Cytokine/Chemokine Release Patterns and Transcriptomic Profiles of LPS/IFNγ-Activated Human Macrophages Differentiated with Heat-Killed Mycobacterium obuense, M-CSF, or GM-CSF

Macrophages (Mφs) are instrumental regulators of the immune response whereby they acquire diverse functional phenotypes following their exposure to microenvironmental cues that govern their differentiation from monocytes and their activation. The complexity and diversity of the mycobacterial cell wall have empowered mycobacteria with potent immunomodulatory capacities. A heat-killed (HK) whole-cell preparation of Mycobacterium obuense (M. obuense) has shown promise as an adjunctive immunotherapeutic agent for the treatment of cancer. Moreover, HK M. obuense has been shown to trigger the differentiation of human monocytes into a monocyte-derived macrophage (MDM) type named Mob-MDM. However, the transcriptomic profile and functional properties of Mob-MDMs remain undefined during an activation state. Here, we characterized cytokine/chemokine release patterns and transcriptomic profiles of lipopolysaccharide (LPS)/interferon γ (IFNγ)-activated human MDMs that were differentiated with HK M. obuense (Mob-MDM(LPS/IFNγ)), macrophage colony-stimulating factor M-MDM(LPS/IFNγ)), or granulocyte/macrophage colony-stimulating factor (GM-MDM(LPS/IFNγ)). Mob-MDM(LPS/IFNγ) demonstrated a unique cytokine/chemokine release pattern (interleukin (IL)-10low, IL-12/23p40low, IL-23p19/p40low, chemokine (C-x-C) motif ligand (CXCL)9low) that was distinct from those of M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ). Furthermore, M-MDM(LPS/IFNγ) maintained IL-10 production at significantly higher levels compared to GM-MDM(LPS/IFNγ) and Mob-MDM(LPS/IFNγ) despite being activated with M1-Mφ-activating stimuli. Comparative RNA sequencing analysis pointed to a distinct transcriptome profile for Mob-MDM(LPS/IFNγ) relative to both M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ) that comprised 417 transcripts. Functional gene-set enrichment analysis revealed significant overrepresentation of signaling pathways and biological processes that were uniquely related to Mob-MDM(LPS/IFNγ). Our findings lay a foundation for the potential integration of HK M. obuense in specific cell-based immunotherapeutic modalities such as adoptive transfer of Mφs (Mob-MDM(LPS/IFNγ)) for cancer treatment.

The complexity and diversity of the mycobacterial cell wall have empowered mycobacteria with a potent immunomodulatory capacity [9]. Such mycobacteria-associated immunomodulatory activities have been mainly attributed to various mycobacterial cell wall components, serving as pathogen-associated molecular patterns (PAMPs) that are recognized by specific pathogen recognition receptors (PRRs) expressed on various innate immune cells, including monocytes, and macrophages [10,11]. Over the past 10 years, there has been an increasing interest in exploring the immunotherapeutic potential [12][13][14] and the in vitro immunomodulatory properties of a heat-killed (HK) whole-cell preparation of Mycobacterium obuense (M. obuense), also known as IMM-101 [15][16][17][18]. In a phase I clinical trial, HK M. obuense was shown to be safe and well tolerated when used alone in patients with stage III/IV metastatic melanoma [12]. A phase II clinical trial reported that a combination treatment of HK M. obuense and conventional gemcitabine chemotherapy was safe and resulted in a significant increase in the overall survival of patients with metastatic pancreatic cancer [13]. HK M. obuense is currently being investigated as an immunotherapeutic agent in combination with anti-PD-1 in a phase II trial (NCT03711188) involving patients with advanced melanoma. In addition, an ongoing phase III clinical trial (NCT04442048) is examining whether immunization with HK M. obuense can diminish the frequency of severe respiratory and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)related infections among cancer patients. Our group previously demonstrated the ability of HK M. obuense to modulate the surface expression of different categories of cells surface receptors on whole-blood human monocytes [16], as well as to trigger the differentiation of human monocytes into a nonactivated Mϕ type (Mob-MDM) that is quite distinct from the nonactivated Mϕ phenotypes generated by M-CSF (M-MDM) or GM-CSF (GM-MDM) [18]. Nonactivated Mob-MDMs exhibited significantly increased and spontaneous release of the proinflammatory cytokines IL-6 and TNF-α and chemokine (C-x-C) motif ligand (CXCL) 8, relative to nonactivated M-MDMs and GM-MDMs. Moreover, all three nonactivated Mϕ types did not release spontaneous detectable levels of IL-10, IL-12p40, IL-12p70, and IL-23p19/p40. In the same study, genome-wide gene expression profiling combined with functional enrichment analysis revealed unique gene sets and networks suggestive of a proinflammatory M1-like Mϕ phenotype in Mob-MDMs compared with both M-MDMs and GM-MDMs [18].

Discussion
Mϕ activation is a multifaceted and dynamic mechanism that can generate various Mϕ states with a diverse range of functional outcomes [21,22]. However, the preactivation process whereby monocytes are differentiated into Mϕs plays a dominant role in shaping the final activation state of Mϕs irrespective of the adopted stimulus. Traditionally, GM-CSF and M-CSF have been widely used to induce monocyte-to-Mϕ differentiation and to prime Mϕs toward the nonactivated pre-M1 (GM-MDM) and pre-M2 (M-MDM) Mϕ phenotypes, respectively [23]. Evidence from previous studies has confirmed the immunomodulatory properties and immunotherapeutic potentials of HK whole-cell mycobacteria namely, M. vaccae and M. obuense. More importantly, the safety of both HK mycobacteria is proven and well documented in several studies whereby mild adverse events were noted following the administration of either preparation to cancer [12,13,24] and human immunodeficiency virus (HIV) [25] patients. Thus, HK whole-cell preparations of M. vaccae and M. obuense have a major advantage over Bacillus Calmette-Guérin (BCG) in terms of their favorable safety and tolerability. Despite its successful usage in the treatment of nonmuscle/noninvasive bladder cancer, BCG might cause in certain cases severe and life-threatening adverse events such as local or systemic infections with BCG [26]. It has been well documented from earlier studies that various viable and killed mycobacterial species can alter the activation state of human Mϕs [27,28]. In this manner, data from our previous phenotypic, functional, and transcriptomic studies have indicated that HK M. obuense, an immunomodulator of both the innate and adaptive immune responses, promotes the differentiation of human monocytes into a nonactivated Mϕ type (Mob-MDM) that acquire proinflammatory properties suggestive of an M1-like Mϕ phenotype [18]. In this study, we aimed to gain deeper insights into the cytokine/chemokine secretion patterns and transcriptome profiles of Mob-MDMs, M-MDMs, and GM-MDMs following their LPS/IFNγ-mediated activation.
In the present work and consistent with previous reports, GM-MDM(LPS/IFNγ) produced significantly higher levels of CCL22 [7,29], IL-6 [29,30], IL-12/23p40 [7,29,31], and IL-23p19/p40 [19] and significantly lower levels of IL-10 [29-32] as compared to M-MDM(LPS/IFNγ). On the other hand, a study by Vogel et al. did not find any significant difference in IL-10 secretion between M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ) [8]. Moreover, in sharp contrast to Verreck et al., who reported the absence of detectable levels of IL-12/23p40 in M-MDM(LPS/IFNγ) culture supernatants [19], findings from this study and from previous studies [7,31] have found M-MDM(LPS/IFNγ) to release detectable amounts of IL-12/23p40, although, to a significantly lower level than that released by GM-MDM(LPS/IFNγ). Additionally, previous studies have reported that IL-12p70 [7,32] and TNF-α [30,32] were secreted at significantly higher levels by GM-MDM(LPS/IFNγ) relative to M-MDM(LPS/IFNγ). This study only showed an overall trend, but with no statistical significance, toward increased secretion of both cytokines by GM-MDM(LPS/IFNγ) relative to M-MDM(LPS/IFNγ). Contrary to previous studies [29,32], which found CCL5 to be secreted at significantly higher levels by GM-MDM(LPS/IFNγ) relative to M-MDM(LPS/IFNγ), results from the current study demonstrated the opposite. Hence, the above-mentioned variations in the differential secretion of cytokines and chemokines between M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ) could be related to differences in the activation method employed by different studies, more specifically the period of exposure to LPS and IFNγ and the concentration of each activating stimulus. The profile and level of cytokines/chemokines secreted by Mϕs have been widely adopted as signature markers to discriminate between the two activated M1-and M2-Mϕ phenotypes [33]. Classically activated human M1-Mϕs are typically characterized by their IL-10 low , IL-12/23p40 high and/or IL-12p70 high , and IL-23p19/p40 high phenotype [1,34]. In contrast, alternatively activated M2-Mϕs have an IL-10 high , IL-12/23p40 low and/or IL-12p70 low , and IL-23p19/p40 low phenotype [35]. Several reports utilizing CSF-differentiated human MDMs as in vitro models to study Mϕ have indicated that M-CSF and GM-CSF are crucial determinants for shaping the functional profiles of activated Mϕs [29,36]. In support of this notion, this study has shown that M-MDM(LPS/IFNγ) display a cytokine secretion profile (IL-10 high , IL-12/23p40 low , IL-12p70 low , and IL23p19/p40 neg ) characteristic of the M2-Mϕ phenotype, while GM-MDM(LPS/IFNγ) display a cytokine secretion profile (IL-10 low , IL-12/23p40 high , IL-12p70 high , and IL23p19/p40 high ) characteristic of the M1-Mϕ phenotype. These striking differences were also noted at the transcript level whereby RNA-Seq analysis performed in this study revealed higher expression of IL10 and lower expression of IL12B (encodes for IL-12/23p40) and IL23A (encodes for IL-23p19) in M-MDM(LPS/IFNγ) as compared to GM-MDM(LPS/IFNγ). This is in agreement with previous studies where similar patterns of differential mRNA expression between M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ) were reported for IL10, IL23A [36], and IL12B [31]. Therefore, despite being activated with the M1-activating stimuli, LPS and IFNγ, M-MDMs remain trapped in an M2-Mϕ response mode imposed by M-CSF during the stage of monocyte-to-Mϕ differentiation, and such a mode seems to be quite independent of the nature of the subsequent activating stimuli. A key finding in the current study was the demonstration that Mob-MDM(LPS/IFNγ) exhibit a distinct cytokine/chemokine secretion profile from that of M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ). The unique cytokine secretion profile of Mob-MDM(LPS/IFNγ) (IL-10 low , IL-12/23p40 low , IL-12p70 neg , and IL23p19/p40 low ) did not correlate either with that of M1-Mϕs (IL-12 high ) or with any of those of the M2-Mϕ subtypes, M2a, M2b, M2c, or M2d, whereby high IL-10 production is a common feature among the four of them [37]. Moreover, the distinctiveness of Mob-MDM(LPS/IFNγ) phenotype was further reinforced by their pattern of low CXCL9 production as compared to both M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ). A previous study demonstrated that cell wall extracts from HK M. obuense and M. vaccae can induce the production of IL-12 and TNF-α by human THP-1-derived macrophages. In this study, IL-12 levels induced by cell wall extract of either of the two HK mycobacteria were comparable to those induced by BCG, while TNF-α levels induced by HK M. vaccae cell wall extract were significantly higher than those induced by BCG [38].
In  Table  S2). Mob-MDM(LPS/IFNγ) selective transcripts encompassed a specific group of upregulated proinflammatory chemokines, namely CXCL1, CXCL2, CXCL3, and CXCL5. These four chemokines are known to share a common chemokine binding receptor, chemokine (C-X-C) motif receptor 2 (CXCR2), and they serve as potent chemoattractants for neutrophils [33,39]. On the other hand, a different family of proinflammatory chemokine transcripts (CXCL9, CXCL10, and CXL11) was found to be uniquely downregulated in Mob-MDM(LPS/IFNγ). This family of IFN-inducible chemokines exhibits effective T lymphocyte chemoattractant activities and is capable of binding to the cognate chemokine receptor, CXCR3 [40,41]. Moreover, other signature proinflammatory cytokines (e.g., IFNG, IL1B, and IL36G) were significantly upregulated in Mob-MDM(LPS/IFNγ) versus both M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ). Several studies have highlighted the contribution of IL-36γ (encoded by IL-36G) in polarizing immune cells toward type-1 immune responses, which are pertinent to antitumor immunity [42,43]. It has been previously indicated that human IL-36γ expression is inversely associated with lung and melanoma cancer progression [44]. Transcriptomic data analysis of various immune cell populations purified from human colorectal cancer (CRC) microenvironment has shown that IL36G is expressed at significantly higher levels in M1-Mϕs compared to all other immune cells, including M2-Mϕs [45,46]. Further exploration of the tumor microenvironment has uncovered a correlation between the pattern of Mϕ IL-36γ expression and two positive prognostic markers for CRC which comprise CD4 + central memory T-cell infiltration and augmented density of B cells in tertiary lymphoid structures [46]. IL-36γ has been also reported to confer protection against severe influenza infection through enhancing the survival of mouse lung alveolar Mϕs and restraining the replication of the virus [47]. Alveolar Mϕs isolated from uninfected IL-36γ knockout mice exhibited an M2-like Mϕ phenotype and demonstrated an increased and rapid rate of apoptosis post influenza infection [47]. Taking into account that HK M. obuense is currently being assessed in an ongoing clinical trial (NCT04442048) for its capacity to decrease the occurrence and severity of coronavirus disease 2019 (COVID-19)-related symptoms in cancer patients, it is noteworthy to further investigate whether there is a correlation of serum IL-36γ levels with the clinical outcomes in those patients. Results of our study identified a group of solute carriers that were selectively modulated in Mob-MDM(LPS/IFNγ), whereby SLC27A2, SLC7A11, and SLC51B displayed an increased expression, while SLC8A1, SLC2A8, SLC2A5, and SLC6A9 showed a decreased expression. Key physiological functions for the solute carrier family include: transport of ions, uptake of nutrients, and elimination of cellular waste [48]. However, a limited number of studies have addressed the functions of the above-modulated solute carriers in human Mϕs. For instance, heat-inactivated M. tuberculosis lysate was found to induce SLC7A11 expression in U937-derived human Mϕs [49], while SLC8A1 mediated the release of the proinflammatory cytokine, TNFα, by human lung Mϕs [50]. SIGLEC10 was among the top 10 transcripts of its category (others) that were uniquely and significantly downregulated in Mob-MDM(LPS/IFNγ). The interaction between sialic acid-binding Ig-like lectin 10 (Siglec-10) on the surface of Mϕs with CD24, commonly overexpressed on numerous human cancers, has been previously described to be involved in the suppression of Mϕ-mediated immune responses to cancer [51]. Targeting the CD24-Siglec-10 signaling axis stands as a promising approach for cancer immunotherapy whereby blocking Siglec-10 on human MDMs resulted in a significant increase in the phagocytic capacity of tumor cells [52]. In this context, it would be of interest to evaluate the cytotoxic activity of Mob-MDM(LPS/IFNγ) in vitro against CD24-expressing human cancer cell lines as well as in vivo in humanized mouse cancer models.
RNA-Seq profiling coupled with pathways network analysis predicted nuclear factor-κB (NF-κB) and triggering receptor expressed on myeloid cells 1 (TREM1) signaling pathways to be significantly activated in Mob-MDM(LPS/IFNγ) relative to both M-MDM(LPS/ IFNγ) and GM-MDM(LPS/IFNγ). The transcription factor, NF-κB, is a master regulator of inflammatory gene expression in Mϕs [53], and it also plays a crucial role in maintaining Mϕ viability [54]. It would be reasonable to expect that activation of the NF-κB signaling pathway in Mϕs would induce the release of proinflammatory mediators (e.g., IL-6, inducible nitric oxide synthase (iNOs), and TNF-α) [55], hence promoting the M1-Mϕ phenotype as previously indicated [56,57]. Of note, TREM1 has been previously reported to be an inducer of the M1-Mϕ phenotype [58] and to extend the survival of inflammatory Mϕs [59]. TREM1 activation was shown to initiate a cascade of downstream signaling events that would lead to the secretion of various M1 proinflammatory cytokines [58]. Upstream regulator analysis identified colony-stimulating factor 2 (CSF2; also known as GM-CSF) as the top activated upstream regulator in Mob-(LPS/IFNγ) compared with both M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ). GM-CSF has been previously reported to play a pivotal role in controlling M. tuberculosis infection in human MDMs in vitro, whereby the degree of infection control was positively correlated with GM-CSF secreted levels and with activated GM-CSF signaling pathways in infected Mϕs [60,61]. Further studies are clearly warranted to evaluate the antimycobacterial properties of Mob-MDM(LPS/IFNγ). Taken together, the cytokine/chemokine release pattern of Mob-MDM(LPS/IFNγ), combined with their transcriptome profile, strongly point out to a unique Mϕ phenotype, which is quite distinct from that of M-MDM(LPS/IFNγ) and GM-MDM(LPS/IFNγ).
Previous work by our group indicated that nonactivated Mob-MDMs possess a potent in vitro antitumor activity whereby they displayed a higher cytostatic effect against the human pancreatic cancer cell line, BxPC3, when compared to nonactivated M-MDMs, but their effect was comparable to that of nonactivated GM-MDMs [18]. In fact, HK M. obuense was shown to exhibit marked immunotherapeutic effects when given in combination with chemotherapy, in patients with metastatic pancreatic cancer [13], or in combination with checkpoint inhibitors in patients with advanced melanoma [14]. In view of the aforementioned findings, it is reasonable to suggest that the antitumor activity of nonactivated or LPS/IFNγ-activated Mob-MDMs is independent of their ability to release IL-12 and might be attributed to a unique set of Mob-MDM-related genes (e.g., CSF2 and IL36G). Thus, the potential association of these candidate genes with HK M. obuense-mediated antitumor activity needs to be pursued in future studies. More recently, there has been a strong but renewed interest in novel Mϕ-based approaches in cancer immunotherapy whereby the use of chimeric antigen receptors Mϕs (CAR-Ms), currently under development, has proven to be a promising immunotherapeutic tool against cancer [62,63]. In this manner, results from this study might pave the way for a potential application of HK M. obuense in novel cell-based immunotherapeutic modalities such as the adoptive cell transfer of autologous Mob-MDM(LPS/IFNγ) to patients with solid tumors.

Blood Collection
Human whole-blood samples (150-200 mL) were obtained from anonymous phenotypically healthy adult donors through the blood bank at Nini Hospital. Whole blood was collected in plastic bags containing citrate-phosphate-dextrose-adenine and stored at room temperature for 1 h prior to being processed.

Monocyte Isolation and Generation of MDMs
Whole blood was diluted with an equal volume of complete Roswell Park Memorial Institute (RPMI; Merck, Darmstadt, Germany) 1640 medium (containing 100 U/mL penicillin, 100 µg/mL streptomycin penicillin, and 2 mM L-glutamine; Merck, Darmstadt, Germany). Diluted blood was layered over Ficoll-Paque Plus (GE Healthcare, Buckingham, UK) and centrifuged at 400× g for 40 min at 20 • C. Peripheral blood mononuclear cells (PBMCs) were removed from the interface and washed twice with Dulbecco's phosphate-buffered saline (DPBS; Merck, Darmstadt, Germany) by centrifugation at 200× g for 15 min at 20 • C to get rid of platelets. PBMCs were assessed for viability using the trypan blue exclusion method and were always >95% viable. Later, PBMCs were seeded into 75 cm 2 tissue culture flasks at a density of 1.5 × 10 6 cells/mL of complete RPMI medium and incubated overnight at 37 • C in a 5% CO 2 humidified incubator. Cells were then washed thoroughly with DBPS, and adherent monocytes were permitted to differentiate for 5 days into Mϕs in complete RPMI 1640 medium supplemented with 7.5% heat-inactivated pooled human AB serum (ZenBio, Durham, NC, USA) and with optimal concentrations [7,18,32] of M-CSF (100 ng/mL), GM-CSF (100 ng/mL) (R&D Systems, Abingdon, UK), or 30 µg/mL HK M. obuense (NCTC13365, Immodulon Therapeutics, UK) to generate M-MDMs, GM-MDMs, and Mob-MDMs, respectively. MDM purity, as assessed by flow cytometry, was~80-85% after 5 days of differentiation.

RNA Extraction from MDM
Total RNA was extracted from LPS/IFNγ-activated MDMs with the RNeasy Plus kit (Qiagen, Courtaboeuf, France) following the manufacturer's recommendations. The RNA yield of samples was quantified using the NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). RNA Quality was evaluated by calculating RNA integrity numbers (RINs) through the use of the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer's guidelines. RNA samples had a mean RIN of 8.9 (range = 8.7-9.2).

Library Preparation and RNA-Seq
A total of 800 ng of RNA was subjected to ribosomal RNA depletion using the Low Input RiboMinus Eukaryote System v2 (Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturer's protocol. Samples were then concentrated by vacuum centrifugation, followed by the construction of whole-transcriptome libraries using the Ion Total RNA-Seq Kit v2 (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions. Assessment of the size distribution of samples was carried out by using the Agilent 2100 Bioanalyzer coupled with the 6000 RNA Pico kit. Library concentrations were determined using the Agilent DNA 1000 assay on the 2100 Bioanalyzer according to the manufacturer's procedure. Barcoded whole-transcriptome libraries were then diluted to 66 pM and pooled equally (13 µL each for a total of 26 µL) for sequencing with two samples per template preparation. All template reactions were carried out on the Ion Chef Instrument with the Ion PI Hi-Q Chef 200 kit (Thermo Fisher Scientific, Waltham, MA, USA) based on the manufacturer's protocol. Finally, barcoded samples were loaded onto Ion PI Chips v3 (Thermo Fisher Scientific, Waltham, MA, USA), and sequencing was performed with an Ion Proton sequencer (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions.

RNA-Seq Data Analysis
An average of 36 million reads/sample were sequenced and attained 90% uniformity of coverage. Alignment of reads was performed using Partek flow and a two-step alignment process. Unaligned reads were initially aligned to hg19 (human reference genome) using the STAR algorithm [64]. After the initial alignment, all unaligned reads were then realigned with Bowtie2 version 2.1.0 [65], whereby reads from both alignment steps were then pooled. Transcripts were then quantified using a modified version of the expectationmaximization (E/M) algorithm, as previously defined [66]. A pseudocount was added to all values with reads per kilobase per million (RPKM) < 1.0 followed by log (base 2) transformation and quantile normalization. The fixed-effects model with ANOVA was employed to identify transcripts significantly differentially expressed (p < 0.  Table S1). Analysis was carried out in the R language environment [67]. Select transcripts were also evaluated in pairwise comparisons (e.g., M-MDM(LPS/IFNγ) compared to GM-MDM(LPS/IFNγ)) using t-tests with random variance models. Hierarchical clustering analysis was applied to distinguish, among the differentially expressed genes, those with different patterns of expression among the three LPS/IFNγ-activated MDM types [68,69]. Functional pathways analysis, including gene-set enrichment and gene-gene network analysis, of differentially expressed transcripts was conducted using Ingenuity Pathways Analysis as previously described [70]. Raw data for M-MDM(LPS/IFNγ), GM-MDM(LPS/IFNγ), and Mob-MDM(LPS/IFNγ) will be deposited into the Gene Expression Omnibus (GEO) under the dataset series GSE102492.

Statistical Analysis
Statistical analysis was carried out using GraphPad Prism software (version 6; Graph-Pad Inc., San Diego, CA, USAe). Cytokine/chemokine data are presented as mean values ± standard error of the mean (SEM). One-way ANOVA, followed by Tukey's multiple comparison post-hoc test, was used to analyze cytokine/chemokine data. Values of p < 0.05 denoted the existence of a statistically significant difference between compared groups.