Human specimens were obtained by surgical resections at Hôpitaux Universitaires de Strasbourg, France. Their tumoral status was confirmed by anatomopathological analysis. The anatomopathologist confirmed the presence of a tumor on mirror samples based on morphology studies with a hematoxylin & eosin staining ( Supp. Data Sheet Figure 1 ). Healthy lung tissue samples were obtained from the peri-tumoral tissue of the same donor. Patients’ informed consents were managed by the Centre de Ressources Biologiques (CRB). For information, lung cancer patients that did not present a driver mutation were selected for this study.

After the reception, human specimens were washed in PBS to remove blood excess, then enzymatically digested using the tumor Dissociation Kit, Human (ref. 130-095-929, Miltenyi Biotech) with the gentleMACS™ Octo dissociator with heaters. This kit was optimized for facilitating the maintenance of tumor-infiltrating lymphocytes while preserving important cell surface epitopes. The content of the C tube (ref. 130-093-237, Miltenyi Biotech) is filtered through a 70µm cell strainers. Strained cells were centrifuged, and pellets were resuspended in 1mL lysis buffer (ref. R7757-100mL, Sigma) to remove the remaining blood cells. A second centrifugation was performed, and the left cells were resuspended in 1mL DMEM-F12 (ref. BE12719F, Lonza).

As described before in our previous studies, for the formation of patient-derived tumoroids (PDTs) and healthy organoids (PDHOs) in a matrix-free condition, supportive cells such as adipose-tissue-derived microvessels (ad-MVs) were added (12). They were prepared according to the protocol from the supplier Advanced Solutions® and the previous studies.

PDTs and PDHOs were prepared according to a mix of 5000 patients’ cells and 5000 ad-MVs (within 1 PDT or PDHO). Cells’ suspension was then diluted in cell culture media which is composed of a mix of DMEM high glucose (ref. 41966-029, Gibco), RPMI (ref. 10101-145, Sigma) and TexMACS™ media (ref. 130-097-196, Miltenyi Biotech). This cell culture media was supplemented with different growth factors to improve patient ‘s cells in vitro culture, maintenance of tumor-infiltrating immune cells and ad-MVs (details in Table 1 ).

After that, 200µL of this cell suspension was transferred on a ULA U bottom 96 wells plate (ref. 174929, Thermofisher) to form PDTs and PDHOs. The medium was changed every 3-4 days by replacing IL-2 with two other interleukins: IL-7 at 155UI/mL (ref. 130-095-361, Miltenyi Biotech) and IL-15 at 290UI/mL (ref. 130-095-762, Miltenyi Biotech).

Primary tissue samples were fixed in 10% neutral-buffered formalin (ref. HT501128-4L, Sigma-Aldrich) and processed for histologic examination including hematoxylin and eosin (H&E) (Supp. Data Sheet Figure 1). 5µm thick tissue sections from selected paraffin blocks for each specimen were used for immunohistochemical analysis. PDTs and PDHOs were fixed in 4% PFA (ref. 416250397, Roti-Histofix®) for 1 hr at room temperature (RT) and embedded in Histogel specimen processing gel (ref. HG-4000-012, ThermoFisher Scientific) before dehydration in the Pearl. After paraffin inclusion, the blocks were sectioned on the Leica microtome at the thickness of 5μm. Then, IHC staining were performed on these sectioned slides using the LEICA Bond-II system with Bond Polymer Refine Detection based on the Novolink-polymer (ref. 7161, Leica), which is a horseradish peroxidase (HRP)-based polymer conjugated fluorescent dye that labels anti-rabbit antibody. The first steps of IHC are the dewaxing and the antigen unmasking (either High pH buffer or citrate buffer during 20min). Saturation of endogenous peroxydases with 10min of incubation in 3% H2O2 (ref. H1009, Sigma) and blocking step with goat serum for 30min (ref. G6767, Sigma) were performed. Primary antibodies were incubated for 1 hr at RT ( Table 2 ). Then, after 1 hr of incubation with the primary antibody, secondary antibodies [or post-primary which is a rabbit anti-mouse IgG (ref. 7161, Leica)] and tertiary antibody (Novolink-polymer which is an anti-rabbit Poly-HRP) were incubated during 30 min at RT sequentially (ref. 7161, Leica). The post-primary was applied for primary mouse antibodies only. The final step was the signal amplification based on TSA using Perkin Elmer kit (ref. SAT701B, Perkin Elmer), for 10min at RT. Cell nucleus were stained with DAPI (dilution 1:10000 in PBS; ref. B2883, Sigma) during 10min at RT. The washing steps were performed between each step with Bond wash solution 1X (ref. AR9590, Leica). Images were captured using the fluorescence microscope Nikon Eclipse 90.

The value of dilution is based on the concentration of the antibody in the stock solution.

H&E staining was also performed on paraffin sections that were prior deparaffinized in xylene (ref. 185566, Honeywell) for 5 min twice and in different ethanol solutions (3 min twice in ethanol 100%, 5 min in ethanol 70%, and 5 min in ethanol 30%). Colorations with hematoxylin (ref. HHS-16, Sigma) for 3 min and eosin (ref. 318906, Sigma; diluted in distilled water at 1/50) for 30 sec were done. Differentiation for 2 min with ethanol 80% was done, followed by dehydration with ethanol 100% for 2 min and xylene for 2 min. Mounting was done with the Eukitt (ref. 045798, D. Dutscher).

PDTs and PDHOs were fixed for IHC staining of CD4, CD8 and CD20 in 4% PFA (ref. 416250397, Roti-Histofix®) for 1 hr at room temperature (RT) and embedded in Histogel specimen processing gel (ref. HG-4000-012, ThermoFisher Scientific) before dehydration in the Pearl. After paraffin inclusion, the blocks were sectioned on the Leica microtome at the size of 5μm. Then, IHC stainings were performed on these sectioned slides (Supp. Data Sheet Figure 7). IHC staining was performed on the LEICA Bond-II system using Bond Polymer Refine Detection based on the Novolink-polymer (ref. 7161, Leica), which is a horseradish peroxidase (HRP)-based polymer conjugated fluorescent dye that labels anti-rabbit antibody. The first steps of IHC are the dewaxing and the antigen unmasking (either High pH buffer or citrate buffer during 20 min). Saturation of endogenous peroxydases with 10 min of incubation in 3% H2O2 (ref. H1009, Sigma) and blocking step with goat serum for 30 min (ref. G6767, Sigma) were performed. The first primary antibody (anti-CD4) was incubated for 1 hr at RT. Then, after 1 hr of incubation with the primary antibody ( Table 3 ), post-primary, a rabbit anti-mouse IgG (ref. 7161, Leica) was incubated for 30 min, and a tertiary antibody (Novolinkpolymer which is an anti-rabbit Poly-HRP) was then incubated during 30 min at RT (ref. 7161, Leica). The final step was the signal amplification based on TSA using Perkin Elmer kit (ref. FITC FP1018, Perkin Elmer), for 10 min at RT. A prior step to eliminate the first primary antibody and to inhibit the enzymes was used with the Linblock reagent (ref. RAG0149UK, Linaris). It was incubated twice for 2 min at RT, before saturation of endogenous peroxydases and blocking step with goat serum. The second primary antibody (anti-CD8) was incubated for 1 hr at RT, followed by the post-primary and tertiary antibody as described before. The following step was the signal amplification based on TSA using Perkin Elmer kit (ref. Cy5 FP1117, Perkin Elmer), for 10 min at RT. Then, the linblock reagent was used a second time to perform the staining of anti-CD20 and the protocol was the same, except that the signal amplification was based on TSA with the fluorescent dye Cy3 (ref. FP1046). At the end of the protocol, cell nuclei were stained with DAPI (dilution 1:10000 in PBS; ref. B2883, Sigma) during 10 min at RT. The washing steps were performed between each step with Bond wash solution 1X (ref. AR9590, Leica). Images were captured using the fluorescence microscope Nikon Eclipse 90.

VACV or TG6002 is a replication-competent Copenhagen Strain vaccinia (20). The complete DNA sequence of vaccinia virus: thymidine kinase gene-inactivated, ribonucleotide reductase gene-inactivated.

VACV GM-CSF+ or JX-594 is a replication-competent Wyeth strain vaccinia, thymidine kinase gene-inactivated (18, 19).

The oncolytic viruses were amplified using chicken embryo fibroblasts and purified using tangential flow filtration (TFF). Briefly, the crude harvest containing infected cells and culture supernatant and conserved at -20°C, was thawed at room temperature and the viral suspension was homogenized using a homogenizing mixer equipped with an in-line chamber. Large cellular debris were then eliminated by depth filtration using filters of 5 µm pore size. The clarified viral suspension was subsequently concentrated and diafiltered in a formulation buffer by using filtration and 0,2 µm pore size hollow fibers. Finally, the purified virus was further concentrated using the same tangential flow filtration system and aliquoted before storage at -80°C until use.

The viral titer of the purified material was then measured using plaque assay on Vero cells (described in section 2.6.).

Oncolytic viruses were prepared according to a multiplicity of infection (MOI) of 0,1. This means that one viral particle is applied to ten cells. Mock is the condition containing only DMEM media. PDTs and PDHOs were incubated with oncolytic viruses for 96 hrs.

PDTs and PDHOs were fixed in 4% PFA (ref. 416250397, Roti-Histofix®) for 24 hrs at 4°C. The day after, PDTs and PDHOs were incubated with PBS-Triton X100 0,25% (ref. X100, Sigma) during 30 min at RT then with PBS-BSA 5% (ref. A9647-100G, Sigma) during 4h at RT. Primary antibody (Table 4) was prepared in PBS-BSA 5% and incubated for 3 days at 4°C.

Four PBS-washings were performed at 30 min intervals (at RT), including one overnight washing at 4°C. Then, secondary antibody (Table 5) was diluted in PBS-BSA 5% and incubated for 24 hrs at 4°C.

Three PBS-washings were performed at 30min intervals (at RT), and then, DAPI (ref. B-2883, Sigma; Dilution 1:2500) was added to PDTs and PDHOs for 1 hr at RT. PBS then replaced DAPI until acquisition on confocal microscopy LSM Zeiss 800.

Prior to titration, supernatants, and PDTs or PDHOs were harvested in duplicate for each condition and stored at -80°C before use.

The day of titration, samples were thawed in water bath at 37°C and then refrozen in dry ice. Samples were thawed and refrozen alternatively at least three times to lyse the PDTs and PDHOs. Vero cells (ATCC CCL-81™) were split and seeded at 5.10⁴ cells/well in a six wells-plate. Cell culture media was DMEM (ref. 41966029, Gibco), supplemented with 1% gentamycin and 10% FBS. The day after, serial dilutions (E^-1 to E^-6) of supernatants comprising lysed PDTs and PDHOs were prepared. Dilutions were done in PBS (ref. D8537-500mL, Sigma) supplemented with 1% FBS and 1% of cations [Mg (CH₃COO)₂ (Cf = 10g/L) and CaCl₂ 2(H₂O) (Cf= 10g/L)]. These different solutions were then incubated for three days at 37°C. Solutions containing viral particles produce some lysis zones on Vero cells, which are also called viral plaques. These viral plaques are counted after three days of incubation by using the neutral red solution 10% (ref. N2889-100mL, Merck-Millipore) diluted in DMEM media with 20% agarose (Cf = 50g/L) (ref. A9045-250g, Sigma).

In our study, we generated PDTs and PDHOs from lung adenocarcinoma and the healthy counterpart of the tumor respectively (refer Method section: lung tumor and healthy tissue engineering). They were cultured according to the methods described in previous research (12). PDTs and PDHOs were infected with a multiplicity of infection (MOI) of 0,1 to evaluate the permissivity of patient’ s cells to oncolytic viruses, for 96 hrs ( Figure 1A ). We assessed main lung adenocarcinoma marker, TTF-1 (21) on primary tumor tissues ( Figure 1B ) and primary healthy tissue ( Supp. Data Sheet Figure 2 ), then verified its expression on respective PDTs and PDHOs. Other main biomarkers of lung adenocarcinoma such as cytokeratin 7 (CK7) and mucin-1 (MUC1) were also assessed by IHC on PDTs and PDHOs. Globally, PDTs could maintain the expression of these three biomarkers, and more cells were stained in PDTs than in PDHOs ( Supp. Data Sheet Figures 3 , 4 ).

The biomarker Ki-67 defines a rapid cell division and high metabolic activity; we then evaluated the biomarker Ki-67 in concomitance with TTF-1. We can observe that patients 22T 441, and 23T 37 (both in stages IIB) were the ones presenting more Ki-67 expression in primary tumor tissues and PDTs ( Supplementary Table 1 ; Figure 1B ). The quantification of viral particles produced in cells was evaluated with the plaque assay. We can observe that the number of viral particles increases when the biomarker Ki-67 is predominant generally ( Figure 1C ). This biomarker was more expressed for patients 22T 441 and 23T 37.

Furthermore, the number of viral particles was more important in PDTs than in PDHOs. This observation is consistent as we observe low expression of Ki-67 in healthy primary tissues and matched PDHOs ( Supp. Data Sheet Figure 2 ).

The presence of vaccinia virus particles can be further confirmed with immunofluorescence using an antibody targeting vaccinia virus ( Figure 1D ; Supp. Data Sheet Figure 5 ). For organoids from patient 22T 441, we could observe a positive staining for vaccinia virus with both oncolytic viruses. On PDTs, VACV infected on the periphery whereas VACV GM-CSF+ infected cells with more spreading ( Figure 1D ). On PDHOs, VACV and VACV GM-CSF+ infected fewer cells compared to PDTs. This observation supports onco-virotherapy specificity for tumoral cells.

Viral replication is known to be supported by the host cellular metabolic state within the TME. We could point out that not all tumor-derived cells were metabolically active. Few co-staining of biomarkers Ki-67 and TTF-1 was observed in both type of samples. As both oncolytic viruses were deleted for enzymes involved in VACV DNA synthesis, this makes them rely heavily on the pool of these enzymes present in the target tumoral cells to support their proliferation (20). Contrary to what we might think, not all live tumoral cells are highly proliferative. These low metabolic states and quiescent states were already described in solid tumors (22). And this lack of metabolic activity could hinder oncolytic viruses’ efficiency. In this part, results supported specific replication of both oncolytic viruses within PDTs.

Prior to performing proteomics to evaluate immunostimulatory properties of oncolytic viruses, we checked the maintenance of primary tumor immune cells within PDTs and PDHOs. We have optimized the cell culture media to maintain them along the culture ( Figure 2 ; Supp. Data Sheet Figures 6 , 7 ). These immune cell subpopulations from the primary tissue are likely to have a better predictive value of prognosis and therapeutic responses (23). Maintaining these tumor-infiltrating immune cells in our 3D model was more relevant to reflect the tumor immune microenvironment better. We used the biomarker CD45 to assess the maintenance of tumor-infiltrating immune cells within PDTs, PDHOs, and their corresponding primary tissues ( Figure 2 ; Supp. Data Sheet Figures 6 , 7 ). We could observe CD45 biomarker expression in both type of samples (tumor tissues and PDTs). Two patients 22T 384 and 23T 37 presented more immune cells in their primary tumoral tissues, and this was also reflected on their PDTs. Then, healthy primary tissues presented fewer immune cells compared to the tumoral primary tissue. This was also observed on PDHOs compared to PDTs ( Supp. Data Sheet Figure 6 ). As for patients 22T 345 and 22T 351, no CD45 expression was observed in corresponding PDTs. In patient 22T 351, we even observed a loss of CD45 expression, this could be due to the difficulty of maintaining immune cells in vitro.

Immune or tumoral cells can express PD-L1 to reduce antitumor immunity and inhibit T cell activation. The knowledge of PD-L1 tumoral cells expression is important as this ligand has emerged as a potential biomarker to predict responses to immunotherapy (24, 25). Then, we assessed PD-L1 biomarker on PDTs as well and, we could find the expression of PD-L1 in PDTs ( Figure 2 ). When there’s no colocalization of PD-L1 with CD45, we suggest that it is expressed by tumoral cells. This was the case for all patients except patient 22T 345 whose PD-L1 expression is 0%. In contrast, PDTs from patient 22T 441 presented more PD-L1 expression, which is up to 60% in clinical features ( Supplementary Table 1 ).

These PDTs and PDHOs could maintain immune cells from the primary tissue. Hence, from these PDT and PDHO model, we could assess immune-related effects following oncolytic viruses’ treatment.

The use of genomics on fresh tumor tissues or formalin-fixed paraffin-embedded (FFPE) blocks has permitted to have insights on genetic variants that predict responses to drugs (26, 27). The cancer genome can be completed by having a look on the protein products from these genes (intracellular and membranous proteins) (28). These proteins represent most of the human proteome besides secreted proteins. They are the molecules mostly responsible for cell growth and cancer progression (28). To better understand oncolytic viruses’ effect at the protein level, we measured proteins relative expression with the proximity extension assay (PEA) technology. This technology is supported by the Olink platform (Olink Target 96 Immuno-Oncology). We mainly investigated the immune responses at the intracellular proteins ( Figure 3A ).

To investigate oncolytic viruses’ global effect on the five patients, specifically selected for their different histology, proteomic data can be summarized under a Volcano Plot representation ( Figures 3B, C ). This data representation permits an overview of differentially expressed proteins on PDTs and PDHOs, either treated by VACV or VACV GM-CSF+. After both oncolytic viruses’ treatment, levels of IFN-γ were upregulated ( Figures 3B, C ). This upregulation is a sign of antiviral responses in the TME (29). Then, both oncolytic viruses mainly presented a downregulation of proteins involved in tumoral cell progression and survival, such as interleukin-8 (IL-8), C-C motif chemokine ligand 20 (CCL20), and galectin-1 (Gal-1) ( Figures 3B, C ). IL-8 (30) is a chemokine known for its pro-inflammatory and pro-angiogenic effects in NSCLC. CCL20 (31) is an oncogenic chemokine favoring tumoral progression, and Gal-1 (32) is known to favor adhesion, proliferation, and metastatic processes of tumoral cells. On PDTs, VACV GM-CSF+ could induce downregulation of other proteins involved in tumoral progression, such as CX3CL1 and CXCL12 chemokines. Regarding the effect of oncolytic viruses on PDHOs, we could note a downregulation of the TNFRSF21 (33) protein, mainly involved in the negative regulation of T lymphocytes ( Figures 3B, C ). Besides the downregulation of pro-tumorigenic effects, oncolytic viruses can elicit immune responses.

To assess the global effect of viral infection on patients’ PDTs, a gene set enrichment analysis using KEGG pathways was also done to study the pathway significantly enriched in our comparisons infected versus mock (Supplementary Table 2, Supp. Data Sheet Figure 8 ). The main pathway enriched by both oncolytic viruses was the chemokines pathway. Among the enriched chemokines, we can note the decrease of CCL20 expression observed previously and CCL4 (34) and that promote cancer progression. We can also note a decrease expression of chemokine implied in decreased activity of CD8+ T cells such as CXCL5 (35) by both oncolytic viruses ( Supp. Data Sheet Figure 8 ).

The heatmap permits to cluster the proteins’ levels by the nature of the samples (tumor or healthy), patients’ individuality and treatments’ conditions. The proteins’ levels were defined according to the normalized protein expression values (NPX) on a log₂ scale. Oncolytic viruses’ effects varied from one patient to another.

If looked into details, PDTs derived from patients 22T 345 and 22T 351 showed some lower protein z-score than other patients. This observation was following clinical features which are a low PD-L1 expression and the absence of CD45 expression ( Supplementary Table 1 ). These features reflect a cold tumor (36) ( Figure 2 ; Supplementary Table 1 ). Patient 22T 345 presented a slight immunomodulatory effect under VACV and VACV GM-CSF+ ( Figure 4 ). Indeed, we could observe that the CD27 (37) protein was 1,5 to 2-fold higher in PDTs treated with both oncolytic viruses.

In contrast with these two latter patients, PDTs derived from patients 23T 37 and 22T 441 could demonstrate some immunomodulatory responses, mostly under VACV GM-CSF+ treatment. We will describe each case individually and briefly. In PDTs from patient 22T 441, we could observe an increase of IFN-γ (38), which is also a cytokine mediated by cytotoxic T cells in addition to be linked to antiviral responses. As for patient 23T 37, we observed more immune-mediated cytotoxicity effect. Indeed, we observed that proteins involved in NK-cell mediated cytotoxicity were 2-fold higher in PDTs VACV GM-CSF+ treated, with the increase of proteins Natural Cytotoxicity Triggering Receptor (NCR1) (39) that mediates MHC non-restricted cytotoxicity, and Killer cell immunoglobulin-like receptor (KIR3DL1) (40) which is the ligand for the surface protein KLRD1, involved in NK cell signaling. We can also note a slight increase of IFN-γ (38), TNF-α (41) and FASLG that participate to effector functions of NK and T lymphocytes. Then, we could observe a slight increase of Cytotoxic and Regulatory T cells Molecule (CRTAM) (42) protein involved in CD8+ T cells response ( Figure 4 ; Supp. Data Sheet Figure 8 ). In summary, this patient has more response to oncolytic virotherapy. This patient presented a PD-L1 expression > 90% regarding its clinical criteria. As mentioned previously, PD-L1 expression level is among the biomarkers to define a response to immunotherapy. This suggests to integrate PD-L1 expression with tumor immune cells infiltrate to refine onco-virotherapy selection for patients (43) ( Figure 4 , framed in pink). Indeed, PD-L1 expression and a high T-cell infiltration are part of features to define a “hot” tumor which is mostly effective to immunotherapies (44).

In general, VACV GM-CSF+ presented more immunomodulatory effects than VACV mainly because of the GM-CSF transgene which might have potentiated an antitumoral immune response. The encoded transgene GM-CSF was mostly released in PDTs infected by VACV GM-CSF+ than in VACV ( Supp. Data Sheet Figure 9 ).

As the research for biomarkers often relies on the secreted proteins, we also studied the proteomic signature in the secretome of two other patients 22T 31 and 22T 67 according to the same treatment workflow (Supplementary Table 3, Supp. Data Sheet Figure 10A ). Indeed, biological fluids such as blood samples are used to predict outcomes on patients under ICIs treatment (45). Clinical biomarkers such as TTF-1, Ki-67, CD45 and PD-L1 were assessed in the same way as other patients ( Supp. Data Sheet Figures 10B, C ). Our experimental design permits us to establish the signature of secreted proteins after VACV and VACV GM-CSF+ treatments under a heatmap representation ( Supp. Data Sheet Figure 10D ). From PDTs of patient 22T 31, we can note that more change was seen at the level of secreted proteins. Due to its higher clinical relevance, we analyzed in greater detail the secreted proteins under VACV GM-CSF+ infection. We could observe the increase of some proteins involved in T cells activation such as CD28 (46), CD40L (47), CD70 (37), CD244 (48), CRTAM (42) and IL12 receptor subunit beta 1 (IL12RB1) (49). Also, NK-cell mediated cytotoxicity was observed with the increase of proteins NCR1 (39), KIR3DL1 (40), and Killer cell lectin-like receptor subfamily D member 1 (KLRD1) (50) or CD94 which is a surface protein involved in NK cell signaling ( Supp. Data Sheet Figure 10D , proteins marked with a pink star). PDTs from patient 22T 67 showed less response (similarly to the intracellular proteins’ signature), we only observed a slight increase of IL2 (51), IL5 (52), and KIR3DL1 (40) ( Supp. Data Sheet Figure 10D , proteins marked with a green star). Interestingly, we also observed a decrease of lymphocyte-activation gene 3 (LAG3) protein (53), which is involved in the TCR inhibition of T cells. Its inhibition is beneficial to an anti-tumoral effect. Overall, different profiles of drug sensitivity were observed; this could be explained with patients’ heterogeneity which is a hurdle to drugs efficiency (54).

As observed with the intracellular proteins, the immunomodulatory response induced by oncolytic viruses, especially VACV GM-CSF+, was also reflected with the secreted proteins. This demonstrates the relevance of combining the study of secreted and intracellular proteins to map the expression of the proteins under treatment. Thus, we have shown the usefulness of proteomics with tumoral organoids to bring more insights into onco-virotherapy responses. The use of proteomics gives multiple readouts in protein profiling and permits to stratify responders from non-responders.