Chimeric antigen receptor (CAR) T cell therapy targeting B-cell antigens has led to unprecedented success in inducing complete response (CR) rates in refractory B-cell non-Hodgkin lymphomas (NHLs). We previously reported the results of a pilot trial testing a 3rd generation CAR with 4-1BB and CD28 costimulatory domains targeting CD20 in patients (pts) with relapsed B-cell lymphomas (NCT00621452). 3 pts received cyclophosphamide (CY) 1 g/m2 IV for lymphodepletion 2 days prior to a series of 3 CAR T cell infusions 2 to 5 days apart in escalating doses followed by 2 weeks of low-dose IL-2. Here, we report the long term follow up of 2 of these pts. The third pt, UPN-02, died of progressive mantle cell lymphoma (MCL) and myelodysplastic syndrome at ~26 months (mos) post-treatment. Pt UPN-03, with relapsed MCL, had no evaluable disease after an excisional biopsy following protocol therapy. He relapsed after 2 years with a left flank mass, which was then radiated. He was re-treated with CY lymphodepletion, a dose of 1.8 x 109 CAR T cells/m2, and low-dose IL-2. At 6 weeks post infusion, he developed a small mass at the margin of his previous radiation field suspected to be recurrent MCL. This mass spontaneously regressed 1 mo later. Subsequent CT scans showed a residual soft tissue mass that gradually decreased over time and met criteria for CR by CT criteria. He remained in remission, albeit with detectable minimal residual disease (MRD) by ClonoSEQ assay, for the next 8.4 years until developing a subcutaneous chest mass and axillary adenopathy with biopsy showing recurrent MCL. Pt UPN-04, with relapsed follicular lymphoma (FL), had stable disease by CT at 1 mo but went on to have a partial response at ~3 mos. By 12 mos post-CAR T treatment there was evidence of progressive disease. However, between 2.5 and 3 years post-treatment he experienced resolution of all palpable nodes, and a near-CR was confirmed by CT at 35 mos after treatment. He remained without any clinical signs of progression with low level MRD by ClonoSEQ for the next 4 years. At 7.2 years post-CAR T cell treatment, he was found to have recurrent FL. CAR T cells were detectable for up to 12 mos in UPN-03 and 9 mos in UPN-04 by qPCR assay. B-cell recovery over time in both pts confirmed a lack of functional CAR T cell persistence. Given these findings, along with the durable responses observed in both pts, we hypothesized that these late remissions were caused by endogenous immune responses triggered by CAR T cell therapy. To evaluate this, we performed western blot assays incubating post-infusion serum with autologous tumor cell lysates followed by incubation with an anti-human IgG antibody, which showed new bands at post clinical response timepoints in both pts. ELISPOT assays with PBMCs co-incubated with autologous tumor cells exhibited spikes in IFN-γ secreting cells compared with control cells at timepoints proximal to their clinical responses. These findings suggested the occurrence of both humoral and cellular anti-tumor responses. To better characterize the T-cell response, we performed high-throughput TCR-β sequencing (Adaptive ImmunoSEQ) for UPN-04. We conducted a differential abundance analysis to identify possible tumor-reactive T cells, with the hypothesis that tumor-reactive T cell clones would be increased near the timepoints with clinical responses at ~3 and 35 mos. We identified 12 and 7 sequences, respectively, at these timepoints that were significantly increased compared to baseline (FDR < 0.05) and not present in the CAR-T product; 3 sequences were present at both 3 and 35 mos, and 1 of these was also significantly enriched in a post-infusion tumor biopsy, suggesting possible tumor-specificity. These results support the hypothesis that the late responses were mediated by endogenous T cells rather than CAR T cells. In summary, we report long-term remissions in 2 of 3 pts treated with CD20-targeted CAR T cell therapy, with evidence of endogenous anti-tumor immune responses. Together with previously published reports of epitope spreading after CAR-T for solid tumors, and in mouse models of lymphoma, these results suggest that CAR T cells have the capacity to elicit endogenous anti-tumor immune responses in lymphoma. Our use of CY without fludarabine as lymphodepletion may have facilitated such responses. Future research is warranted to enhance epitope spreading after CAR-T therapy as a bulwark against escape by antigen loss or loss of CAR-T persistence.