Four patients (three patients with ET and one patient with post-ET myelofibrosis) that completed our CALRmut vaccination trial and displayed an immune response to the CALRLong36 epitope were chosen as donors for generation of specific T-cell cultures. Cultures were generated as previously described by stimulation with the CALRLong36 peptide (13). Specific T cells were enriched by overnight incubation after restimulation with CALRLong36 peptide and labelled with fluorescence labelled antibodies after 18 hours of stimulation. Specific T cells were sorted using a fluorescence activated cell sorting (FACS) Melody cell sorter (BD Biosciences, San José, CA, USA) with the sorting gate set on live, CD3⁺, CD4⁺ or CD8⁺, CD137⁺/CD107a⁺ double positive events ( Supplementary Material 1 , top). Enrichment purity was assessed, and pellets of specific cells were kept at -80 °C. Sorting of bone marrow derived T cells was performed on cryopreserved bone marrow derived mononuclear cells. T cells were enriched as live CD3⁺ positive events ( Supplementary Material 2 , bottom). Enrichment purity was assessed after FACS, and enriched T cells were pelleted and cryopreserved as described above. A visual abstract of the methods described above is provided in Figure 1 . A list of fluorescence labelled antibodies is provided in Supplementary Material 2 .

Autologous peripheral blood lymphocytes (PBL) and monocytes were enriched using a BD FACS Melody live cell sorter with monocytes sorted as live CD14⁺ events, and PBL sorted as live CD14^- events ( Supplementary Material 3 ). The sorted cells were setup in an IFN-γ Enzyme Linked ImmunoSPOT (ELISPOT) assay which was coated, developed and analyzed as previously reported (13). In short, CALRmut specific T cells were aliquoted in 10⁵ aliquots in triplicates. T cells were stimulated with either CALRLong36 peptide (positive control), 10⁴ autologous monocytes, 10⁴ autologous PBL, or left unstimulated (negative control). To assess the cytokine activity from monocytes and PBL, 10⁴ autologous monocytes and 10⁴ autologous PBL were plated alone in triplicate wells serving as an additional negative control. Normalized counts were calculated by subtracting the spot count in the wells with unstimulated T cells and the wells with monocytes and PBL respectively from the spot count in wells with monocyte stimulated T cells and PBL stimulated T cells respectively.

Patient peripheral blood mononuclear cells (PBMC) were thawed and treated with Fc-block (Human TruStain FcX receptor blocking solution, Biolegend, San Diego, CA, USA) and stained with fluorescent antibodies ( Supplementary Material 2 ). Cells were acquired on a Novocyte Quanteon Flow cytometer (Agilent, Santa Clara, CA, USA) for the assessment of HLA-I expression, and a FACS Canto II flow cytometer (BD Biosciences, San José, CA, USA) for assessment of HLA-II, both with appropriate compensation controls. Mean fluorescence intensity (MFI) for HLA-I and HLA-II was calculated by subtracting the MFI of the isotype control sample from the HLA-I/II stained sample.

Performed as previously described (14, 15).

Total RNA purification was performed on cell pellets of CALRLong36-specific T cells or bone marrow derived T cells using the RNEasy Plus Mini Kit (Qiagen, Germantown, MD, USA) according to the manufacturer’s supplied protocol. RNA concentration was measured on a NanoDrop2000 Spectrophotometer (Thermo Scientific, Waltham, MA, USA). T cell receptor sequencing was performed using the SMARTer Human TCR a/b Profiling kit (Takara 635016, San Jose, CA, USA) according to the manufacturer’s protocols. The TCR libraries were sequenced on an Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA) using read lengths of 150 bp read 1, 8 bp i7 index, 150 bp read 2, 8 bp i5 index and 20% PhiX. TCR reads were analyzed using the Cogent NGS Immune Profiler Software v1.0 which enable data preprocessing, UMI based analysis, clonotype calling and statistical analysis. Output from the Cogent Profiler was further processed in R using the software package Immunarch (16). Clonotypes with less than two reads were excluded from further analysis. To filter out redundant TCR clonotypes identified in the specific T-cell cultures, clonotypes with a fraction< 1% were excluded from the analysis as they were deemed redundant non-specific T cells in the culture. Assessment of the number of clonotypes detected was performed using the repExplore-function and tracking of clonotypes using the trackClonotypes-function.

We based our calculations on the estimates reported by Sender and Milo (17) on total numbers and turnover of cells in the human organism. The total tumor burden can be computed from the number of neutrophils (N) plus monocytes (M) multiplied by the variant allele frequency (VAF) of mutant CALR multiplied by a factor of 2 since the mutation is heterozygous: 2·(N+M)·VAF.

The number of CALRmut specific T cells is a fraction (b) of the total T cell pool (T), CALRmut specific T cells=b·T. The number of kills per CALR mut specific T cell needed to eradicate the tumor load is then:

Each of these kills take a defined period of time, τ. The total time needed for the CALRmut specific T cells to eradicate the tumor load is then:

From Sender and Milo (17) we get the estimates: N=6·10¹¹ cells, with a lifetime of 6,6 days and M=1·10¹⁰ cells with a lifetime of 3,5 days and T=7·10¹¹. The neutrophils and monocytes have characteristic lifetimes, hence, the total kill time must be faster than the self-renewal time of tumor cells for successful eradication:

This relation was plotted for a range of parameter values using Matlab R2020b. The maximal VAF that can be eradicated is then: VAF ≤ 6,6 days·b·T/(2·τ· (N+M))

Statistics were performed using Graphpad Prism version 9 software. ELISPOT responses were evaluated using the distribution free resampling method by Moodie et al. (18) using the statistical software package R.

Our goal was to clarify if T cells specific to mutant CALR enrich in the bone marrow in patients after completion of therapeutic cancer vaccination against mutant CALR. Hence, four patients that were included in our vaccination trial (11) and displayed strong in vitro immune responses to the CALRLong36 vaccination epitope were selected as donors to generate T-cell cultures specific to the CALRLong36 peptide. Patient 1 and 9 (both ET) gave rise to CD4⁺ T-cell cultures, and patient 2 (PMF) and 7 (ET) to CD8⁺ T-cell cultures. The in vitro ELISPOT immune responses in these patients can be found in Figure 2A in the report on the vaccine trial (11). Specific T cells from these cultures were enriched using live cell FACS by stimulating specific cells with the CALRLong36 peptide for 18 hours and enriching live T cells co-expressing the T-cell activation markers CD107a and CD137 with unstimulated cells used to set the sorting gate ( Figure 2A , top and middle). Purity analysis showed satisfactory enrichment ( Figure 2A , bottom). Next, bone marrow derived T cells were enriched using live cell FACS on cryopreserved bone marrow mononuclear cells acquired at baseline and after trial completion ( Figure 2B , top) with purity assessed after enrichment ( Figure 2B , bottom). The quantity and purity of the enriched populations is provided in Table 1 . The medical records from all patients showed that the acquired bone marrow aspirations were indeed bone marrow aspirations and not peripheral blood, a feat that is not uncommon in patients with MPN. Additionally, the description by the hematopathologists showed that all bone marrow specimens harbored lymphocyte infiltrates.

As the presence of tumor specific T cells in the tumor microenvironment is necessary to attain a clinical response to cancer immune therapy (19), we investigated if T cells specific for mutant CALR enriched in the bone marrow after completion of therapeutic cancer vaccines. Enriched CALRmut specific T cells underwent TCRSeq to identify the CALRmut specific TCR clonotypes. Next, FACS sorted bone marrow derived T cells were analyzed by TCRSeq. Only clonotypes with two or more reads were included in the in silico analysis. TCRSeq of bone marrow derived T cells showed a high variation in the frequency of TCRs identified ( Figures 3 , 4 and Table 2 ). For example, only 3601 different α-chains were identified in the sample acquired from patient 1 at baseline ( Figure 3A and Table 2 ), whereas 52996 different β-chains were identified in the baseline sample from patient 7 ( Figure 4A and Table 2 ). Naturally, the TCR repertoire from the specific cultures was much lower compared to the bone marrow derived T cells ( Figures 3 , 4 ). Low frequency TCRs in the specific T cells could be attributed to background activity from allogenic feeder cells used to expand the specific T cells, and in the downstream analysis it was assumed that only TCRs with a frequency of > 1% in the specific T-cell samples represented CALRmut specific T cells, thereby reducing the amount of TCRs identified in the specific T-cell fraction ( Table 2 ).

Generally, the TCRs in the baseline and end-of-trial samples remained relatively constant as clonotype tracking of the 100 most frequent clonotypes in the end-of-trial samples were present at high frequencies in the baseline samples too ( Figures 3B, D , 4B, D ).

Next, we identified TCRs that were present in the specific T-cell cultures with a frequency of > 1% and tracked these clonotypes to the baseline and end-of-trial samples ( Figure 5 ). Surprisingly, clonotypes identified in the CALRmut specific T-cell fraction were identified in very low fractions in both baseline and end-of-trial samples ( Figure 5 and Table 2 ). For patient 1, 23 α- and 22 β-chains were identified in the specific T cells. None of these were identified in the baseline sample, but three α− and six β-chains were identified in the end-of-trial bone marrow T cells albeit at a very low frequency ( Table 2 and Supplementary Material 4 ). For patient 2, five α- and five β−chains were identified in the specific T cells, however none of the α−chains were identified in neither the baseline nor the end-of-trial sample. No β-chains from the specific cells were identified in the baseline sample, and only one β-chain was found in the end-of-trial sample at a frequency of 0,001% of all clonotypes ( Table 2 and Supplementary Material 4 ). The specific T- cell culture from patient 7 was almost monoclonal displaying only one β-chain and two α-chains. The β-chain was not identified in neither of the bone marrow samples, and only one of the α-chains was found in the end-of-trial sample at a low frequency of 0,018% ( Table 2 and Supplementary Material 4 ). In the patient 9 derived specific T cells, we managed to detect 19 α-chains and 13 β-chains respectively. Of the former none were detected in neither baseline nor in the end-of-trial samples. Of the latter, one β-chain was identified in the baseline sample at a frequency of 0,016% but was not identified in the end-of-trial sample. Another β-chain from the specific T-cell culture was identified in the end-of-trial sample at a frequency of 0,046%. However, in the specific T cell culture, the clonotype was only the 8^th most frequent clonotype and displayed a frequency of only 3,5% of the total T cell repertoire making it unlikely that this clonotype represented a TCR specific to mutant CALR ( Table 1 and Supplementary Material 4 ). Thus overall, we have found that CALRmut specific T cells do not enrich in the bone marrow in patients that have received therapeutic cancer vaccines against mutant CALR.

We have previously demonstrated that CALRmut specific T cells recognize autologous PB myeloid cells and hematopoietic stem cells (HSC) (7, 8). However, the failure to identify enrichment of specific T cells in the bone marrow and target the transformed cells herein could be explained by the fact that the enriched T cells in this study are unable to recognize autologous CALRmut cells. Thus, we investigated the recognition of autologous myeloid cells and PBL. These were enriched by FACS ( Figure 6A ) and used as target cells in an IFN-γ ELISPOT assay. In accordance with earlier findings (7, 8), the specific T cells recognized myeloid derived target cells, whereas PBL were not recognized ( Figure 6B ). Analysis by ddPCR confirmed that the monocytes were indeed CALRmut with a CALRmut VAF ranging from 0,94% to 52%. Surprisingly, all PBL fractions also showed some degree of mutants with a CALRmut VAF ranging from 0,67% to 14%. The enriched cells showed a high purity of all fractions, but we cannot rule out the possibility that some PBL fractions were contaminated with monocytes. However, low/absent responses to PBL could also be explained by the lower expression of both HLA-I ( Figure 6C ) and HLA-II ( Figure 6D ) on PBL compared to monocytes.

Previous data suggests that the CALRmut VAF in PB provides a good estimate of PB myeloid cell CALRmut VAF (20). Thus, a 40% CALRmut VAF in PB indicates a PB myeloid cell CALRmut VAF of approximately the same value. As essentially all patients with CALRmut are heterozygous (21), this translates into 80% of all myeloid cells carrying the mutation and consequently posing as targets of specific T cells. Both monocytes and neutrophils are produced in high quantities in man (17) implying a high tumor burden in CALRmut MPN. Thus, failure of T-cell enrichment in the bone marrow could be explained by an overload of transformed cells in the periphery.

Based on assumptions described in the materials section we performed in silico analyses to quantify the transformed neutrophils and monocytes in the human organism and compare this to the amount of CALRmut specific T cells. The calculations rely on the fraction of CALRmut specific T cells of the entire T-cell pool (b). An ex vivo ELISPOT on PBMC from patient 2 and 7 showed no response in the former, whereas a response was identified in the latter with approximately 45 of 6 x10⁵ assayed cells recognizing the CALRLong36 peptide ( Figure 7 ). This provided a fraction of CALRmut specific T cells in the entire T-cell pool (b) of approximately 10^-4. As patient 2 did not show any ex vivo response 10^-5 was included as a value, as were more optimistic T-cell fractions of 10^-1, 10^-2 and 10^-3. The period of time (τ) a T cell requires to kill a cancer cells was derived from Halle et al. (22) and three different killing times (90 minutes, 4 hours and 12 hours) were assumed. The fraction between the CALRmut cells in the periphery and CALRmut specific T cells was plotted displaying the number of kills required by each specific T cell to break even with the number of transformed cells ( Figure 8A ). This showed that even with a high value of b (10^-1), the amount of CALRmut cells outnumbers the T cells. The median CALRmut VAF in the vaccination trial was 41%, and calculations showed that even if a highly unrealistic 1/10 of all T cells were specific to mutant CALR, the CALRmut specific T cells would require to kill several CALRmut cells. Provided a more realistic fraction of specific T cells of 1/10 000 (b=10⁻⁴) and a CALRmut VAF of 40%, each specific T cell needs to kill approximately 7 000 target cells ( Figure 8A , left). Another factor to account for is the time required for the T cell to kill the target cell. Thus, we calculated the total kill time required provided different values of τ ( Figures 8B–D ). Assuming a killing time of 4 hours per kill, 10⁻⁴ specific T cells and a CALRmut PB VAF of 40%, approximately 400 days is needed for the cellular immune system to clear the mutant cells ( Figure 8C , left). Providing a killing time of only 90 minutes, cellular immunity can only match a CALRmut VAF< 1%. With τ=90 minutes and b=10⁻⁴, the maximal VAF that can be eradicated is 0,6% ( Figure 7B , right). Next, the maximal time a T cell should use to kill a transformed cell to balance the amount of newly formed transformed cells was analyzed ( Figure 8E ). A realistic T-cell fraction of 10⁻⁴ shows that even with a killing time of only one hour, the maximally allowed VAF that the T cells can control is 1% ( Figure 8E ). Finally, the maximum CALRmut VAF which the immune system can control was calculated ( Figure 8F ). Even provided a short killing time and a high fraction of specific T cells, the CALRmut VAF is not controllable by T-cell immunity. In conclusion, by using in silico analyses we provide data which imply that the levels of transformed cells in CALRmut MPN massively exceeds the number of cells which the cellular immune system is able to curtail.