Analysis of TCR CDR3 sequencing data
Sequencing data were analyzed from blood samples of DLBCL patients and healthy controls. A total of 873937 unique TCR CDR3 amino acid sequences were obtained, with an average of 14565.6 per sample. The number of unique CDR3 in the patient group was lower than that in the control group (P < 0.0001), and the median value was 11412 (9071.25 ~ 16014.75) and 16691 (15159.5 ~ 18890), respectively. The V gene of the patient group (50, range from 50 to 52) was lower than that of the control group (52, range from 51 to 54), which was statistically significant (P = 0.0006) (Table 2). In addition, the median number of 502 V-J pairs (range from 479.5 to 516.75) identified from the DLBCL patient group was lower than that of healthy controls (537, range from 518.75 to 560.75) (P < 0.0001), indicating antigenic stimulation and certain specific V-J combination amplification.
Table 2
Sequencing data of TCR CDR3 repertoires for DLBCL patients and controls
|
Patient
|
Normal
|
P
|
Unique CDR3
|
11412(9071.25 ~ 16014.75)
|
16691(15159.5 ~ 18890)
|
< 0.0001
|
Unique V family
|
50(50 ~ 52)
|
52(51 ~ 54)
|
0.0006
|
Unique J family
|
13(13 ~ 14)
|
13(13 ~ 14)
|
0.188
|
V-J pair
|
502(479.5 ~ 516.75)
|
537(518.75 ~ 560.75)
|
< 0.0001
|
All 57 V genes and 14 J genes were observed, and the frequency of these gene usages was analyzed in each sample. The 27 V genes and 3 J genes between DLBCL patients and controls were significantly differences by Mann-Whitney U test. TRBV28, TRBV18, TRBV12-3, TRBV27, TRBV14, TRBV7-7, and TRBV12-4 were more common in patients, while others were used more frequently in the control group (Table S2). Among all J genes, only TRBJ2-3, TRBJ1-4, and TRBJ1-2 showed different frequency of use between the two groups, and the control group was higher (Table S3).
Comparison of the diversity of TCR repertoires between DLBCL patients and controls
We used Shannon index, unique clonotypes and D50 to quantify the diversity of the TCR repertoires. Compared with the control group, the D50 and the number of clonotypes in DLBCL patients (Fig. 1B, 1C) were significantly lower (P < 0.0001; P < 0.0001). However, no difference was observed in the three indicators before and after treatment.
The relationship between TCR CDR3 diversity and clinical characteristics
We assessed the relationship between TCR CDR3 diversity and clinical characteristics in DLBCL patients. Patients with stage III-IV had less TCR CDR3 diversity than patients with stage I-II (P = 0.003). Compared with younger patients, patients with a median age of 53.5 years or older had lower D50 values (P = 0.023). Regarding the IPI score and ECOG performance, the D50 value of patients with a score greater than 1 decreased (P = 0.036) (Fig. 2). In order to study how the diversity of peripheral blood TCR reflects the immune status of DLBCL, we evaluated the relationship between lactate dehydrogenase (LDH) and β-2 microglobulin and CDR3 diversity, and found their negative correlation (Figure S1).
The relationship between TCR diversity and clinical response
According to the efficacy evaluation after four cycles of treatment, DLBCL patients were divided into CR group and Non-CR group. The TCR diversity of non-CR group before treatment was lower than that of CR group (Figure S2). Since TCR V and J genes show different usages between individuals, we continue to investigate whether the use of TRBV and TRBJ genes has potential therapeutic efficacy predictors in patients. We compared the use of V and J genes in the blood before treatment between the two groups, and found that in the CR group with good clinical response, TRBV5-1, TRBV6-1, TRBV6-7, TRBV6-9, TRBV7-4, TRBV7- 7. The frequency of TRBV11-1 and TRBJ2-5 is significantly higher. Then we established a Logistic regression model to predict the efficacy. The model was evaluated by the receiver operating characteristic curve (ROC). As a result, the combined prediction model including TRBV7-7 and TRBV11-1 classified the efficacy prediction (AUC = 0.891) (Fig. 3).
The high TCR similarity index is related to the poor efficacy of DLBCL patients
In order to quantitatively evaluate whether the change in the distribution of clones is treatment-induced TCR repertoires reconstruction or treatment-induced new clones generation, we applied the Bhattacharyya coefficient to calculate the similarity index between the pre- and post-treatment samples. Our study showed that after effective treatment in DLBCL patients, there was a significant correlation between the similarity index and the response to treatment, and the similarity index of the CR group decreased(Figure 4A, 4B, P = 0.048, AUC = 0.712).
To further describe the frequency of all TCR clones pre- and post- treatment, we use pre-treatment/post-treatment clones as the X/Y axis. TCR clones with the same frequency are on the Y = X diagonal, and the identified new clones or changed clones are located above or below the diagonal. The frequency of T cell clones increased or decreased significantly in each treated patient, and more significant changes were observed in CR patients (Fig. 6D). Next, we evaluated the increase and decrease of TCR clones in patients with DLBCL pre- and post- treatment. In 30 patients, the median percentages of expanded and contracted TCR clones were 12.18% and 14.73% (Fig. 6C), indicating that the percentages of increased and decreased TCR clones after treatment were similar.