Establishment and Characterization of Free-Floating 3D Macrophage Programming Model in the Presence of Cancer Cell Spheroids

Reprogramming of tumor-associated macrophages (TAMs) is a promising strategy for cancer immunotherapy. Several studies have shown that cancer cells induce/support the formation of immunosuppressive TAMs phenotypes. However, the specific factors that orchestrate this immunosuppressive process are unknown or poorly studied. In vivo studies are expensive, complex, and ethically constrained. Therefore, 3D cell interaction models could become a unique framework for the identification of important TAMs programming factors. In this study, we have established and characterized a new in vitro 3D model for macrophage programming in the presence of cancer cell spheroids. First, it was demonstrated that the profile of cytokines, chemokines, and surface markers of 3D-cultured macrophages did not differ conceptually from monolayer-cultured M1 and M2-programmed macrophages. Second, the possibility of reprogramming macrophages in 3D conditions was investigated. In total, the dynamic changes in 6 surface markers, 11 cytokines, and 22 chemokines were analyzed upon macrophage programming (M1 and M2) and reprogramming (M1→M2 and M2→M1). According to the findings, the reprogramming resulted in a mixed macrophage phenotype that expressed both immunosuppressive and anti-cancer immunostimulatory features. Third, cancer cell spheroids were shown to stimulate the production of immunosuppressive M2 markers as well as pro-tumor cytokines and chemokines. In summary, the newly developed 3D model of cancer cell spheroid/macrophage co-culture under free-floating conditions can be used for studies on macrophage plasticity and for the development of targeted cancer immunotherapy.


Introduction
Tumor-associated macrophages (TAMs) are a major leukocyte population in the tumor microenvironment (TME). TAMs support tumor progression by inhibiting the antitumor immune response, induction of tumor growth, and angiogenesis [1]. Conditionally, macrophages are divided into classically activated macrophages (M1) and alternatively activated macrophages (AAM or M2). High infiltration of the immunostimulatory M1 phenotype is associated with a Th1-dominant anti-tumor immune response and, therefore, a favorable prognosis [2]. M1 can be polarized upon stimulation with interferon-γ (IFNγ) and Toll-like receptor (TLR) agonists such as lipopolysaccharide (LPS) and synthetic lipopeptide Pam3SCK4 [3,4]; this state can be characterized by highly expressed CD38 [5], MHC II, and inducible nitric oxide synthase (iNOS), as well as by the secretion of large amounts of interleukins (IL)-1β, IL-6, IL-12, and tumor necrosis factor-α (TNFα) [5][6][7]. The immunosuppressive M2 phenotype can be induced by the anti-inflammatory cytokines IL-4, IL-10, and IL-13; this state is characterized by highly expressed arginase-1 (Arg1), scavenger receptor CD163, and mannose receptor CD206 [8,9]. A high M2-to-M1 ratio is associated with a poor cancer prognosis. 2 of 32 Macrophage reprogramming toward the M1 phenotype is a promising immunotherapy approach for cancer treatment [10]. Reprogramming can be achieved using proinflammatory cytokines, such as TNFα, IL-12, and IFNγ [11][12][13][14]. In previous studies, we reported that the alphavirus-based IFNγ therapy showed a significant decrease in the population of cells bearing myeloid cell markers CD11b, CD38, and CD206 in a murine triple-negative breast cancer model [15]. Even though in vivo therapy reduces the volume of the tumor and increases the M1-to-M2 ratio, the mechanism of macrophage polarization and the factors orchestrating macrophage phenotypes remain unclear.
Traditionally used cell co-cultivation in monolayers does not represent tumor architecture, cell-cell and cell-matrix contacts, or nutrient and oxygen gradients, and therefore is insufficiently appropriate for cancer cell-macrophage interaction studies [16]. Furthermore, passaging/transferring macrophages is practically impossible after differentiation in monolayers, such manipulations can traumatize them and reduce their viability. This problem can be solved using 3D approaches, such as the spheroid model [17]. Spheroid formation using round-bottom ultra-low attachment (ULA) wells, agarose-coated microwells, and the hanging drop technique was reported [18]. The cultivation of macrophages with cancer cell spheroids under free-floating non-adherent conditions may represent a significant advantage over traditional monolayer co-cultures [19].
This study aimed to investigate macrophage interactions with cancer cells and to evaluate reprogramming abilities towards the M1 phenotype in vitro under free-floating 3D conditions. For this purpose, we established a murine mammary cancer 3D model with macrophages at different polarization states (non-treated or M0, M1, and M2). The model was characterized by determining the changes in cell surface marker expression and the secretome composition. We hypothesized that cancer cells switch macrophage polarization towards the M2 phenotype. M2 was expected to be involved in cancer cell migration as well as in the direct promotion of tumor growth. Reprogramming toward the M1 phenotype was expected to reverse these processes. This study represents the first attempt to culture macrophages under 3D free-floating conditions with the comprehensive quantitative characterization of surface markers and cytokine/chemokine profiles. This novel approach allows for the examination of cellular behavior and programming properties within a more physiologically relevant environment, deviating from traditional 2D culture systems. The assessment of the macrophage phenotypes co-cultured with cancer cell spheroids potentially will facilitate the modeling of spatial cellular interactions for therapeutic TME programming.

BMDMs Programming under Free-Floating 3D Conditions
Conventional 2D cultivation of mouse bone marrow-derived macrophages (BMDMs) does not represent in vivo cell-cell interactions and architecture [16,18]. Furthermore, the detachment of 2D (monolayer) polarized macrophage traumatizes cells because standard trypsinization techniques are not applicable to macrophage cultures. In this study, we have investigated the ability of BMDMs to polarize under 3D free-floating conditions. To represent free-floating conditions, the macrophages were detached and plated in a 96-well ultra-low attachment plate (2.9 × 10 4 cells per well) to achieve an undifferentiated M0 phenotype. A total of 50 ng/mL IFNγ and 100 ng/mL TLR2/1 ligand Pam3SCK4 were added to achieve the M1 phenotype, and 20 ng/mL IL-4 was used to achieve the M2 phenotype. Changes in the cell morphology were detectable after 48 h: M1 macrophages formed net-like structures, while M2 macrophages formed cell clumps, compared to M0 (Figure 1a). When 72 h had passed after programming, the nitric oxide (NO) and secreted cytokines were quantified in cell media, and flow cytometry analysis of cell markers was performed by cell staining with the respective antibodies (Figure 1b-d). BMDMs programming under free-floating 3D conditions. BMDMs were plated in a 96well black round bottom ultra-low attachment plate in a medium containing the following programming factors: 50 ng/mL recombinant mouse IFNγ and 100 ng/mL Pam3SCK4 to achieve M1 phenotype or with 20 ng/mL IL-4 to achieve M2 phenotype. M0 represents the initial BMDM phenotype.  BMDMs programming under free-floating 3D conditions. BMDMs were plated in a 96-well black round bottom ultra-low attachment plate in a medium containing the following programming factors: 50 ng/mL recombinant mouse IFNγ and 100 ng/mL Pam3SCK4 to achieve M1 phenotype or with 20 ng/mL IL-4 to achieve M2 phenotype. M0 represents the initial BMDM phenotype. (a) Bright-field microscopy of M0/M1/M2 48 h after activation. (b) Nitric oxide test 72 h after activation (n = 3). (c) Flow cytometry determined marker expression of 3D polarized macrophages 72 h after activation; data are presented as % of all cells. Flow cytometry analysis showing representative staining of CD38 on the x-axis and CD206 on the y-axis. (d) The concentration of cytokines in the medium of 3D polarized macrophages 72 h after activation was determined using ELISA or Luminex assay. Medium-RPMI-1640 medium supplemented with 10% FBS, 50 U/mL penicillin, 50 µg/mL streptomycin, 2 mM L-glutamine, and 10% L929-conditioned medium (CM). Data are shown as the mean of two experiments ± SEM (n = 2, each experiment is a pool of 4 biological replicates). * p < 0.05; ** p < 0.01; *** p < 0.001; ns-non-significant. M1 macrophages were found to express higher levels of CD38 (96%), secrete NO (230 µM), and a high amount of the following pro-inflammatory cytokines: IL-12 (>4.0 ng/mL), TNFα (0.6 ng/mL), and IL-6 (9.8 ng/mL), as described in previous studies of macrophages cultured in monolayers [6,20]. Although M0, M1, and M2 macrophages expressed similar levels of CD11b ( Figure S1a), M1 polarized macrophages had lower amounts of CD11b hi expressing cells (M1 vs. M2 = 73% vs. 93%, p = 0.0039; Figure 1b). The M2 phenotype was identified by an increase in the mannose receptor CD206 (10%) and intracellular arginase-1 hi (Arg1 hi ; 50%). MHC II was upregulated in both M1 and M2 types of macrophages with predominant expression in M1 (Figure 1c).
All macrophages were found to secrete IL-1β, IL-2, and IL-16 ( Figure 1d and Figure S1a). Interestingly, M2 macrophages showed an increased level of IL-1β. Even though IL-1β is conventionally referred to as M1 cytokine [6,21], several studies have shown that M2 cells are also able to secrete it [22][23][24]. These findings suggest that the secretion of IL-1β is not limited to M1 macrophages and that M2 macrophages can also contribute to its production under free-floating conditions. Nevertheless, we suppose that further studies are necessary to assess the IL-1β increase in M2 macrophages under free-floating conditions.
Although both M1 and M2 were able to produce IFNβ, the M1 medium had significantly higher levels of IFNβ (M1 vs. M2 = 2.1 vs. 1.0 ng/mL, p = 0.0070). Previous studies showed that IFNβ has a growth inhibitory effect on cancer cells [25], upregulates the percentage of CD11b low macrophages, promotes the expression of IL-10, and reduces the expression of IL-12 [26]. Also, both M1 and M2 produced similar amounts of GM-CSF compared to undifferentiated M0 macrophages ( Figure S1a). No or very low levels of IFNα and IL-10 were detected similar to previous findings ( Figure S1a) [3,27].
M2 programming provoked macrophage proliferation. M2 macrophages had significantly higher cell numbers compared to M0 (Figure S1a, p = 0.0254, Day 4), as IL-4 is a well-known proliferative agent [28]. In contrast, M1 had significantly lower viability and reduced cell number compared to M0 (p < 0.0001), as the M1 activating factor IFNγ has an antiproliferative effect ( Figure S1a) [29]. Furthermore, NO induces oxidative stress. In summary, the obtained data show the successful programming of BMDMs under free-floating conditions.

BMDMs Reprogramming under Free-Floating 3D Conditions
The reprogramming of M2 macrophages towards the M1 phenotype is a promising cancer immunotherapy method. To perform reprogramming in a 3D free-floating model, BMDMs were activated for 48 h as described above, and then the NO test of cell media was performed to control M1 polarization (Figure 2a,b, Day 2). Next, the cells were washed and an equal number of cells was transferred to a medium without activation factors to eliminate the possibility that both programming and reprogramming factors are affecting cells at the same time. After 24 h, the reprogramming factors were added (Figure 2a M1→M2 and M2→M1). A total of 50 ng/mL IFNγ and 100 ng/mL Pam3SCK4 were used to achieve M2→M1 phenotype, and 20 ng/mL IL-4 was used to achieve M1→M2. Groups containing M1 or M2 macrophages cultivated in the medium without activating factors (w/o) and not reprogrammed were stated as reprogramming-negative controls (Figure 2a   Figure 2b Nitric oxide). Interestingly, (w/o) M2 showed a decrease in the CD206 marker, indicating possible depolarization in the absence of an IL-4 stimulus.
The induction of pro-inflammatory cytokines and metabolic markers (Arg1) may be the result of cellular overstimulation during reprogramming. On the other hand, as macrophages possess high plasticity, we suppose that they might lose phenotype features when the activation factors are removed [30,31]. Thus, (w/o) control macrophages had lower levels of secreted cytokines, except for IL-1β. Similar findings were previously described by Tarique et al. [32].
Although reprogramming under free-floating conditions is possible, both M1→M2 and M2→M1 macrophages achieved mixed phenotypes expressing MHC II, Arg1 hi , CD38, secreting IFNβ, IL-1β, IL-2, and IL-16 (Figure 2c,d and Figure S1b). These findings show that macrophages do not change their phenotype completely following reprogramming, and some features from the previous phenotype are still present. The profiles of all those mentioned above and tested macrophage phenotype markers, cytokines, and chemokines are summarized in Section 2.6, Figure 7. The results of the chemokine analysis are described in Section 2.5: "Chemokines secreted by macrophages and 4T1/GFP spheroids".

BMDMs Programming in the Presence of Cancer Cell Spheroids
It is widely accepted that cancer cells influence/stimulate tumor macrophages to acquire pro-tumoral phenotype. However, the mechanism of such "education" is unclear. Here, we tested the changes in macrophages induced by cancer cells and evaluated their potential programming abilities in the presence of cancer cell spheroids under free-floating conditions. First, the 4T1/GFP breast cancer cell spheroids were generated, and then the non-treated M0 macrophages were added (Figure 3a). BMDMs were incubated with cancer cell spheroids for 24 h before programming to achieve a potential TAM-like phenotype. Then, the programming factors (M1 and M2) were added, as described above. Later, 72 h after programming, morphological changes were recorded; M0 and M2 macrophages induced the migration of cancer cells out of the spheroid, and cancer cell satellite spheroids were formed (Figure 3b).

Chemokines Secreted by Macrophages and 4T1/GFP Spheroids
Chemokines are important for immune cell recruitment and immune control. In this study, macrophage reprogramming resulted in new, previously undescribed macrophage phenotypes, M1→M2 and M2→M1. Furthermore, we observed cancer cell (4T1/GFP) migration out of the spheroids in the presence of M0 and M2, but not M1 macrophages. To explain the cancer cell migration and formation of spheroid satellites, as well as to further characterize the reprogrammed macrophage phenotypes, we assessed the secreted chemokine profiles under 3D free-floating conditions. First, the chemokines secreted by cancer cell spheroids without macrophages and activation factors were determined ( Figure 5a). The 4T1/GFP spheroids (S) secreted CCL5 (53 pg/mL) and CXCL5 (290 pg/mL), which are associated with a negative cancer prognosis [34,35]. The spheroids also secreted CXCL16 (181 pg/mL), which was also described in previous studies and is considered a positive prognosis [36]. S-spheroid; medium-RPMI-1640 medium containing 10% FBS, 50 U/mL penicillin, 50 µg/mL streptomycin, 2 mM L-glutamine, and 10% L929-CM. Data are shown as the mean of two experiments ± SEM (n = 2, each experiment is a pool of 4 biological replicates). * p < 0.05; ** p < 0.01; *** p < 0.001; ns-non-significant.
Continuing the characterization of programmed macrophages, the chemokines secreted by M2 macrophages in 3D monoculture (M2) and co-culture with cancer cell spheroids (S + M2) were determined (Figure 5b). M2 secreted an increased amount of CCL1 (2.7 pg/mL) as well as the following IL-4-induced chemokines: CCL17 (0.3 ng/mL), CCL22 (1.5 ng/mL), and CCL24 (21 ng/mL). CCL17 and CCL22 attract type II T helpers (Th2) and regulatory T cells (Tregs); thus, these cytokines are associated with immunosuppressive TME and poor prognosis [40]. Macrophage production of CCL22 was also observed in human ovarian cancer [41]. Contrary to the observations in the M1 macrophage group, the CCL12 level was high only in the presence of cancer cell spheroids (M2: 0.8 ng/mL; S + M2: 3.4 ng/mL, p = 0.0004). CCL12 was also upregulated in M2-like MHC II low TAMs [39] and in the presence of 4T1-CM [33], suggesting that the established 3D co-culture model reliably represents the properties of TAMs.

The Influence of Cancer Cell Spheroids on BMDMs
This study represents a comprehensive evaluation of cytokine, chemokine, and surface marker profiles of macrophages co-cultured with cancer cell spheroids, as well as an assessment of their reprogramming potential. In this Section, we have highlighted the data demonstrating the influence of 4T1/GFP cancer cell spheroids on macrophage surface/intracellular markers and secretome profiles. Data are shown as the mean of two experiments ± SEM (n = 2, each experiment is a pool of 4 biological replicates). * p < 0.05; ** p < 0.01; *** p < 0.001; ns-non-significant.

The Influence of Cancer Cell Spheroids on BMDMs
This study represents a comprehensive evaluation of cytokine, chemokine, and surface marker profiles of macrophages co-cultured with cancer cell spheroids, as well as an assessment of their reprogramming potential. In this Section, we have highlighted the data demonstrating the influence of 4T1/GFP cancer cell spheroids on macrophage surface/intracellular markers and secretome profiles. Figure 7a summarizes the obtained surface marker and secretome characteristics of programmed/reprogrammed macrophages in the presence/absence of cancer cell spheroids. The heatmap data visualization presented in Figure 7b allows us to clearly identify the cell signatures that are stably induced by pro- and IL-4, respectively). For example, CD38 and CCL5 are stably expressed by M1 macrophages, including (w/o) M1 and M1→M2, and are slightly affected by the presence of cancer cell spheroids. In contrast, IL-12 and CXCL10 in M1 are highly dependent on the presence of IFNγ/Pam3SCK4 in the medium and are completely downregulated in (w/o) M1.  Markers. How do 4T1/GFP spheroids affect the BMDMs phenotype? After incubation of M0 with cancer cell spheroids (S + M0) for 96 h (Figure 3a, schematic representation), the macrophages demonstrated elevated levels of CD38 (from 5% (M0) to 17% (S + M0), p = 0.0038; Figure 8a, Non-polarized). Although CD38 is a good marker of M1 macrophages in vitro [7], several studies have associated the presence of CD38 in TME with a negative prognosis [43]. Furthermore, in the macrophage-cancer cell co-culture, the levels of macrophage MHC II were downregulated compared to M0 cultivated alone (from 23% to 4.5%, p = 0.0003, Figure 8b). The decrease in MHC II levels after BMDMs cultivation with 4T1 conditioned media was previously reported by Madera et al. [44].
Macrophages reprogrammed in the presence of cancer cell spheroids had decreased levels of CD11b hi (M1→M2: from 94% to 67%, p < 0.0001; M2→M1: from 69% to 40%, p = 0.0116). S + M1→M2 had downregulated levels of MHC II from 68% to 50%, p = 0.0007, and CD38 from 75% to 57%, p = 0.0007, but upregulated levels of CD206 from 13% to 22%, p = 0.0734 (Figure 8a, Reprogramming). We suppose that co-cultivation with cancer cell spheroids reduces the ability of macrophages to induce an immune response via antigen presentation due to a decrease in MHC II levels. Moreover, a decrease in the proportion of cells expressing the integrin CD11b hi can negatively affect the phagocytic activity of macrophages.
M2 macrophages polarized in the presence of spheroids (S + M2) had increased levels of chemokines CXCL1 (from M2 0.14 ng/mL to S + M2 0.45 ng/mL, p = 0.0013) and CCL12 (from 0.8 to 3.4 ng/mL, p = 0.0004). Increased levels of CX3CL1 (from 75 to 115 pg/mL, p = 0.0056) and CCL11 (from 13 to 20 pg/mL, p = 0.0077) were also observed in S + M2 ( Figure S5a), which can be explained by the interaction of 4T1/GFP cancer cells with the M2 activation factor IL-4 ( Figure 5a). M1 macrophages cultured without activation factors (w/o) after interaction with the cancer cell spheroid showed increased CCL20 (from (w/o) M1 3 to S + (w/o) M1 15 pg/mL, p = 0.0013) and CXCL1 (from 0.6 to 1.5 ng/mL, p = 0.0032) levels, decreased CCL3 (from 1.5 to 0.2 ng/mL, p = 0.0205) and CXCL13 (from 0.20 to 0.11 ng/mL, p = 0.0144) ( Figure S5b). A decrease in CCL3 gene expression was also detected in macrophages after co-culture with ovarian cancer cells [47]. On the other hand, in colorectal cancer, CCL20 secreted by TAMs is associated with the recruitment of Tregs [50]. S + (w/o) M2 macrophages showed increased levels of CXCL1 (from (w/o) M2 0.13 ng/mL to S + (w/o) M2 0.32 ng/mL, p = 0.0175) and CCL24 (from 2.7 to 9 ng/mL, p = 0.0016). It is theoretically possible that a leaky amount of activation factors (IL4, or IFNγ/Pam3SCK4) are present in the (w/o) macrophage environment and may interact with the cancer cell spheroid. However, taking into account the loss of the macrophage phenotypic characteristics, we suppose that the levels of the remaining activation factors are too small to affect chemokine production.
Changes in cell surface markers and secretome profiles after the co-culture of macrophages with cancer cell spheroids are summarized in Table 1. Overall, CXCL1 was elevated in all samples of cancer cell spheroids. However, it is unclear whether the CXCL1 is elevated due to macrophage-spheroid interactions or due to the spheroid response to IFNγ/Pam3SCK4. Yoshimura et al. described that GM-CSF induces macrophage CXCL1 and CCL17 production [33]. Since spheroids secrete a small amount of GM-CSF in the absence of activation factors and higher amounts in the presence of activation factors ( Figure S2), they can induce macrophage expression of CXCL1 and CCL17.

BMDMs Do Not Affect the Spheroid Size
The size of the cancer cell spheroids was measured using fluorometry and by counting total cell fluorescence in microscopy images. Relative spheroid sizes were determined using flow cytometry (total cell number). The results of all three methods were similar, supporting the same tendency (Figure 9a). The macrophage migration to the 4T1/GFP cell culture medium was used as a control. Data are shown as mean ± SEM. * p < 0.05; ** p < 0.01; *** p < 0.001; ns-non-significant.
Spheroids cultivated with M1 and M2→M1 macrophages were significantly smaller than the spheroids cultivated with M0 macrophages (Figure 9a, Fluorometry, p = 0.0026 and p < 0.0001, respectively). As IFNγ has a direct antiproliferative effect on cancer cells [51], we used controls that included spheroids with IFNγ/Pam3SCK4 or IL-4 but without macrophages. The spheroids cultivated with IFNγ/Pam3SCK4 were significantly smaller than those without the same factors (Figure 9a, Fluorometry, p < 0.0001). Furthermore, the spheroids cultivated with IFNγ/Pam3SCK4 had lower viability than the spheroids cultivated in the absence of factors (p = 0.0136) and spheroids cultivated with IL-4 (p = 0.0151), which can be explained by the toxic effect of these factors (Figure 9b).
The effect of M1 and M2→M1 macrophages on spheroid size in the presence of IFNγ/Pam3SCK4 was not significantly different from the spheroids with IFNγ/Pam3SCK4 factors without macrophages (p > 0.05). Although macrophages were functionally programmed (or reprogrammed) to express pro-inflammatory markers and cytokines, they were not capable of decreasing the size of the cancer cell spheroids in this model.

4T1 Cells and Macrophages Migrate toward Each Other
To study cell migration, 8.0 µm pore inserts were used. The migration of cancer cells to macrophages in different polarization states was determined. The cancer cells were able to migrate to M0 and M2, and migrated significantly less to M1 plated in a standard monolayer culture 24-well plate (M2 vs. M1 p = 0.0171; Figure 9c). Similarly, the formation of satellite cancer cell spheroids was observed in the co-culture with M0 and M2 (Figures  3b and 4b). In addition, M2 macrophages secreted CCL17 and CCL22 (Figure 5b) ligands for the receptor CCR4 expressed by 4T1 cells [48]. A macrophage culture medium was used as a negative control in the migration assay. The cancer cells did not migrate to the culture medium, suggesting that cancer cell migration was induced only in the presence of macrophages. Interestingly, in the literature, 4T1/GFP cells migrated to the M1 macrophage culture medium [52].
Next, we examined the ability of M0, M1, and M2 macrophages to migrate toward 4T1/GFP cancer cells. For this, cancer cells were seeded in a 24-well plate, while BMDMs were added to the inserts with M0, M1, or M2 activation medium. Macrophages were able to migrate to cancer cells through an 8.0 µm pore membrane, but not to the 4T1/GFP culture medium (Figure 9d). M1 and M2 migrated more significantly than M0 (p < 0.001). Xuan et al. showed that CXCL10 induces chemotaxis in M1 macrophages [53]. As we found after stimulation of cancer cells with IFNγ/Pam3SCK4, cancer cells begin to secrete CXCL10 (Figure 5a). In addition, Vogel et al. demonstrated that M2 macrophages migrate to CCL5 [27], which is secreted at low levels by cancer cells after stimulation with IL-4 ( Figure 5a).

Discussion
In this study, we have developed and characterized a macrophage programming and reprogramming model in the presence of cancer cell spheroids under 3D free-floating conditions. In total, 6 surface/intracellular markers, 11 cytokines, and 22 chemokines were analyzed, providing a comprehensive characterization of the phenotypic and functional properties of BMDMs. Many studies on macrophage profiling are focused on the specific markers, cytokines, or chemokines [24,[54][55][56][57][58]. The advantage of this study is that it represents the global assessment of macrophage programming features that are monitored in terms of the presence/absence of the respective stimuli. Moreover, for the first time, the reprogramming of BMDMs co-cultured with cancer cell spheroids under free-floating conditions was characterized and the respective changes were determined.
A few studies have shown that macrophages can be reprogrammed in vitro [32,59], and some studies have described reprogramming in vivo [10,58,60,61]. Although the principal possibilities of macrophage reprogramming have been partially investigated, no studies that explain the reprogramming process in detail have been conducted. The majority of conclusions are based on the presence of a few markers and functional features. It is insufficient to draw conclusions regarding the cells with high plasticity. In this study, the phenotypes of M1→M2 and M2→M1 reprogrammed macrophages were investigated by analyzing the surface markers, amounts of secreted cytokines and chemokines. It was concluded that reprogramming leads to a mixed phenotype expressing both M1 and M2 markers. M1→M2 continued to express the M1 marker CD38 and the respective chemokines CXCL1, CXCL16, CCL5, and CCL12 (Figures 4c and 6). On the other hand, M2→M1 continued to express the M2 characteristic marker Arg1 hi and the chemokines CCL17 and CCL24 (Figures 4c and 6). Practically, this signature and other features should be considered in the development of reprogramming-based therapy.
The tumor microenvironment represents a complex and dynamic system where the factors produced by the cancer cells generate an immunosuppressive "cold" state of the immune cells [62,63]. Many aspects of cancer-related immune suppression remain unclear. We used a 3D free-floating model to investigate how the murine mammary cancer cell spheroids affect the programming of macrophages. Madera et al. demonstrated that macrophage exposure to 4T1-secreted factors enhanced macrophage responses to bacteriaderived TLR agonists [44]. As a result, macrophages had higher levels of nitric oxide and pro-inflammatory cytokines, as well as enhanced phagocytosis activity. Our studies showed that co-culture of M1 macrophages with 4T1 cells indeed enhanced the secretion of proinflammatory factors such as TNFα after stimulation with the TLR2/1 ligand Pam3SCK4 and IFNγ (Figure 8d). At the same time, our data showed that co-cultivation led to the production of metastasis-promoting chemokine CXCL1 in M1 macrophages (Figure 5b) and downregulated the macrophage (M0, M1, and M2) expression of CD11b hi and MHC II (Figure 8a), the surface molecules that are involved in phagocytosis and antigen presentation [64]. Furthermore, non-polarized M0 macrophages showed increased expression of CD38 after co-culture with cancer cell spheroids (Figure 8a), which can be linked to the myeloid-derived suppressor cell (MDSC)-like phenotypic changes [65,66]. Co-culture of M1 macrophages with cancer cell spheroids significantly increased the expression of the M2-associated marker, Arg1 hi (Figure 8a). Previous studies also showed that macrophages co-cultured with non-small cell lung cancer (NSCLC), ovarian cancer, or triple-negative breast cancer (TNBC) cells obtain an immunosuppressive M2-like phenotype [45,47,67]. In summary, our data indicate significant changes in the macrophage phenotype after cultivation with cancer cell spheroids.
Chemokines are chemotactic cytokines that provide signals for cell migration, which is critical for immune cell composition in tumors [68]. Macrophages can be recruited to the TME by CCL2, CCL3, CCL4, CCL5, CCL7, CCL8, and CXCL12 chemokines [69]. Our findings show that 4T1 cell spheroids secrete chemokines CCL5, CXCL5, CXCL16 (Figure 5a), and GM-CSF ( Figure S2). Elevated levels of the chemokine CCL5 and its receptor CCR5 have been found in more than 58% of TNBC and HER-2 + subtype breast cancer clinical tumor samples [70]. In addition, it was previously shown that GM-CSF secreted by 4T1 cells induces macrophage CCL2 production [33]. CCL2, known as monocyte chemoattractant protein 1 (MCP-1), attracts monocytes and was shown to be associated with the spread of breast cancer in the body [71]. In this study, we do not see the enhancement of CCL2 production by macrophages co-cultured with cancer cell spheroids, probably because the macrophage cultivation medium supplemented with L929-CM already contained a significant amount of CCL2 (2.6 ng/mL) ( Figure S4b).
Cancer cells cultured with macrophage activation factors (IFNγ/Pam3SCK4, or IL-4) change their chemokine profile. Upon stimulation with IFNγ/Pam3SCK4, the spheroids secrete increased levels of IL-6, IL-16, CCL5, CXCL10, and CXCL11 ( Figure S2 and Figure 5a); and also produce CCL3, CCL20, CCL22, CXCL1, CXCL5, and CX3CL1 (Figure 5a). On the other hand, upon stimulation with IL-4, the spheroids secrete low but significant levels of CCL5, CCL11, CCL17, and CCL19 (Figure 5a). Interestingly, CCL11, known as the eosinophil chemotactic factor, is much higher induced by IL-4 than by IFNγ/Pam3SCK4. The high CCL11 level was shown to increase the proportion of immunosuppressive CD4 + CD25 + Foxp3 + Tregs in a breast cancer model [72], indicating the pro-tumoral effect of the IL-4/CCL11 interaction. In another study, the expression of IFNγ-induced chemokine CXCL10 in the tumor negatively correlated with tumor size and microvessel density (MVD) but positively correlated with the number of infiltrating CD8 + cells [73]. CXCL10 elicits strong inflammatory effects and is also known to recruit the tumor antigenspecific CD8 + T cells into tumors; however, its stability and downstream functions in the tumor environment are not clear [74]. Therefore, IFNγ/CXCL10 can be linked to the antitumor immune response. Other IFNγ-induced chemokines CCL5, CCL20, and CXCL1 were shown to be related to cancer progression and poor prognosis [34,75,76]. The association of inflammatory chemokines with different prognoses is confusing. The chemokines represent a complex network where one group of chemokines may contribute to acute inflammation, while another one to chronic inflammation [77]. The tumor microenvironment is characterized by chronic inflammation, which does not induce an anti-tumor immune response, but in contrast, it promotes disease progression [78]. The coordination of chemokines and their specific role in cancer cell escape from immune surveillance remain to be elucidated.
The developed 3D free-floating cell system represents a useful tool for the analysis and modeling of the cytokine/chemokine profiles of immune cells in the presence/absence of cancer cells. Figure 7a,b provides a comprehensive summary of the surface marker and secretome profiles of macrophages that were programmed and reprogrammed in the presence or absence of cancer cell spheroids. These profiles offer valuable insights into the phenotypic and functional properties of macrophages under different conditions. Potentially, the model allows generating more complex heterotypic cultures with macrophages, T-cells, NK-cells, fibroblasts, etc., resembling the native tumor composition. Simultaneous coordination of different cytokine and chemokine levels is critical for the therapeutic programming of immune cells.
In the established model, macrophages stimulated the migration of 4T1/GFP cancer cells out of the spheroid. In the M2 group, the formation of satellite GFP-positive spheroids was observed. Cancer cells were found to migrate to M0 and M2 macrophages, but not to M1, whereas M2 and M1 macrophages migrated to cancer cells (Figure 9c,d). Thus, the reciprocal migration of M2 macrophages and cancer cells possibly results in satellite spheroid formation. Vogel et al. [27] showed that M2 macrophages migrate to CCL5, which, as we have found, is released by cancer cells following IL-4 stimulation. Therefore, we suppose that the satellite spheroids are formed in a CCL5-dependent manner. It can be concluded that the developed model represents a new free-floating system for cell migration studies where the potential cancer cell distribution can be studied.
Although macrophages were functionally programmed/reprogrammed and expressed inflammatory markers and cytokines, they did not significantly affect the size of the cancer cell spheroids. Furthermore, IFNγ in combination with the TLR2/1 ligand Pam3SCK4 significantly reduced the size of the spheroid without the presence of macrophages. On the other hand, when these factors are removed, macrophages start to lose their phenotypic features. This makes potential therapeutic reprogramming challenging.
Although the established model represents TME better than conventional 2D cultivation, it has certain limitations. First, the volume of the wells is limited and the model does not supply nutrients. As a result, the developed model is not suitable for long-term cell cultivation. The use of microfluidic devices can potentially solve the problem; however, it can be difficult to accomplish under free-floating conditions. Second, the traditional macrophage cultivation protocol requires the use of a medium supplemented with L929-CM, in which monocyte and eosinophil chemotactic factors CCL2, CCL7, VEGF-A, CCL11, and CXCL1 have been found ( Figure S4b). In this context, the application of synthetic media can be useful. Third, BMDMs represent monocyte-derived macrophages and not tissue-resident macrophages. The molecular profiles of BMDMs and tissue-resident macrophages, as well as TAMs, can differ in response to co-culture with cancer cells [79]. Nevertheless, TAMs in mammary tumors are mostly blood-derived and primarily originate from the recruitment and differentiation of inflammatory CCR2 + Ly6C hi CX3CR1 low monocytes [39]. As TAMs are mainly monocyte-derived, BMDMs are suitable for representing the interactions between cancer cells and macrophages. Finally, it is important to mention that the polarized macrophage population is heterogeneous, which means that non-polarized M0 can still be present, as shown by the percentage of cells expressing specific cell surface markers (Figure 1c). Furthermore, the use of the IFNγ co-stimulatory TLR ligand Pam3SCK4 instead of the more standard LPS also needs to be carefully considered. LPS has a non-specific effect on the immune system, thereby potentially leading to unfavorable consequences upon its administration [4,80,81]. On the other hand, Horuluoglu et al. showed that Pam3SCK4 can stimulate monocytes to obtain an M2-like phenotype [82]. The specific effects of LPS, Pam3SCK4, and other TLR ligands should be further evaluated in the proposed model in terms of macrophage phenotyping.
In summary, a new spheroid-based cancer cell model for macrophage programming under 3D conditions was established and characterized. Unlike matrix-based 3D models, the use of an ultra-low attachment plate allows for the easy manipulation of cells and secreted factors. This model can be used for in vitro immunotherapy studies to develop a tumor micromodel with other cancer-associated cell types, such as fibroblasts or immune cells. The co-culture allowed us to evaluate how cancer cell spheroids affect macrophages in different polarization states, and how macrophages affect spheroid growth. The described model can be used to study TME cell-cell interactions and can be potentially applied as a screening tool for TAM-targeted therapies. Fibroblast-like L929 cells were obtained from ATCC ® (CCL-1™) and cultivated in RPMI-1640 medium supplemented with 10% FBS, 1% PEST, 2 mM L-glutamine at 37 • C, and 5% CO 2 in a humidified atmosphere.

Isolation and Generation of Bone Marrow-Derived Macrophages (BMDMs)
Mouse macrophages were differentiated from bone marrow cells according to an established protocol using an L929 cell-conditioned medium as a source of macrophage colony-stimulating factor (MCSF) [3]. The L929 cell-conditioned medium was prepared as follows: L929 cells were plated in RPMI-1640 with 10% FBS, 2 mM L-glutamine, and 1% PEST in T175 flasks (Cat. No. 156502; Nunc, Thermo Fisher Scientific, Waltham, MA, USA) to achieve 60-70% confluence, the medium was replaced with fresh 30 mL medium, and the cells were incubated for 8-9 days in a humidified 5% CO 2 incubator at 37 • C. After cultivation, the conditioned medium (CM) of L929 cells was collected, centrifuged at 400× g for 10 min, transferred to a new tube, and centrifuged again at 13,000× g for 20 min at 4 • C to remove the cell pellets. The CM of L929 cells was aliquoted and stored at −20 • C until use.
Bone marrow from the femurs and tibia of the hind legs of 8-12 weeks old female BALB/c mice was flushed with RPMI-1640 medium containing 10% FBS under sterile conditions and passed through a 70-µm cell strainer (Cat. No. 800070; Bioswisstec, Schaffhausen, Switzerland). Red blood cells (RBC) were removed by incubation in an ammoniumchloride-potassium (ACK) lysing buffer (Cat. No. A10492-01; Gibco, Life Technologies) for 10 min at room temperature. After centrifugation and washing, the remaining cells were cultured in 10 cm non-tissue culture-treated dishes (Cat. No. 0030702018; Eppendorf, Hamburg, Germany) in RPMI-1640 medium (10% FBS) containing 30% L929-cell-CM. Bone marrow cells (Passage 0) were cultured for 5 days, after which non-adherent cells were washed off using phosphate-buffered saline (PBS), and the adherent macrophages were cultured for an additional day.

Reprogramming of BMDMs
BMDMs were polarized to the M0/M1/M2 phenotype for 48 h, as described above. Then, cells were collected, washed with PBS, counted, and plated 2.87 × 10 4 cells in 100 µL per well in a fresh macrophage medium. After 24 h, 50 µL of macrophage cultivation medium supplemented with 50 ng/mL recombinant mouse IFNγ and 100 ng/mL Pam3SCK4 was added to the M2 pre-polarized macrophages and 20 ng/mL IL-4 was added to the pre-polarized M1. The samples (w/o) M1 and (w/o) M2 received 50 µL of macrophage cultivation medium to keep the total volumes equal in all wells.

Co-Cultivation of Macrophages with the Cancer Cell Spheroid
When 48 h passed after spheroid formation, M0, M1, and M2 macrophages (polarized as described above for 48 h) were washed, counted, and 2.87 × 10 4 cells in 100 µL per well were added to the cancer cell spheroid. Macrophages were then co-cultivated with 4T1/GFP spheroids for 24 h before programming or reprogramming was performed. A total of 50 ng/mL recombinant mouse IFNγ and 100 ng/mL Pam3SCK4 were added to achieve M1 and M2→M1 phenotypes, and 20 ng/mL IL-4 was added to achieve M2 and M1→M2 phenotypes.

Flow Cytometry
For flow cytometry, cells from 4 wells were combined, the suspension was centrifuged at 410× g for 5 min and washed with PBS. Spheroids were treated with 0.05% trypsin solution for 5 min. Then, the cells were washed with PBS supplemented with 10% FBS (PBS-FS10%) and centrifuged at 410× g for 10 min.

Analysis of Cytokines and Chemokines by ELISA and Luminex Assays
Supernatants of macrophages programmed and reprogrammed in the presence and absence of cancer cell spheroids were subjected to cytokine and chemokine analyses using ELISA and Luminex assays. The media was combined from at least 4 wells, centrifuged for 5 min at 500× g, aliquoted, and stored at −20 • C until use. Then, the cytokines were determined using Mouse IL-12 ELISA Kit For the determination of chemokines, Mouse Chemokine Panel 31-Plex (Cat. No. 12009159; Bio-Rad, Hercules, CA, USA) was used. Briefly, the beads were washed twice, and then 50 µL of the samples and standards were added to the beads. To ensure proper mixing, the plate was covered with foil and incubated at RT on a shaker at 850 rpm for 30 min. Following the incubation, the plate was washed three times before the addition of the detection antibodies. The plate was incubated on a shaker in the dark. After 30 min, the plate was washed three times. Then, streptavidin-phycoerythrin (SA-PE) was added to the plate. The plate was incubated in the dark at 850 rpm for 10 min. After the incubation, the plate was washed three times to remove any excess SA-PE. The beads were resuspended in an assay buffer and shaken for 30 s prior to analysis with Luminex 100/200™ and xPONENT 3.1.971.0 software package for the data acquisition system (Luminex Corporation, Austin, TX, USA).

Nitric Oxide (NO) Assay
A nitric oxide assay kit was used (Cat. No. EMSNO; Invitrogen) to determine the level of NO in the cell culture medium. NO is measured by the determination of nitrites. Briefly, 50 µL of the cell culture medium was collected from each well, centrifuged, and used for NO quantification. The standards provided in the kit were utilized to generate a standard curve and quantify the nitrites. A spectrophotometer was employed to measure optical density at 540 nm.

Migration Assay
The 8.0 µm pore polycarbonate membrane inserts (Cat. No. 3422; Corning, Life Sciences) were used to investigate the migration of cancer cells and the differently activated macrophages. To investigate the migration of cancer cells towards macrophages, BMDMs were plated on the bottom of the 24-well plate (Cat. No. 30024; SPL Life Sciences, Pocheonsi, Gyeonggi-do, Republic of Korea). When the macrophages reached confluency, they were programmed towards the M0/M1/M2 phenotype, as described above. After 4 h, 4 × 10 4 4T1/GFP cells in 300 µL complete medium were added to the upper inserts. Then, the inserts were placed into wells containing the differently activated macrophages.
To investigate the migration of BMDMs towards cancer cells, 5 × 10 4 4T1/GFP cells were plated on the bottom of the 24-well plate. After 24 h, the BMDMs were stained with CellTracker™ CM-DiI Dye (Cat. No. C7001; Invitrogen), and 300 µL of 7.5 × 10 4 cells were added into the upper inserts. The inserts were then inserted into the wells.
After 18 h, the non-migrated cells were removed using a cotton swab. Migrated cells localized on the opposite membrane site were counted using a Leica DM-IL fluorescent microscope. Cell migration against media (4T1/GFP, M0, M1, and M2 media, respectively) was used as a negative control.

Determination of Cancer Cell Spheroid Growth
Fluorometry. The spheroid growth was measured by a GFP fluorometry assay using Victor3V 1420-040 Multilabel HTS Counter (PerkinElmer): emission filter 485 nm, detection filter 535 nm.
Microscopy. Fluorescent microscopy images were obtained using a Leica DM-IL microscope. Total fluorescence was calculated by ImageJ using the following formula: Total Fluorescence = Area Integrated Intensity − (Area × Background Average Fluorescence) (1) Flow cytometry. The number of GFP-positive cells per sample was determined by a FACSAria II BD Hardware flow cytometer using FACSDiva 6.1.3 Software.

Statistical Analysis
Statistical analysis was performed using the GraphPad Prism 8.02 software. Data were expressed as mean ± standard error of the mean (SEM). Statistical analysis was performed using the Student's t-test and one-way ANOVA. An unpaired one-tailed Student's t-test was used to compare groups. Each experimental group in this study consisted of two biological replicates. To increase the statistical power, each biological replicate was created by pooling four independently treated samples, which were processed under the same experimental conditions. p < 0.05 was considered statistically significant (* p < 0.05; ** p < 0.01; *** p < 0.001; ns-non-significant). Statistical analysis of larger groups (n ≥ 3) was performed using one-way ANOVA.
Heatmaps were generated using a normalization approach depending on the type of data being analyzed. For macrophage markers, the scale was based on the percentage of maximal expression, ranging from 0% to 100%. In the case of cytokines and chemokines, a normalized scale ranges from 0 to 1 and represents the relative amount of secreted cytokines.

Conclusions
In this study, we created a 3D macrophage programming model in the presence of 4T1 breast cancer cell spheroids. The model was characterized using surface markers, secreted cytokines, and chemokines. Unlike many previous studies that focused on specific markers or cytokines, this research offers a global assessment of macrophage signatures.
Macrophage reprogramming was performed, and it was established that reprogramming causes a heterogeneous phenotype that expresses both M1 and M2 markers. In a 3D spheroid model, cancer cells were demonstrated to alter macrophage characteristics, promoting M2-like phenotype features. This novel cancer cell spheroid-based model for macrophage programming in 3D settings can be exploited for in vitro immunotherapy studies. It offers opportunities for more complex heterotypic cultures resembling the native tumor composition and facilitates studies on immune cell programming in a clinically relevant context. Although the established model represents a step forward in mimicking the tumor microenvironment, it has certain limitations. Addressing these limitations, such as nutrient supply and the use of synthetic media, will further enhance the relevance and applicability of the model.