIICs. Finally, eight immune cells including naive B cells, activated memory CD4+T cells, follicular helper T cells, gd T cells, resting Dendritic cells, activated Dendritic cells, Neutrophils and activated Mast cells were selected (tabel2). Then, eight immune cells were filled into multivariate Cox PHR analysis to construct an immune risk score model. Formula is as follows: Risk8 = 8.49 * naive B cells +15.7 * activated memory CD4+T cells + 17.1 * follicular helper T cells + 6.07 * gd T cells + 27.9 * resting Dendritic cells + 24.7 * activated Dendritic cells - 26.8 * Neutrophils -7.38 * activated Mast cells. Each sample will be given a risk score based on the model. Patients were divided into high-risk group and low-risk group according to the median risk score. Kaplan-Meier curves indicated that patients in high-risk group had a poorer prognosis than those in low-risk group (p= 8.303e-04, figure3A). The ROC curve showed that the immune risk score model is reliable in predicting the prognosis of patients with osteosarcoma (AUC=0.724, figure3B). In addition, figure 3C, 3D and 3E respectively showed the risk score, survival status and 8 immune cell infiltration of patients with osteosarcoma. Then, the prediction ability of the immune risk score model was verified in Target-OS dataset based on the same cutoff value as GEO dataset. The result indicated that is the prediction ability of the immune risk score model is reliable (figure 4).
Independent predictive power of eight immune cells
We also used multivariate analysis to explore whether these 8 immune cells were independent prognostic factors for osteosarcoma patients. The results indicated that naive B cells (HR = 4.9e+03; 95% CI: 5.5-4.3e+06; p = 0.014), activated memory CD4+T cells (HR = 6.6e+06; 95% CI: 2.6e+02-1.7e+11; p = 0.002), follicular helper T cells (HR = 2.6e+07; 95% CI: 1.1e+03-6.0e+11; p <0.001), resting Dendritic cells (HR = 1.3e+12; 95% CI: 5.0e+04-3.3e+19; p = 0.001), activated Dendritic cells (HR =5.3e+10; 95% CI: 4.8e+02-5.9e+18; p = 0.009) were independent predictors for osteosarcoma patients (figure5, tabel2).
Construction of nomogram model
To assess 3-year survival in patients with osteosarcoma, we constructed a nomogram based on eight screened immune cells. In the nomogram model, there are several vertical lines passing through the value of each variable, and different integrals can be obtained on the integral line at the top of the nomogram. The integrals of all the variables are added to obtain the total score, and the predicted survival probability values at the corresponding time points are calculated from the line perpendicular to the total score on the prediction line at the bottom of the nomogram. Probability of death = 1- probability of survival. According to the nomogram, we can score the influence of eight immune cells on the prognosis of patients, and then use the total score to evaluate the 3-year survival rate of patients with osteosarcoma (figure6).