key: cord-0307464-oytpkh3r
authors: van Puffelen, Jelmer H.; Novakovic, Boris; van Emst, Liesbeth; Kooper, Denise; Zuiverloon, Tahlita C.M.; Oldenhof, Ursula T.H.; Witjes, J. Alfred; Galesloot, Tessel E.; Vrieling, Alina; Aben, Katja K.H.; Kiemeney, Lambertus A.L.M.; Oosterwijk, Egbert; Netea, Mihai G.; Boormans, Joost L.; van der Heijden, Antoine G.; Joosten, Leo A.B.; Vermeulen, Sita H.
title: Intravesical BCG in patients with non-muscle invasive bladder cancer induces trained immunity and decreases respiratory infections
date: 2022-02-21
journal: bioRxiv
DOI: 10.1101/2022.02.21.480081
sha: 15f9bbf148b5e7e859e5f62d7916e73f927ed22b
doc_id: 307464
cord_uid: oytpkh3r

Bacillus Calmette-Guérin (BCG) is recommended as intravesical immunotherapy to reduce the risk of tumor recurrence in patients with non-muscle invasive bladder cancer (NMIBC). Currently, it is unknown whether intravesical BCG application induces trained immunity. Here, we found that intravesical BCG does induce trained immunity based on an increased production of TNF and IL-1β after heterologous ex-vivo stimulation of circulating monocytes 6- 12 weeks after intravesical BCG treatment; and a 37% decreased risk (OR 0.63 (95% CI 0.40- 1.01)) for respiratory infections in BCG-treated versus non-BCG-treated NMIBC patients. An epigenomics approach combining ChIP-sequencing and RNA-sequencing with in-vitro trained immunity experiments identified enhanced inflammasome activity in BCG-treated individuals. Finally, germline variation in genes that affect trained immunity was associated with recurrence and progression after BCG therapy in NMIBC, suggesting a link between trained immunity and oncological outcome.

Intravesical instillations with Bacillus Calmette-Guérin (BCG) is the recommended adjuvant therapy in patients with high-risk non-muscle invasive bladder cancer (HR-NMIBC). HR-NMIBC is associated with a high risk of tumor recurrence and progression to muscle-invasive bladder cancer. BCG immunotherapy generally consists of an induction course with 6 weekly instillations, followed by 3-weekly maintenance courses at months 3 and 6 and then every 6 months for up to 3 years 1 . Although BCG is more effective in preventing tumor recurrences than intravesical chemotherapy 2, 3 , non-responsiveness to BCG is observed in more than 25% of HR-NMIBC patients within 5 years 4, 5 . The exact immunological mechanisms through which BCG mediates anti-tumor immunity in bladder cancer are still unclear. Initiation of a T helper 1 (T h 1) cell immune response is important to achieve a good clinical response 6, 7 , and cellmediated anti-tumor activity is achieved via immune cells such as CD8+ T cells 8, 9 , NK cells 10, 11 and macrophages 12, 13, 14 .

It is known that activated myeloid cells can further increase T h 1 immune responses by producing TNF, IL-1β, IL-6, IL-12 and IL-18 15, 16, 17, 18, 19 . These cytokines are produced after stimulation of Toll-like receptors (TLR) and NOD-like receptors (NLR) by BCG 20, 21, 22 . TLR2 and TLR4 are especially important for the production of IL-1β by monocytes after BCG stimulation 23 . Additionally, BCG vaccination and mechanistic studies have demonstrated that following BCG stimulation, myeloid cells can develop trained immunity (TI) 24 , which is characterized by long-term functional and epigenetic reprogramming of myeloid cells, with enhancement of their function 25 .This TI phenotype allows myeloid cells to respond with an increased cytokine response upon encountering a secondary stimulus or challenge 24, 25 . It contributes to the increased protection against non-tuberculosis infections, including respiratory infections, after BCG vaccination 26, 27, 28 . Mechanistic studies found that autophagy 29 , IL-1β signaling 26, 30, 31 , IL-32 signaling 32, 33 , NOD2-RIPK2 signaling 34 , epigenetic enzymes 35, 36 , metabolic reprogramming 37 , and reprogramming of hematopoietic stem and progenitor cells (HSPCs) in the bone marrow 38, 39 , are important for the induction of TI by BCG.

Whether repeated BCG instillations in NMIBC also induce TI, and whether this is relevant for clinical responses, has not been systematically studied. There are no data on epigenetic modifications and metabolic changes in myeloid cells of NMIBC patients during BCG therapy.

Reports of increased urinary TNF, IL-1β, IL-2, IL-6, IL-8, IFNγ, M-CSF and GM-CSF after BCG instillations only provide indirect evidence for (systemic) TI 6, 40, 41, 42, 43 . Ex-vivo LPS stimulation experiments with blood monocytes found that TNF, IL-1β and IL-6 production was generally increased during a BCG induction cycle 29, 44, 45, 46 , indicating induction of TI, but these experiments lacked relevant long-term data. Here, we extensively investigate and describe induction of TI by BCG instillations in NMIBC by i) analyzing long-term induction of systemic TI in blood monocytes of BCG-treated NMIBC patients, ii) evaluating whether BCG instillations result in reduced incidence of respiratory infections, and iii) exploring the relation between BCG-induced TI and oncological NMIBC outcome via genetic studies.

To determine whether BCG instillations induced TI in peripheral blood mononuclear cells (PBMCs) we performed a prospective cohort study ('Tribute') and isolated PBMCs collected before BCG therapy and at 8 time points during BCG therapy (Fig. 1a) . A total of 17 BCGnaïve HR-NMIBC patients were included (see Supplementary Table 1 for inclusion and exclusion criteria), with a median age at inclusion of 67 years and 16 male patients (94%).

Twelve patients received a complete induction regimen of 6 cycles of full-dose BCG. Four patients discontinued BCG therapy due to side effects, two patients due to disease progression, one patient because of BCG shortage, and for two patients the last BCG maintenance regimen was not performed (see Supplementary Table 2 ). Six patients received 1/3 dose BCG or a reduced number of BCG instillations (or both) due to BCG shortage.

Immediately after PBMC isolation, we performed ex-vivo stimulation experiments for 24 hours and measured production of four innate cytokines, i.e., IL-1β, IL-6, TNF, and IL-1 receptor antagonist (IL-1Ra) (see Methods). The fold change in cytokine production between pre-BCG1 (baseline, BCG naïve) and time points during BCG therapy was used as indicator of the TI response, in line with methods used by BCG vaccination studies 47 . We focused on pre-BCG7, pre-BCG10 and pre-BCG13 (i.e., pre-BCG maintenance time points). At these time points, the patients had not received a BCG instillation for multiple weeks or months and hence are most informative for TI 48 .

Before the start of the first 3-weekly BCG maintenance cycle (pre-BCG7, ±6 weeks after BCG6, Fig. 1a) , the LPS-induced production of TNF and IL-1β was elevated compared to pre-BCG1

(median fold change (MFC) TNF: 1.70 (p=0.035); IL-1β: 1.50 (p=0.020)) ( Fig. 1b and 1c) . The TNF and IL-1β production after P3C stimulation was also increased, but not at the level of nominal statistical significance (MFC TNF: 1.26 (p=0.110); IL-1β: 1.05 (p=0.194)) ( Fig. 1d and  1e ). Similar findings were observed for IL-6 (LPS: MFC 1.21 (p=0.080); P3C: MFC 1.55 (p=0.058)) and IL-1Ra (LPS MFC 1.12 (p=0.020); P3C MFC 1.16 (p=0.058)) ( Supplementary   Fig. 1 ).

At the start of the second BCG maintenance cycle (pre-BCG10), approximately 3 months after BCG9, TNF and IL-1β production after stimulation was still increased compared to pre-BCG1 ( Fig. 1) . TNF was increased with a MFC of 1.62 and 1.87 for LPS and P3C stimulation, respectively (p=0.008 and p=0.016), and the MFC for IL-1β was 1.48 and 1.41 (p=0.016 and p=0.055). Production of IL-6 and IL-1Ra was not increased compared to pre-BCG1 ( Supplementary Fig. 1 ). Between pre-BCG7 and pre-BCG10 there was no further increase in TNF production after LPS or P3C stimulation. The three patients from whom PBMCs were isolated at the pre-BCG13 time point did not show an increased production of TNF, IL-1β, IL-6 or IL-1Ra compared to pre-BCG1, but small numbers prevent meaningful conclusions.

Thus, for the first time, we showed that BCG instillations induce long-lasting increased innate cytokine production by circulating monocyte-enriched PBMCs and we confirmed the presence of a persistent TI phenotype before the first and second BCG maintenance cycle.

We also assessed the short-term effects of BCG on innate immune responses during the BCG induction regimen and found that one week after the first BCG instillation (at BCG2) the total number of white blood cells in the circulation was increased (p<0.01, Supplementary Table 3) compared to pre-BCG1. More specifically, the neutrophil (p<0.01) and monocyte (p<0.01) cell populations were increased. The increase in neutrophil count after a single BCG instillation is in line with a study that found increased peripheral blood neutrophil counts in human newborns 3 days after vaccination with BCG 49 . For other time points, changes in innate immune populations were either less strong or absent (Supplementary Table 3 ). At BCG2, we also observed increased TNF, IL-1β and IL-1Ra production by PBMCs after ex-vivo stimulation with LPS or P3C. IL-6 production was not increased ( Fig. 1 and Supplementary Fig. 1 ). IL-1β production after LPS and P3C stimulation remained increased at BCG6 compared to pre-BCG 1, whereas on a group level, TNF, IL-6 and IL-1Ra production was not increased ( Fig. 1 and Supplementary Fig. 1 ).

Epigenetic modifications at the gene promoters of IL6, TNF and IL1B are hallmarks of trained immunity 26, 50 . Using ChIP-qPCR we assessed histone 3 lysine 4 trimethylation (H3K4me3) modifications, an activating epigenetic modification linked to BCG-induced TI 50 , at these promoters in a subset of 7 patients (those 7 that were included first in Tribute). Again, we focused on the pre-BCG1, pre-BCG7 and pre-BCG10 time points to assess innate immune memory. Surprisingly, in contrast to observations for cytokine production, we did not find an increase on group level in H3K4me3 modifications at pre-BCG7 or pre-BCG10 compared to pre-BCG1 (Fig. 2 ). An increased H3K4me3 signal was only observed in 2 to 4 out of 7 patients for the various cytokine gene loci. Importantly though, these 2 to 4 patients all had an increased TNF and IL-1β production after LPS stimulation. Furthermore, 3 out of the 4 patients with an increase in H3K4me3 at region 1 of the IL6 promoter showed a strong increase in IL-6 production after LPS stimulation (MFC of 2 or higher). However, there were also patients with increased IL-6, TNF or IL-1β production upon LPS or P3C stimulation that showed decreased H3K4me3 signal.

The analysis of H3K4me3 modifications via ChIP-PCR only covered specific parts of the promoter regions of TNF, IL6 and IL1B. To exclude the possibility of missing out on modification signals in other parts of the promoter due to the selection of specific primers only, we investigated changes in H3K4me3 signal in other parts of these promoters as well via analysis of genome-wide ChIP-sequencing data of these 7 patients. This more in-depth analysis revealed that, on a group level, there were no changes in H3K4me3 signal at either the gene promoter or gene body of IL6, TNF or IL1B. This result was consistent for all comparisons between pre-BCG1 and post-BCG time points. Notably, both patients without 

We further analyzed the ChIP-seq data to evaluate genomic regions that changed in H3K4me3 signal on a genome-wide scale. In total, there were 481 dynamic peaks (increasing or decreasing in H3K4me3 signal) that showed unadjusted p<0.05 between pre-BCG1 and any of the measured post-BCG time points (i.e., BCG6, pre-BCG7, BCG9, pre-BCG10) of which 375 were unique (Supplementary Table 4 ). Seventeen peaks showed an adjusted p<0.05. The strongest statistical evidence for an increase in H3K4me3 signal was found for pre-BCG10 and a region near CIDEC, coding for Cell-death-inducing-DFF45-like-effector-C and important for LPS-induced IL-1β production by epithelial cells 51 ( Fig. 3c ; at pre-BCG10); the signal was increased at BCG6 and BCG9 as well (Supplementary Table 4 ). CIDEC is also involved in the formation of pro-inflammatory macrophage foam cells 52 and controls inflammasome activation in mouse adipocytes 53 . In contrast, H3K4me3 near ATF7IP, a binding partner and regulator of SETDB1 54, 55 , which is a histone 3-lysine 9 (H3K9) methyltransferase, was decreased ( Fig. 3c and Supplementary Overall, the ChIP-seq results suggested that H3K4me3 modifications at the promoter regions of TNF, IL6, and IL1B can be induced by BCG instillations but not in all patients and without a clear link to clinical response in our small patient series. Analysis of the top hits and candidate genes suggested that genes involved with IL32, autophagy and H3K9-remodeling epigenetic enzymes, important for TI responses, may be epigenetically modified.

To determine how BCG instillations affect gene expression in monocytes, we performed RNAsequencing on circulating Percoll-isolated monocytes isolated from six (out of 7 ChIP-seq) patients for timepoints pre-BCG1, BCG6 and pre-BCG7 (Fig. 4a) . Note that the monocytes were not stimulated ex-vivo before RNA-seq analysis. We did not find statistically significant changes in gene expression of TNF, IL6, IL1B or IL1RN at BCG6 or pre-BCG7 compared to pre-BCG1 (Supplementary Table 6 and 7 and Supplementary Fig. 2 ). This is in line with BCG vaccination results described by Arts et al. 26 . Among the top hits for BCG6 versus pre-BCG1, Increased APOL2 expression is associated with polarization of macrophages to the M1 (proinflammatory phenotype) 57 . Furthermore, two defined TI candidate genes, GBP1 and GBP2, involved in inflammasome activation 58, 59, 60, 61 , showed strong but borderline non-significant evidence for increased gene expression (log2 FC: 0.83; p adj =0.052 and log2 FC: 0.54; p adj =0.114). Among the candidate genes that did not reach statistical significance after multiple testing, we found that in addition to GBP1 and GBP2, also GBP4 and GBP5 were upregulated ( Fig. 4b ) Interferon regulatory factor 1 (IRF1), an upstream transcription factor that regulates GBP gene expression, and the cytosolic DNA sensing receptor AIM2, which is activated by GBPs and IRF1 58, 62 , were upregulated as well (at p unadj <0.05). Interestingly, IL8, a major chemokine for neutrophil recruitment, was downregulated (Supplementary Table 6 ). Motif enrichment analysis revealed that 17% of the scanned promoters from the differentially expressed genes at BCG6 are enriched for an interferon-sensitive response element (ISRE), an IRF responsive motif, compared to just 1.3% of all genes (Fig. 5a) . These results suggest that IRFs, which are also transcription factors that control M1 macrophage differentiation 63, 64 , are important mediators between BCG exposure and gene induction 38 , also when BCG is instilled in the bladder. In line with this observation, we found upregulated expression of IRF1 and IRF9 at BCG6 (Fig. 5b ) and increased IFNγ concentration in blood plasma (Fig. 5c ).

At the start of BCG maintenance (pre-BCG7), 6 weeks after BCG induction there were no genes upregulated that reached the level of statistical significance after multiple testing (Supplementary Table 7 ).

A total of 26 genes remained upregulated at p unadj <0.05 from BCG6 to pre-BCG7 and 27 genes remained downregulated (Supplementary Table 8 ). Among those that were upregulated, we identified four genes involved in inflammasome activation: GBP1, AIM2, CASP5 and KCNMA1 65, 66 , and one gene involved in positive regulation of autophagy: RUFY4 67, 68, 69 .

Interestingly, three genes coding for hemoglobin subunits (HBA1, HBA2, HBB) remained downregulated.

GBPs regulate intracellular innate immune responses and activate the inflammasome for IL-1β production 58, 59, 60, 61, 62 , also after LPS stimulation 58, 60, 70 . GBPs also mediate the release of DNA from vacuoles into the cytosol for activation of DNA-sensing receptors such as CGAS and AIM2 71 . Thus, our RNA-seq data suggested that GBP-inflammasome signaling and DNAsensing by AIM2 may be increased.

The increased IL-1β production by PBMCs and the upregulation of AIM2 and GBPs suggests that inflammasome activity is enhanced and may be an important mechanism of TI during BCG therapy. To determine whether GBP and AIM2 gene expression is also upregulated in monocytes trained by direct exposure to BCG vaccine, we performed in-vitro experiments with monocytes from 5 healthy donors using a standard TI protocol 37, 72 (Fig. 4c) , and measured gene expression at baseline, after 4 hours, 24 hours, 6 days, and 6 days plus 4 hours after LPS restimulation. In line with the RNA-seq data from the BCG-treated bladder cancer patients, we found that in-vitro induction of TI by two strains of BCG increased the expression of GBP1, GBP2, GBP4, GBP5 and AIM2 (Fig. 4d ) compared to the non-BCG trained control condition, especially after LPS restimulation. These results suggested that BCG-trained monocytes, compared to the non-trained monocytes, had a primed AIM2-GBP inflammasome prior to restimulation which is then enhanced even more upon LPS restimulation. BCG may thus increase AIM2-mediated DNA-sensing and subsequently lead to increased inflammasome activation and IL-1β production.

BCG vaccination is associated with a reduced risk of pneumonia and influenza due to TI 27, 73, 74, 75 . If BCG instillations also induce a TI phenotype, we would expect to see a reduced frequency of respiratory infections. We studied this association by comparing self-reported data on respiratory infections experienced in the 18 months prior to reporting (i.e., between November 2018 and May 2020) from 407 BCG-treated NMIBC patients and 250 non-BCG treated NMIBC patients. Out of the 407 BCG-treated patients, 109 were considered BCGexposed during the whole respiratory infection outcome assessment period; the remaining 298 BCG-treated patients were exposed during a part of this period or prior to this period (partially exposed) (see Methods and Supplementary Table 9 for patient characteristics). Results of the univariable regression analysis showed a 37% decreased risk of respiratory infections for the BCG-exposed vs BCG-unexposed group (OR 0.63 (95% CI 0.40-1.01)), and a 17% reduced risk for the partially BCG-exposed versus BCG-unexposed (OR 0.83 (95% CI 0.58-1.18)) ( Table 1 ). The strongest evidence for a risk reduction in BCG-treated patients was seen for pneumonia and common cold. Multivariable analyses that included potential confounders showed very similar results (Supplementary Table 10 ). These findings suggest that BCG instillations induce heterologous protective effects against infectious diseases, in line with the induction of a systemic TI phenotype.

We previously described an association between single-nucleotide polymorphism rs3759601 in ATG2B, that influenced both the in-vitro and in-vivo training effect of BCG, and response to BCG in NMIBC patients from the Nijmegen Bladder Cancer Study (NBCS) 29, 76 . Here, we extended our genetic research and assessed whether rare to common DNA variants in the 34 TI candidate genes (Supplementary Table 5 ) affect clinical outcome of BCG-treated patients.

We analysed available exome chip data of 215 BCG-treated NMIBC patients from the NBCS (see Supplementary Table 11 for patient characteristics). Gene-based association analysis could be performed for 29 of the 34 genes and revealed no multiple testing-adjusted statistically significant finding for any of the candidate genes (Supplementary Table 12 ).

However, an unadjusted p<0.06 was found for candidates ATG2B, ATG7, ATG16L1, HNF1B, and EHMT2, with either recurrence or progression after BCG induction or BCG induction and maintenance ( Table 2 ). These results suggest that genetic variation in genes known to affect TI induction by BCG, i.e., autophagy genes 29 , a gene controlling myelopoiesis 39 , and a mediator of H3K9 methylation 35 , affect the clinical response after BCG therapy.

In this study we showed for the first time that intravesical BCG treatment induces trained immunity at a systemic level, reflected by the increased cytokine production by PBMCs isolated from peripheral blood and increased protection against respiratory infections. Also, we found that germline DNA variants in genes important for the induction of TI influence the clinical oncological outcome after BCG instillations in NMIBC.

Previous research assessing the effects of intravesical BCG application on TI responses is scarce and provides incomplete evidence. Conti et al. showed that after 18 hours of ex-vivo LPS stimulation, monocytes from three BCG-treated NMIBC patients had a significantly higher production of TNF and IL-1α compared to monocytes from three non-BCG treated NMIBC patients or three healthy age-matched controls 45 . Furthermore, Buffen et al. found an increased production of TNF, IL-1β and IL-6 after the sixth BCG instillation compared to before the first BCG instillation 29 . Kim et al. observed that TNF production increased with consecutive BCG instillations of an induction cycle and peaked after the fourth BCG instillation, but for three out of seven patients the TNF production declined between the fourth and sixth instillation 46 . Graham et al. described variable post versus pre-BCG ratios of cytokines released by monocytes following LPS stimulation in 33 NMIBC patients during BCG induction, with approximately half of the patients showing an increase in ratios 77 . A limitation of all these studies is that they only assessed the cytokine production during a BCG induction cycle which may be influenced by the immunological processes of priming or tolerance 48 .

We now showed that repeated intravesical BCG instillations induce systemic TI and added new long-term data that showed that the TI phenotype is present at the start of BCG maintenance cycle 1 and 2, despite a 6 to 12-week time interval without BCG instillations. TI may already be induced after a single BCG instillation given the observed increased cytokine production one week after the first BCG instillation by us and others 46 . Importantly, some patients showed a decreased cytokine production capacity at the sixth BCG instillation, as also observed by Kim et al. 46 , which could indicate a phenotype of immune tolerance.

We also found large inter-patient variation in in-vitro cytokine production and in activating epigenetic modifications (H3K4me3) at promoters of hallmark trained immunity genes. Indeed, variability of the trained immunity responses in human volunteers has been well described, with only approximately half of the individuals being good trainers 78 . Our Tribute patient series was too small to test for associations between patient, tumor, and clinical characteristics and magnitude of TI responses but this is a relevant topic for further research. We could only study epigenetic modifications in a small number of patients, which may explain the lack of statistical significance of changes at the level of the entire group. Furthermore, we were able to assess only one histone mark (H3K4me3), and it may be hypothesized that other epigenetic modifications (such as H3K4me1, H3K27Ac, H3K9me3) could also contribute to the effects of BCG instillations at these time points.

Previous research showed that IL-1β is a key mediator controlling the induction of TI by BCG vaccination and that IL-1β was highly correlated with a reduction of viremia after BCG vaccination in an experimental model of viral infection in humans 26 . Furthermore, IL-1β signaling promotes glycolysis and proliferation of HSPCs in β-glucan trained mice 30, 31 .

Functional reprogramming of HSPCs is also a hallmark of BCG-induced trained immunity in humans 39 and mice 38 . Our data confirmed that IL-1β is also a key mediator of the TI phenotype induced by BCG instillations in humans. We showed that NMIBC patients treated with BCG have the capacity to produce increased amounts of IL-1β compared to pre-BCG1, and most importantly, this capacity is still present at the start of BCG maintenance 1 and 2. This indicates that circulating monocytes of BCG-treated NMIBC patients, which have proliferated in the bone marrow, have acquired a long-term functional program for increased IL-1β production upon immunological stimulation. Indeed, the increased expression of GBPs and AIM2 in our RNAseq data indicates that the BCG-treated patients had an increased capacity to produce IL-1β, which is controlled by the inflammasome. Our results are in line with previous studies which show that GBPs activate the inflammasome upon sensing cytosolic LPS, and thereby increase IL-1β production 59, 60, 79 . Furthermore, the increase in GBP5 expression reflects data from Lim et al., who showed increased GBP5 expression in non-tumor urothelial tissue three months post-BCG compared to pre-BCG 8 . Finally, we showed that GBP and AIM2 expression was increased in in-vitro TI experiments as well. Thus, BCG seems to increase GBP expression in circulating monocytes (GBP5 also in the bladder 8 ) and GBP-enhanced inflammasome activation may control TI responses both in-vivo and in-vitro.

The GBP-AIM2 inflammasome can also indirectly become activated via the signaling cascade of CGAS-STING-IRF1 62, 80 . CGAS (MB21D1) is a DNA-sensing receptor which activates the STING pathway which is important for the production of pro-inflammatory cytokines and type 1 interferon 81 . Recently, a recombinant BCG overexpressing c-di-AMP (a STING agonist) induced trained immunity and improved antitumor efficacy in animal models of NMIBC 82 . To what extent DNA-sensing by CGAS and AIM2 is involved in TI responses during BCG therapy in NMIBC patients remains to be elucidated. In our study we did not stimulate PBMCs with double-stranded DNA, so we cannot confirm whether TI responses to DNA stimulation are also increased. If this is indeed the case, one might hypothesize that (tumor-and BCG-derived) DNA-sensing in the bladder microenvironment may be increased and this may be one of the anti-tumor immune mechanisms of BCG instillations.

We already previously reported the association of genetic variation in ATG2B 29 and IL18 83 with clinical response after BCG therapy. Our exome chip analysis identified associations for additional autophagy genes and other genes that are known to affect TI induction directly at the epigenetic level (i.e., EHMT2 35 ) and the level of hematopoiesis (i.e., HNF1B 39 ). These genetic findings suggest a connection between TI and clinical efficacy of BCG in NMIBC.

However, the genetic associations have not been replicated in independent NMIBC cohorts. Also note that due to the small study sample and incomplete coverage of DNA variation via the exome chip, negative gene findings have low informative value.

In conclusion, we present comprehensive evidence from in-vivo and ex-vivo studies that BCG instillations in patients with NMIBC induce trained immunity. However, the TI responses induced varied widely between patients and this may also be the substrate of variable clinical responses, although this remains to be investigated in future studies. Most patients showed an augmented long-term systemic innate immune response, which may very well boost antitumor and anti-pathogen immune responses. Thus, exploiting TI responses may be of use to further increase efficacy of BCG therapy.

We set up a prospective observational study in high-risk NMIBC patients (i.e. Ta high-grade or T1 with or without concomitant carcinoma in situ (CIS)) that were scheduled to start BCG therapy at the urology departments of two Dutch academic hospitals (Radboud university medical center Nijmegen and Erasmus MC Rotterdam). Inclusion for this study, named Tribute (acronym for 'TRained Immunity induced by BCG in UroThElial carcinoma'), was initiated in June 2018 and ended in April 2021. Eligibility criteria are given in Supplementary Table 3 . In short, all patients were BCG naive and had primary or recurrent HR-NMIBC, with or without CIS. All patients were free of visible papillary tumor at the start of BCG therapy as determined via re-TURT or negative cystoscopy and/or cytology at most 6 weeks before start of BCG therapy. Patients with tumor stage >cN0 or cM1 were excluded from participation, as well as those with upper urinary tract tumors or another malignancy other than basal cell carcinoma of the skin or prostate cancer under active surveillance. The BCG treatment schedule in this observational study was up to the treating urologist and was based on a standard regimen of a 6-week induction course followed by 3-weekly maintenance courses. Clinical data and followup were collected from the medical files of the participating patients. A baseline questionnaire was used to extract patient information on e.g. smoking and BCG vaccination status and a diary was used to collect data on smoking and co-medication during BCG therapy. The study received ethical approval from the medical research ethics committee Arnhem-Nijmegen (METC number: NL60341.091.17).

Blood for the Tribute study was collected (40-50mL in EDTA tubes) at nine time points during the first year of BCG therapy and was drawn at the outpatient clinic, either directly before the BCG bladder instillation or 5-20 minutes after the instillation. Blood was drawn directly before the BCG bladder instillation at study visits: pre-BCG1, pre-BCG7, pre-BCG10 and pre-BCG13.

Blood was drawn 5 to 20 minutes after applying the BCG instillation at study visits: BCG2, BCG6, BCG9, BCG12 and BCG15 (Fig. 1a) .

Complete blood leukocyte counts were determined in 100uL EDTA whole blood using a Sysmex XN-450 analyzer (Sysmex) within 5 hours after blood draw.

Isolation of PBMCs and purification of monocytes was performed as described in previously in an incubator at 37°C and 5% CO2. After 24 hours the plates were spun for 8 minutes at 1400RPM (Rotina 380R Hettich) and supernatants were collected and stored at -80°C until analysis.

Cytokine concentrations in PBMC supernatants were measured to determine whether cytokine production after non-specific stimulation for 24 hours was altered throughout BCG instillation therapy. These measurements were performed for both duplicates and the mean value of both duplicates was used for data analysis. IL-1β, IL-6, TNF and IL-1 receptor antagonist (IL-1Ra)

were measured in thawed supernatants using commercial ELISA kits (R&D Systems). IL-1Ra was measured because it inhibits the binding and bioactivity of IL-1β 85 , thus we expect IL-1Ra to inhibit TI responses.

Isolated monocyte-enriched suspensions were fixed with 1% formaldehyde (Sigma). Fixed cell suspensions were sonicated using a Bioruptor Pico (Diagenode) for 7 cycles (30s on; 30s off).

For each ChIP, chromatin of 0.5x10^6 cells was incubated with 254 uL dilution buffer, 12uL protease inhibitor cocktail (25x), and 1µg of H3K4me3 antibody (Cell Signaling Technology, Danvers USA) and incubated overnight at 4C with rotation. Protein A/G magnetic beads were washed in dilution buffer with 0.15% SDS and 0.1% BSA, added to the chromatin/antibody mix and rotated for 60 min at 4C. Beads were washed with 500ml buffer for 5 min at 4C with five rounds of washes. After washing, chromatin was eluted using elution buffer for 20 min.

Supernatant was collected, 8 µL 5M NaCl, 2 µL proteinase K were added and samples were incubated on a shaking heat-block for 4 hours at 1000rpm, 65C. This elution procedure was also repeated for chromatin input samples, where 0.5x10^6 cells were used for input as well.

As final step, the DNA was isolated using QIAGEN MinElute PCR purification Kit and eluted in 20 µL elution buffer.

We included the first 7 Tribute patients for ChIP-qPCR to have access to chromatin from longterm time points (i.e., pre-BCG10). Immunoprecipitated chromatin was used for RT-qPCR analysis. For qPCR analysis the input samples were diluted 25 times and the ChIP samples were diluted 3 times. Primers used in the reaction are listed in Supplementary Table 12 .

Samples were analysed with a comparative Ct method on the StepOne PLUS qPCR machine (Thermo Fisher Scientific) using SYBR green (Invitrogen) in accordance with the manufacturer's instructions.

ChIP-seq was performed for the first 7 patients that were included in Tribute. Illumina library preparation was performed using the KAPA HyperPrep kit (KAPA Biosystems). Library concentration was measured using the KAPA Library Quantification Kit (KAPA Biosystems); library size was determined using the BioAnalyzer High Sensitivity DNA Kit (Agilent).

Sequencing was performed using an Illumina NextSeq500, and 42-bp paired-end reads were generated.

Chromatin Immunoprecipitation and sequencing (ChIP-seq) reads were aligned to human genome assembly hg38 (NCBI version 38) using bwa 86 . BAM files were first filtered to remove the reads with mapping quality less than 15, followed by fragment size modelling. MACS2 was used to call the peaks. Peaks from all samples were merged into a single 'H3K4me3 peaks' bed file and reads per peak were counted using bedtools coverage 87 . Data (reads/peak) were normalized using the R package DESeq2 and then pairwise comparisons were performed.

After isolation, 1 -5 x 10 6 cells of the monocyte-enriched cell suspension were used for RNA extraction. The cells were spun down at 500g for 5 minutes. Supernatant was removed and the cell pellet was resuspended with 500µL RNAlater solution (Thermo Fisher Scientific) and kept at room temperature for 20 minutes. The resuspended solution was stored at -80C until RNA isolation. The cell pellets which were frozen in RNAlater solution were thawed to RT and centrifuged for 5 minutes at 300g. The cell pellets were then dissolved in RLT buffer (Qiagen) for 20 minutes. RNA was isolated using RNeasy Mini Kit (Qiagen) according to manufacturer's protocol, and including DNAse treatment.

RNA sequencing (RNA-seq) was performed by BGI Genomics (Denmark) for 6 patients (all included in ChIP-seq). The DNBseq RNA transcriptome PE101 pipeline was used with 30 million reads per sample. Read length was 100bp. After sequencing, the raw reads were filtered. Data filtering included removing adaptor sequences, contamination and low-quality reads from raw reads. To infer gene expression levels, RNA-seq reads were aligned to hg19 human transcriptome using Bowtie 88 . Quantification of gene expression levels as RPKM was performed using MMSEQ 89 . Reads/transcripts were normalized using DEseq2 and pair-wise comparisons were performed.

Differentially expressed genes were identified using DEseq2 with fold change > 1.33 and pvalue <0.05, with a mean RPKM > 1. For longitudinal comparisons, the outputs of all pair-wise comparisons (e.g. pre-BCG1 vs BCG6, pre-BCG1 vs pre-BCG7) were combined and separated into groups using k-means clustering in MeV 90 . Dynamic H3K4me3 peaks were identified using DEseq2 with change in signal determined as fold change > 1.5 and unadjusted p-value <0.05, with a mean reads/peak > 25 91 92 . For gene ontology, H3K4me3 peaks were assigned to the nearest transcription start site (TSS) within 1Mb using the GREAT tool 93 . Promoter H3K4me3 peaks were identified as those within 5kb from a transcription start site using bedtools. Motif analysis was performed on gene promoters using HOMER findMotifs 94 . Motif enrichment was calculated using the hypergeometric distributions, comparing the input promoter list to a background list of promoters. In addition to the p-value, we used a fold change cut-off of >2 to designate a motif as significantly enriched.

Genes of interest were validated for BCG responsiveness ex-vivo using the standard trained immunity model 72 with a time-resolved transcriptome analysis approach 91 . Briefly, monocytes from 5 healthy volunteers were exposed to BCG ex-vivo for 24h and RNA was collected at baseline, 4h, 24h, day 6 (5 days after removal of stimulus), and day 6 + 4 hours of LPS stimulation. Time-course RNA expression data, as normalized counts, was extracted for GBP1, -2, -3, -4, 5 and AIM2 and plotted as median log2 fold change relative to baseline. Raw data files for ex-vivo RNA sequencing are available at GSE168468. 

Descriptive statistics are presented as median ± range, as indicated in the legend of each figure, unless otherwise stated. The statistical significance of the differences between study visits was evaluated using matched pair two-tailed Wilcoxon signed-rank test. In figures, asterisks denote statistical significance (*, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001).

Statistical analysis was performed in GraphPad PRISM 8. Here, we assumed that a TI phenotype is fully induced two weeks after the first instillation and will be present up to one year after the last BCG instillation. The other BCG-treated patients were either partially exposed during the outcome assessment period or exposed prior to the outcome assessment period.

Statistical analyses entailed the description of characteristics in cross tables for comparison of groups that differed in their exposure to BCG. Univariable logistic regression analysis was performed to generate odds ratios (OR) and 95% confidence intervals (CIs). Multivariable logistic regression analysis was performed to allow for adjustment of factors that are known to affect the incidence of respiratory infections and may be associated with BCG exposure or NMIBC patient status. More specifically, we considered older age, male sex, lack of flu vaccination (in the year 2018 and/or 2019), smoking cigarettes, and history of chronic lung disease (i.e. asthma, COPD, chronic bronchitis, lung emphysema) as potentially relevant respiratory infection-risk increasing covariables.

BCG-treated NMIBC patients were included from the previously described Nijmegen Bladder Cancer Study 76 . NBCS patients were genotyped using the Illumina HumanExome BeadChip ('exome chip') containing over 240,000 exonic markers of which ~90% with minor allele frequency (MAF) <5%. Variants were called and cleaned according to 97 

Results of the ChIP-PCR analysis for H3K4me3 modifications. H3K4me3 signal was calculated as %input and the fold change (of %input) is shown for pre-BCG7 and pre-BCG10 compared to pre-BCG1. The fold change is shown for 6 genomic regions in the promoter of TNF, IL1B

and IL6 (the primers are displayed in Supplementary Table 13 Comparisons were made between NMIBC patients that were considered BCG exposed, partially BCG exposed, and unexposed to BCG (see Methods). Comparisons in H3K4me3 signal were made between pre-BCG1 and any of the measured post-BCG time points (i.e., BCG6, pre-BCG7, BCG9 and pre-BCG10). Candidate trained immunity genes are indicated in the final column.

The mechanism/pathway based on which this gene is important for trained immunity is displayed in the second column and the reference, if available, is shown in the third column. The clinical outcomes which were assessed were recurrence free survival and progression free survival. Each sheet of the excel file shows the candidate trained immunity genes and their association with RFS or PFS after 5 or more and 7 or more BCG instillations.

RFS=recurrence free survival. PFS=progression free survival. Treatment cycle   Week   1  2  3  4  5  6  7  8  9 Week 1 -6 Comparisons were made between NMIBC patients that were considered BCG exposed, partially BCG exposed, and unexposed to BCG (see Methods).

Unadjusted odds ratio (OR) (95% CI)

Total Exposed Partially exposed 

European Association of Urology Guidelines on Non-muscle-invasive Bladder Cancer (Ta, T1, and Carcinoma in Situ)

Intravesical bacillus Calmette-Guérin is superior to mitomycin C in reducing tumour recurrence in high-risk superficial bladder cancer: a meta-analysis of randomized trials

Intravesical bacillus Calmette-Guerin versus mitomycin C for superficial bladder cancer: a formal meta-analysis of comparative studies on recurrence and toxicity

Contemporary Outcomes of Patients with Nonmuscle-Invasive Bladder Cancer Treated with bacillus Calmette-Guérin: Implications for Clinical Trial Design

The effect of age on the efficacy of maintenance bacillus Calmette-Guérin relative to maintenance epirubicin in patients with stage Ta T1 urothelial bladder cancer: results from EORTC genito-urinary group study 30911

Urinary interleukin-2 may predict clinical outcome of intravesical bacillus Calmette-Guérin immunotherapy for carcinoma in situ of the bladder

T-cell subsets required for intravesical BCG immunotherapy for bladder cancer

Immunological Hallmarks for Clinical Response to BCG in Bladder Cancer

Bacillus Calmette-Guerin (BCG) enhances monocyte-and lymphocytemediated bladder tumour cell killing

NK cells are essential for effective BCG immunotherapy

NKG2D is a Key Receptor for Recognition of Bladder Cancer Cells by IL-2-Activated NK Cells and BCG Promotes NK Cell Activation

Role of Th1-stimulating cytokines in bacillus Calmette-Guérin (BCG)-induced macrophage cytotoxicity against mouse bladder cancer MBT-2 cells

Tumor associated macrophages polarization dictates the efficacy of BCG instillation in non-muscle invasive urothelial bladder cancer

The predominance of M2-polarized macrophages in the stroma of low-hypoxic bladder tumors is associated with BCG immunotherapy failure

Autophagy promotes BCG-induced maturation of human dendritic cells

Molecular mechanisms regulating Th1 immune responses

Th1/Th2 Paradigm Extended: Macrophage Polarization as an Unappreciated Pathogen-Driven Escape Mechanism?

Inflammasomes and adaptive immune responses

The role of TNF superfamily members in T-cell function and diseases

NOD2 controls the nature of the inflammatory response and subsequent fate of Mycobacterium tuberculosis and M. bovis BCG in human macrophages

Mycobacterium indicus pranii and Mycobacterium bovis BCG lead to differential macrophage activation in Toll-like receptor-dependent manner

Simultaneous blocking of human Toll-like receptors 2 and 4 suppresses myeloid dendritic cell activation induced by Mycobacterium bovis bacillus Calmette-Guérin peptidoglycan

Deconvolution of the Response to Bacillus Calmette-Guerin Reveals NF-kappaB-Induced Cytokines As Autocrine Mediators of Innate Immunity

Defining trained immunity and its role in health and disease

Lnc-ing Trained Immunity to Chromatin Architecture

CIDEC Is Involved in LPS-Induced Inflammation and Apoptosis in Renal Tubular Epithelial Cells

Identification of lipid droplet-associated proteins in the formation of macrophagederived foam cells using microarrays

Insulin resistance and white adipose tissue inflammation are uncoupled in energetically challenged Fsp27-deficient mice

ATF7IP-Mediated Stabilization of the Histone Methyltransferase SETDB1 Is Essential for Heterochromatin Formation by the HUSH Complex

ATF7IP regulates SETDB1 nuclear localization and increases its ubiquitination

FANCA safeguards interphase and mitosis during hematopoiesis in vivo

Apolipoprotein L2 contains a BH3-like domain but it does not behave as a BH3-only protein

Guanylate-binding proteins promote activation of the AIM2 inflammasome during infection with Francisella novicida

Human GBP1 binds LPS to initiate assembly of a caspase-4 activating platform on cytosolic bacteria

LPS targets host guanylate-binding proteins to the bacterial outer membrane for non-canonical inflammasome activation

Interferon-inducible guanylatebinding proteins at the interface of cell-autonomous immunity and inflammasome activation

The transcription factor IRF1 and guanylate-binding proteins target activation of the AIM2 inflammasome by Francisella infection

The impact of interferon-regulatory factors to macrophage differentiation and polarization into M1 and M2

Effects of IRF1 and IFN-β interaction on the M1 polarization of macrophages and its antitumor function

MaxiK Blockade Selectively Inhibits the Lipopolysaccharide-Induced IκB-α/NF-κB Signaling Pathway in Macrophages

Involvement of a membrane potassium channel in heparan sulphate-induced activation of macrophages

RUFY4 exists as two translationally regulated isoforms, that localize to the mitochondrion in activated macrophages

RUN and FYVE domain-containing protein 4 enhances autophagy and lysosome tethering in response to Interleukin-4

Role of AIM2 inflammasome in inflammatory diseases, cancer and infection

Constitutive Interferon Maintains GBP Expression Required for Release of Bacterial Components Upstream of Pyroptosis and Anti-DNA Responses

In vitro induction of trained immunity in adherent human monocytes

Nonspecific (Heterologous) Protection of Neonatal BCG Vaccination Against Hospitalization Due to Respiratory Infection and Sepsis

BCG Vaccination Enhances the Immunogenicity of Subsequent Influenza Vaccination in Healthy Volunteers: A Randomized, Placebo-Controlled Pilot Study

Non-specific effects of BCG vaccine on viral infections. Clinical microbiology and infection : the official publication of the European Society of

European genome-wide association study identifies SLC14A1 as a new urinary bladder cancer susceptibility gene

Innate immune memory is associated with increased disease-free survival in bladder cancer patients treated with bacillus Calmette-Guérin

BCG vaccination in humans inhibits systemic inflammation in a sexdependent manner

GBP5 Promotes NLRP3 Inflammasome Assembly and Immunity in Mammals. Science

A noncanonical function of cGAMP in inflammasome priming and activation

Molecular mechanisms and cellular functions of cGAS-STING signalling

Recombinant BCG overexpressing a STING agonist elicits trained immunity and improved antitumor efficacy in non-muscle invasive bladder cancer. bioRxiv

Independent Replication of Published Germline Polymorphisms Associated with Urinary Bladder Cancer Prognosis and Treatment Response. Bladder cancer

Long-lasting effects of BCG vaccination on both heterologous Th1/Th17 responses and innate trained immunity

Inhibition of interleukin 1 (IL-1) binding and bioactivity in vitro and modulation of acute inflammation in vivo by IL-1 receptor antagonist and anti-IL-1 receptor monoclonal antibody

Fast and accurate short read alignment with Burrows-Wheeler transform

BEDTools: a flexible suite of utilities for comparing genomic features

Aligning short sequencing reads with Bowtie

Haplotype and isoform specific expression estimation using multi-mapping RNA-seq reads

RNA-Seq analysis in MeV

β-Glucan Reverses the Epigenetic State of LPS-Induced Immunological Tolerance

KEGG: kyoto encyclopedia of genes and genomes

GREAT improves functional interpretation of cis-regulatory regions

Simple combinations of lineage-determining transcription factors prime cisregulatory elements required for macrophage and B cell identities

Insight into bladder cancer care: study protocol of a large nationwide prospective cohort study (BlaZIB)

The UroLife study: protocol for a Dutch prospective cohort on lifestyle habits in relation to non-muscle-invasive bladder cancer prognosis and health-related quality of life

Illumina human exome genotyping array clustering and quality control

Kernel machine SNP-set analysis for censored survival outcomes in genome-wide association studies

Prognostic relevance of urinary bladder cancer susceptibility loci

We thank all participants of Tribute, UroLife, BlaZIB and NBCS for their participation. We 

 Table 2 -Results of the association analysis of exome chip variants in trained immunity candidate genes and clinical outcome in patients of the Nijmegen Bladder Cancer Study (NBCS).Depicted are the trained immunity candidate genes (Supplementary Table 5