Long-term culture expanded alveolar macrophages restore full epigenetic identity in vivo


 Alveolar macrophages (AM) are tissue resident macrophages of the lung that can be expanded in culture, but it is unknown to what extent culture affects their in vivo identity. Here we show that long-term ex vivo expanded mouse AM (exAM) maintain core AM gene expression but show culture adaptations related to adhesion, metabolism and proliferation. Strikingly, even after several months in culture exAM reacquired full transcriptional and epigenetic identity upon transplantation into the lung and could self-maintain in the natural niche long-term. Changes in open chromatin regions (OCR) observed in culture were fully reversible in transplanted exAM (texAM) and resulted in a gene expression profile indistinguishable from resident AM. Our results demonstrate that long-term proliferation of AM in culture does not compromise cellular identity in vivo. The demonstrated robustness of exAM identity provides new opportunities for mechanistic analysis and highlights the therapeutic potential of ex vivo expanded macrophages.

Alveolar macrophages (AM) are tissue resident macrophages of the lung that can be expanded 29 in culture, but it is unknown to what extent culture affects their in vivo identity. Here we show 30 that long-term ex vivo expanded mouse AM (exAM) maintain core AM gene expression but 31 show culture adaptations related to adhesion, metabolism and proliferation. Strikingly, even 32 after several months in culture exAM reacquired full transcriptional and epigenetic identity 33 upon transplantation into the lung and could self-maintain in the natural niche long-term. Highlights: 45 46 • Specific tissue resident macrophage identity can be maintained through long-term culture 47 48 • Long-term proliferation does not compromise differentiated macrophage identity 49 50 • Epigenetic and transcriptional culture adaptations are fully reversible in vivo 51 52 • Rare example of long-term cultured somatic cells restoring full epigenetic identity in 53 natural niche in vivo 54 55 • Shuttling cells between ex vivo culture and natural niche in vivo provides a valuable 56 system for genetic and biochemical investigation 57 58 Introduction macrophages in that they can expand in culture to large numbers and proliferate for extended 104 periods of time 18,19 . It is therefore of high interest for experimental science and potential 105 cellular therapy applications in cancer immunology, infectious disease and regenerative 106 medicine whether macrophage identity can be maintained through long-term proliferation in 107 culture. 108 109 Here we have addressed these questions by comparing transcriptional and epigenetic identity 110 of mouse alveolar macrophages in long-term culture before and after re-transplantation into 111 the natural niche environment in the lung. We observed that substantial adaptations of 112 alveolar macrophages to the culture environment were transient and did not compromise 113 functional long-term integration into the natural alveolar niche of the lung. Even after long-114 term proliferation in culture, ex vivo expanded alveolar macrophages showed a transcriptional 115 and epigenetic signature after transplantation that was indistinguishable from resident AM 116 that had never transitioned through culture. This indicated that alveolar macrophages 117 sustained sensitivity to critical environmental cues encountered in vivo through long periods 118 of culture. Together our findings establish a unique macrophage culture system of high fidelity 119 for shuttling between ex vivo experimental manipulation and in vivo validation. It also 120 demonstrates that macrophage expansion ex vivo can provide large scale preparations for 121 cellular therapy applications with maintained normal macrophage identity in vivo.

123
Results 124 125 Alveolar macrophages show a massive expansion potential in long-term culture 126 127 We have shown previously that mouse alveolar macrophages obtained from broncho-alveolar 128 lavage (BAL) can be cultured 19 and expanded in GM-CSF containing medium 18 . Here we have 129 further characterized such cultures and named them "exAM" for expanded AM (Fig. 1a). We 130 observed that these cells could be kept in continuous culture for at least 10 months, resulting 131 in 33 theoretical population doublings and an amplification factor of 10 10 (Fig.1b). This 132 correlated with a three-to four-fold higher percentage of cells in S-phase of the cell cycle in 133 exAM cultured for 4 months compared to fresh BAL (Fig.1b), whereas the cell death rate 134 determined by Annexin-V/7-AAD staining was low (Fig. S1a). Furthermore, the exAM cultures 135 could also be taken through freeze/thaw cycles (Fig. S1b) without compromising growth 136 capacity. Together these observations indicated an enormous, potentially unlimited 137 expansion potential of exAM.

139
Expanded alveolar macrophages maintain characteristic phenotype and function in long-140 term culture 141 142 We further analyzed whether exAM cultures maintained typical macrophage phenotypic and 143 functional characteristics in culture. Whereas cells could be kept in culture for at least 10 144 months (Fig. 1a), and possible longer, for practical reasons we performed analyses with 1-4 145 months exAM cultures. These cells had a typical AM phenotypic appearance and Diff-Quik dye 146 staining properties (Fig. 2a). Furthermore, similar to AM freshly isolated by BAL, nearly all cells 147 of the exAM cultures showed typical macrophage acidified lysosomal structures indicated by 148 acridine orange staining (Fig. 2b, Fig S2a). Nearly all cells also showed labeling with Magic Red 149 dye, a sensor of enzymatic activity of the lysosomal protease Cathepsin B (Fig. 2b, Fig S2b).

150
The exAM cultures were also highly active to take up fluorescently labelled latex beads, 151 indicating that they are capable of phagocytosis (Fig. 2c, Fig. S2c). In order to quantify this and 152 demonstrate active phagocytosis, we took advantage of zymosan labelled with the pH 153 sensitive dye pHRhodo that only fluoresces upon acidification of the phagocytosed particles 154 in the lysosomes. FACS quantification of this dye showed the same strong signal in exAM 155 cultures as in freshly harvested AM, that was present only at 37 o C but not at 4 o C (Fig. 2d). This 156 indicated that exAM maintain full ability of AM to phagocytose pathogen-associated material 157 in an active metabolic process that involves the uptake into acidic and enzymatically active 158 lysosomal structures. In order to analyze whether exAM could also mount a macrophage 159 typical immune response, we stimulated exAM cultures with IFNg and E.coli LPS as a TLR4-160 stimulating PAMP and mimetic of infection with a gram-negative bacterial pathogen. As 161 shown in Fig. 2e and 2f, IFNg and LPS stimulation resulted in the production of both reactive 162 oxygen species (ROS) and nitric oxide (NO) in exAM cultures similar to fresh AM. Importantly, 163 exAM were also highly efficient in killing Klebsiella pneumoniae, a common cause of bacterial 164 infection of the lung (Fig 2g). Together, our results indicated that exAM cultures showed many 165 central macrophage characteristics and were functional in mounting a typical macrophage 166 immune response.

168
We further analyzed whether exAM not only maintained general macrophage functions but 169 also a specific AM identity in culture. Flow cytometric analysis demonstrated that exAM 170 maintained typical macrophage surface marker staining including the characteristic AM 171 markers SiglecF and CD11c (Fig. 2h), which were also maintained after passage through a 172 freeze/thaw cycle (Fig. S1b). Furthermore, RNAseq analysis revealed that exAM showed 173 similarly high general expression levels of AM-specific core macrophage genes 3 as fresh BAL 174 AM (showing data from 2 pools of 3 mice each) compared to other tissue-resident 175 macrophage populations such as peritoneal macrophages (Fig. 2i). This was reflected in high 176 expression levels of AM-specific macrophage transcription factors genes 3 , including Pparg, 177 Car4, C/ebpb and Bhlhe41 (Fig. 2j), which have been shown to be important regulators of AM 178 identity and self-renewal 4, 20 21, 22, 23 . Together these results showed that exAM not only 179 maintained macrophage functional characteristics but also AM-specific gene expression 180 through long-term culture.

182
The exAM transcriptome shows substantial adaptations to the culture environment 183 184 The microenvironment has emerged as a potent determinant of macrophage gene 185 expression 2, 5 . For example, the transfer of mouse 9 or human 10 microglia from the brain 186 environment to ex vivo culture results in major changes in gene expression and AM in culture 187 show significant changes in cytokine responsiveness and glucose metabolism compared to in 188 vivo conditions 11 . We therefore wondered whether the culture environment also imposed 189 changes in gene expression on exAM, despite the observed overall conservation of 190 macrophage function and phenotype (Fig. 2). 191 192 Indeed, a general comparison of 2 months cultured exAM to AM in vivo by RNAseq analysis 193 from 2 pools of 3 mice each (Pearson correlation shown in Fig. S3) revealed a substantial 194 number of significant transcriptional changes (FDR<0.05) with 1370 genes up-regulated more 195 than 2-fold and 376 genes upregulated more than 10-fold, whereas 2317 genes were down-196 regulated more than 2-fold and 857 genes were do wn-regulated more than 10-fold (Fig. 3a). 197 198 We hypothesized that some of these changes in gene expression might be related to the 199 expansion capacity in culture. Indeed, consistent with the observations in Fig. 1b we found 200 enrichment of GO term cell cycle gene sets and increased expression of cell cycle regulators 201 (Fig. 3b). We also found enrichment and overexpression of TGFbeta signaling related genes, 202 consistent with reports that AM self-renewal depends on an autocrine TGFbeta signaling 203 loop 24 (Fig. 3c).

205
Although exAM maintained overall similar high expression levels of core AM specific genes 3 206 as AM in vivo, compared to other resident tissue macrophage populations (Fig. 2i), there 207 appeared to be reductions in some genes. To investigate this in more detail, we analyzed the 208 individual expression levels of AM core genes and observed that whereas the majority showed 209 no or low changes in gene expression, several genes were substantially downregulated. These 210 genes included surfactant proteins, lipid metabolism genes and Epcam, a homotypic adhesion 211 molecule of the CAM family ( Fig. 3d, Fig. S4). This is consistent with the loss of the natural 212 alveolar niche environment in culture, where these genes might be obsolete or lack 213 appropriate induction cues.

215
To analyze the nature of gene expression changes in culture in more detail we performed k-216 means clustering. Whereas the majority of expressed genes showed no significant difference 217 between culture and in vivo conditions, we also identified two clusters of up-and down-218 regulated genes, respectively (Fig. 3e). 219 220 Pathway analysis of the up-regulated cluster revealed cytoskeletal, adhesion and migration 221 pathways, consistent with adaptation to the different surface properties of the culture 222 environment (Fig. 3f,g). Furthermore, oxygen response, amino acid and glucose metabolism 223 pathways (Fig. 3f,g and Fig. S4) were upregulated, indicative of the altered oxygen pressure, 224 the increased availability of nutrients and the upregulation of glycolytic pathways in culture 225 that are restricted in vivo 11 . 226 227 Pathway analysis of the down-regulated cluster revealed predominantly immune system 228 pathways, including bacterial and viral response, inflammatory and antigen presentation 229 terms. This is consistent with the fact that the alveolar macrophages were taken from a barrier 230 tissue constantly exposed to environmental microbes to a sterile tissue culture environment 231 (Fig. 3f,g). 232 233 In conclusion, these data indicated that despite the conservation of core AM identity through 234 long-term culture, a substantial number of genes are up-or downregulated in culture 235 reflecting gene expression sensitive to the gained or lost regulatory cues of AM interaction 236 with their microenvironment.

238
Transplanted exAM restores full transcriptional AM identity in vivo 239 240 Based on the observed transcriptional adaptations to the culture environment we wanted to 241 determine whether these changes were stable or reversible. We therefore employed an intra-242 tracheal transplantation protocol to establish whether full transcriptional identity of AM could 243 be restored, once the microenvironmental cues of the natural niche in vivo were provided 244 again. Several protocols have been previously used to transfer different macrophage 245 populations into the alveolar space of the lung. These protocols typically involve 246 transplantation into an empty niche, from which the resident AM have been depleted 247 genetically, chemically or by irradiation 7,21,25,26 . Since these procedures all involve strong 248 disturbance of tissue environment and homeostasis, we considered them not ideal to 249 investigate the effects of the natural niche environment on exAM identity in vivo. Here we 250 therefore established a transplantation protocol into an unmodified niche of wild-type mice. 251 To control for the effects of transplantation itself we compared transplanted fresh AM from 252 BAL (tAM) and transplanted exAM that were for 2 months ex vivo amplified (texAM) to the 253 resident host AM populations (Fig. 4a). Using this protocol, we could demonstrate stable long-254 term contribution of transplanted cells to the AM pool for at least 4 months with highly similar 255 contribution rates of 5-15% for both tAM and texAM (Fig. 4b). Furthermore, FACS analysis 256 demonstrated undistinguishable expression levels between all three populations for the 257 general myeloid and macrophage markers CD11b and CD64 as well as for the AM-specific 258 markers CD11c and SiglecF (Fig. 4c). Together, this showed that despite extended passage 259 time in culture, exAM maintained the capacity of long-term contribution to the AM pool with 260 normal stable phenotype and self-renewal in vivo.

262
In order to further determine whether the global transcriptome could be restored to the in 263 vivo state upon transplantation of exAM from the culture environment to the natural niche in 264 vivo, we investigated the similarity of the different AM populations by RNAseq analysis of 2 265 replicates of pools of 3 mice each. Besides high similarity between replicates, Spearman's 266 correlation analysis further revealed that the differences of exAM to in vivo samples 267 disappeared upon transplantation. texAM showed a high degree of correlation to the other 268 two in vivo samples, both tAM and host AM populations (Fig. 4d, Fig. S3). This was further 269 confirmed by comparison to published data sets from other tissue resident macrophage 270 populations using PCA analysis. As shown in Fig. 4e, texAM were indistinguishable from tAM 271 or resident host AM but different to all other resident macrophage populations. As an 272 indication of the high degree of similarity, our samples were closer to each other than to 273 published AM data sets from another lab generated with a different protocol (Fig. 4e).

275
In a more precise analysis of changes in gene expression between tAM and texAM, we only 276 detected 56 genes that were significantly expressed with more than twofold change 277 (FDR<0.05). This contrasted with 3547 differentially expressed genes between exAM in culture 278 and after transplantation (texAM) (Fig. 4f,g). A similarly high number of 3687 differentially 279 expressed genes was detected when comparing exAM to host AM, which were reduced to 280 only 217 differentially expressed genes after transplantation. When analyzing the genes that 281 differed between texAM and tAM or host AM, only very few genes overlapped ( Fig. 4h), 282 indicating that likely none of the other differences were due to the retention of culture-283 specific genes expression in vivo, but to experimental noise or transplantation-specific effects, 284 as similar differences as for texAM versus host AM were also detected between tAM and host 285 AM (Fig. S5).

287
Finally, the analysis of the core AM signature genes showed that the few genes related to lipid 288 metabolism, surfactant and adhesion molecules that were lost in the culture environment 289 (Fig. 3d) were fully restored after transplantation of exAM into their natural niche (texAM), 290 with nearly no detectable differences to the other in vivo AM samples ( Fig. 4i). Similar 291 observations were made for the few AM-specific genes that had been upregulated in culture 292 ( Fig. 4i).

294
Together, this showed that the substantial transcriptomic adaptations of long-term cultured 295 exAM to the culture environment were fully reversible and could be fully restored to the 296 characteristic signature of resident AM upon exposure to the cues of the natural niche 297 environment in vivo.

299
Epigenetic changes of exAM in culture are fully restored in vivo 300 301 The full restoration of transcriptomic identity upon transplantation of exAM into the natural 302 alveolar space and the complete loss of the transcriptional adaptions to the culture 303 environment opens the question as to the underlying epigenetic regulatory mechanisms. We 304 therefore performed ATACseq analysis to determine the status of chromatin accessibility 305 during these transcriptional changes. As one potential explanation of the complete 306 reversibility of transcriptional changes we considered the possibility of unchanged chromatin 307 accessibility between in vivo and culture conditions. We therefore compared open chromatin 308 regions (OCR) by triplicate ATACseq analysis from pools of 3 biological samples each from 309 cultured exAM and BAL AM in vivo (Fig. S6). This analysis showed a high correlation of 310 replicates but substantial differences between conditions. After bioinformatic pooling of 311 replicates we detected 723 lost and 1504 gained OCR in exAM compared to BAL AM (Fig. 5a), 312 thus excluding the explanation of unchanged chromatin accessibility and highlighting 313 substantial differences between in vivo and culture conditions. 314 315 GO term analysis of the genes associated with these OCR alterations yielded similar terms of 316 adaptation to the cell culture environment as the transcriptomic analysis. Whereas terms for 317 metabolism, cellular proliferation, cell adhesion and migration were enriched, immune 318 response terms were depleted in culture ( Environmental cues can induce long-lasting epigenetic changes despite full reversibility of the 326 transcriptional alterations induced by such signals. For example, we have shown previously 327 that LPS stimulated hematopoietic stem cells (HSC) show transcriptional changes that quickly 328 come back to normal but conserve a long-term epigenetic memory of the stimulus 27 . We 329 therefore wondered whether culture-induced epigenetic changes in exAM were conserved 330 upon re-transfer into the natural environment in vivo. In order to address this question, we 331 used the same transplantation protocol as before (Fig. 4a). Strikingly, culture-specific 332 alterations in chromatin accessibility were fully restored upon transplantation of exAM into 333 the natural alveolar niche. This was indicated by nearly identical OCR in transplanted cells 334 (texAM) compared to resident AM of the recipient (host AM), whereas the comparison of cells 335 in culture (exAM) showed the same difference to host AM as to freshly isolated AM (Fig. 5a). 336 From the about 2500 differential OCR detected between AM in culture and in vivo (exAM 337 versus host AM), only 15 were conserved after re-transplantation of exAM into the natural 338 niche (texAM versus host AM, Fig. 5b). Consistent with this, the comparison of exAM to texAM 339 showed similar differences, whereas BAL and host AM showed a nearly identical OCR profile 340 (Fig. S7). The similarity of all in vivo samples among each other, including transplanted 341 expanded AM (texAM) and the similar strong differences of all in vivo samples to exAM in 342 culture was also indicated by a heatmap of unsupervised cluster analysis (Fig. S8). Importantly, 343 analysis of individual OCR showed that the culture-induced peaks in regulatory regions of cell 344 cycle, adhesion or glycolysis genes were lost again, whereas peaks in immune response genes 345 that were lost in culture were re-established, when cultured cells (exAM) were transplanted 346 into the natural in vivo environment (texAM) (Fig. 5e). 347

348
Analysis of transcription factor binding motifs associated with differential OCR between 349 culture and in vivo samples revealed that besides ATF3 and KLF motifs, binding sites for the 350 core myeloid PU.1, C/EBP and Runx transcription factors were associated with both the 351 differential gained and lost OCR (Fig. S9), suggesting that the changes in chromatin 352 accessibility might occur on pre-existing myeloid enhancer platforms.

354
Finally, analysis of regulatory regions for core AM-specific transcription factors and surface 355 markers identified AM-specific OCR that were not detected in other myeloid cells or other cell 356 types of the immune system, identifying them as markers of epigenetic AM identity (Fig. 5d).

357
For example, AM-specific peaks of the transcription factor Pparg 18 , a key regulator of AM 358 identity, are not detected in other macrophage populations or immune cells, whereas the 359 surface marker CD11c (Itgax) that is expressed in AM and other cell types shows AM-specific 360 and general OCRs. Importantly, these key regulatory regions of AM identity did not change 361 across samples and were equally present in AM in vivo, after expansion in culture (exAM) and 362 upon re-transplantation in vivo (texAM), indicating that epigenetic AM identity was not lost in 363 culture.

365
Together the analysis of epigenetic chromatin accessibility indicates that AM identity was 366 conserved through culture expansion and that adaptations to the culture environment were 367 transient, reversible and fully restored to the in vivo status upon transplantation into the 368 alveolar niche.

370
Expanded alveolar macrophages show full long-term reconstitution of alveolar 371 macrophage niche in vivo 372 373 Given the full restoration of transcriptional and epigenetic identity of exAM in vivo, we 374 wondered whether they were also functionally capable of reconstituting an empty alveolar 375 macrophage niche. Since transplantation into the full niche of WT mice only allows a small 376 contribution of donor cells (Fig. 4a,b), we took advantage of GM-CSFR beta KO (Csf2rb -/-) mice, 377 which are deficient for alveolar macrophages and can be transplanted with different sources 378 of monocytes or macrophages 7,8,25,28,29 . We therefore transplanted CD45.1 exAM that had 379 been amplified 4 months ex vivo into the empty niche of neonatal CD45.2 Csf2rb -/mice, which 380 allows the distinction of donor and recipient cells (Fig. 6a). Interestingly, 8 months after 381 transplantation we detected nearly 100% of CD45.1 donor cells with a SiglecF + /CD11c + AM 382 phenotype in transplanted Csf2rb -/mice, the same proportion as in CD45.1 WT mice of the 383 same age, whereas untransplanted Csf2rb -/mice showed no CD45.1 cells or SiglecF + /CD11c + 384 AM (Fig. 6b,e,f). Transplanted and WT control cells showed the same morphology by 385 brightfield microscopy and staining with Diff-Quik dye of cytospin samples (Fig. 6c). We also 386 obtained the same absolute number of cells from the BAL fluid of transplanted Csf2rb -/mice 387 as from CD45.1 WT control mice (Fig. 6g). This demonstrated that exAM were capable of full 388 long-term reconstitution of an empty AM niche in vivo. A similar full reconstitution was also 389 already seen 3 months after transplantation ( Fig. 6e-h, Fig. S10), demonstrating a similar 390 potency of seeding the niche in vivo as fetal monocytes, the most prominent source of 391 populating the lung in normal development 7, 30 . 392 393 Due to the complete absence of AM, Csf2rb -/mice show a pulmonary alveolar proteinosis 394 (PAP) pathology that is characterized by accumulation of mucus and debris in the alveolar lung 395 fluid. Microscopic inspection (Fig. 6c) or FACS analysis (Fig. 6d,h) demonstrated a strong 396 reduction of debris in the transplanted lungs (Fig. 6d,h) that in some cases showed complete 397 clearance of debris similar to WT control mice (Fig. 6c,d,h). Thus, exAM maintained the key 398 homeostatic function of AM to clear alveolar mucus. 399 400 Together, these results showed the capacity of expanded AM in culture (exAM) to functionally 401 repopulate an empty alveolar niche in vivo and to self-maintain homeostatic numbers long-402 term. 403 404 405 406 Discussion 407 Here we have shown that mouse alveolar macrophages can be maintained through long-term 408 culture and can be expanded massively without losing tissue resident macrophage identity.

409
Although substantial epigenetic and transcriptional culture adaptations occur, they are fully 410 reversible upon transplantation into the natural alveolar niche in vivo (Fig. 7). This is 411 conceptually important as it indicates a robust cell endogenous epigenetic setup that is stable 412 through many cell divisions and provides the flexibility to adapt to different environmental 413 cues without losing identity. In practical terms, expanded alveolar macrophages provide a 414 nearly unlimited source of genetically unmodified and untransformed normal macrophages 415 and a new cell culture system that provides the convenience of a cell line but the ability to 416 quantitatively repopulate the natural niche in vivo with full restoration of epigenetic and 417 transcriptomic identity. The potential to shuttle cells between ex vivo culture and natural 418 niche in vivo provides a unique experimental system for screening and in vivo validation 419 approaches or combined large-scale biochemical and genetic investigation in an 420 untransformed cellular system with direct in vivo relevance.

422
Cell culture has had enormous impact on the progress of biological research ranging from 423 cancer to stem cell research 31 . In immunology it has boosted vaccine development, the 424 generation of monoclonal antibodies and cellular therapy and has been essential for many 425 fundamental discoveries. Despite this general usefulness it has been put into question how 426 close cells in culture are to their in vivo equivalents due to the loss of essential in vivo cues 427 and new culture stimuli not encountered in vivo. For example, mouse and human microglia 428 cells in culture undergo major changes in gene expression and lose in vivo functionality 9, 10, 12 . 429 Similarly, AM show significant changes in cytokine responsiveness and metabolism when put 430 in culture and it has been suggested that cultured macrophages do not reflect macrophage 431 identity in vivo 11 . Indeed, here we also observed substantial changes in gene expression 432 related to the adaptation to the culture environment, including for genes involved in cellular 433 adhesion/migration, proliferation and metabolism. This was in particular the case for genes 434 related to glycolysis, consistent with the high availability of glucose in the culture medium 435 compared to the low levels in the alveolar niche and the shift to glycolytic energy metabolism 436 of AM in culture 11 . We also observed changes in adhesion molecules and cytoskeletal genes, 437 which is consistent with the different surfaces and general physical cues that AM encounter 438 in the culture dish compared to the natural alveolar niche. Furthermore, immune response 439 related gene expression was lost to a large extent in culture, which might be explained by the 440 loss of the stimulation from the microbiome of the lung mucosal surface in the sterile culture 441 environment. These changes were also reflected in the signaling pathways regulating these 442 processes, such as Ras/MAPK, mTOR and TGFbeta pathways for the upregulated gene sets 443 and Il6, TNFalpha, type I and II Interferon and NFkB controlled pathways for the 444 downregulated gene sets (Fig. 3f). Importantly, however, both positive and negative changes 445 in gene expression were reversible and fully restored upon transplantation into the natural 446 alveolar niche in vivo, confirming a potent role of the microenvironment as determinant of 447 macrophage gene expression 2 .

449
The interesting question emerging from these observations is how AM conserve the ability to 450 respond to these different environmental cues. We observed that the changes in gene 451 expression were also accompanied by changes in chromatin accessibility. Differentially gained 452 and lost OCR in culture were similarly associated with adhesion, proliferation, metabolism and 453 immune response terms, respectively (Figs. 5,7). Interestingly, both differentially gained and 454 lost OCR were enriched for binding sites of PU.1, C/EBP and Runx (Fig. S9), transcription factors 455 that determine macrophage identity 32 . It has been suggested that these factors establish a 456 core macrophage enhancer platform on which additional signal induced transcription factors 457 converge 33, 34 . Consistent with this, culture-induced OCR were also enriched for ATF3 and KLF 458 transcription factor binding sites, which are typical signal-induced factors. Together, this 459 suggests that a stable enhancer architecture maintains cellular identity through culture, on 460 which accessory transcription factors can activate environment-specific gene expression.

461
Consistent with this, the expression of core macrophage and alveolar macrophage-specific 462 transcription factors like Runx2, Car4, Pparg, C/ebpb and Bhlhe41 (Fig. 2j) and downstream 463 AM-specific genes (Figs. 2i, 4i) as well as AM specific surface markers (Fig. 2h) were maintained 464 throughout culture. Furthermore, core AM-specific transcription factors and surface markers 465 showed maintained OCR in their regulatory regions (Fig. 5d). This could explain how highly 466 stable AM-specific identity is maintained throughout culture, while allowing a spectrum of 467 responsiveness to environmental cues in vivo and in culture with the corresponding repertoire 468 of adapted gene expression. In more general terms this might also explain the high degree of 469 plasticity of macrophages in responding to a diverse set of stimuli 35 . 470

471
The observed full restoration of normal chromatin accessibility in vivo, despite the significant 472 epigenetic changes acquired in long-term culture, is distinctly different from scenarios of 473 trained immunity. Although in principle both resident 36 and monocyte-derived 21, 26 AM also 474 appear to be capable of trained immunity, the underlying mechanism and inducing signals 475 might be different. For example, we have shown previously that LPS stimulation of HSC 476 induced only short-term transcriptional changes but long-term epigenetic memory through a 477 TRIF and C/EBPb signaling axis 27 . Perhaps, in this case, C/EBPb acts as a pioneering factor 478 establishing new enhancer landing sites for other transcription factors, whereas in the case of 479 AM these enhancer assemblies are already present in vivo and maintained through culture. 480 Beta-Glucan stimulation can also induce trained immunity through epigenetic changes in 481 monocytes in an Akt, mTOR and glycolysis dependent manner 37, 38, 39 . Interestingly, PI3K/Akt 482 and mTOR pathways as well as glycolysis were also induced in exAM (Fig. 3f) but changes in 483 chromatin accessibility were completely restored after re-transplantation in vivo. This 484 indicates that these pathways are necessary but not sufficient for the induction of epigenetic 485 memory in myelo-monocytic cells. It has also been suggested that the induction of FOS/JUN 486 (AP-1) transcription factors is a general mechanism to maintain environment-dependent 487 signals as long-term epigenetic memory 40 . Interestingly, we did not detect FOS and JUN 488 binding sites in motif enrichment and genomic footprint analysis (data not shown), which 489 might explain why differential OCR acquired in culture are not maintained after re-490 transplantation in vivo.

492
Our results are important on three fronts: conceptually, as experimental model and for their 493 therapeutic potential.

495
Our observations extend our previous findings that demonstrated the compatibility of self-496 renewal with the differentiated state of mature macrophages 17, 18, 41 . Here we show that AM 497 can be expanded extensively in culture for almost a year or more, but still assume full 498 epigenetic identity and functionality upon re-integration in their natural environment in vivo.

499
This is conceptually exciting, because it demonstrates that neither culture adaptations nor 500 long-term proliferation erase the epigenetic and transcriptional identity of AM. Such behavior 501 is normally only known for stem cells that can be expanded extensively or indefinitely in 502 culture and still functionally integrate into a tissue environment, for example pluripotent stem 503 cells that can integrate into an embryo and give rise to a whole new organism or epithelial 504 stem cells that can provide functional skin tissue for transplantation 31, 42 . 505 506 Furthermore, our results show that expanded AM are a long-term reliable culture model for 507 resident macrophages that provides similar advantages as a cell line, in terms of providing 508 large quantities of cells and the possibility of starting and stopping cultures by freeze/thawing 509 without being dependent on complex differentiation protocols. In addition, the possibility to 510 shuttle between in vivo and culture environment provides a unique experimental system for 511 genetic and biochemical manipulation that cannot be provided by other macrophage culture 512 systems.

514
Finally, although obtained with mouse cells, our results have highly promising implications for 515 potential macrophage cellular therapies. T-cells have been highly successful in cancer 516 immuno-therapy but their use has been largely limited to hematological cancers, because they 517 typically fail to infiltrate solid tumors 15 . As major constituent of the tumor stroma, 518 macrophages have a high potential to treat solid tumors. Furthermore, macrophages fitted 519 with a chimeric antigen receptor (CAR-macrophages) have shown promise to treat tumors in 520 animal models 43,44 . Unfortunately, the use of macrophage cellular therapy has so far been 521 limited by the difficulty to grow large numbers of macrophages in culture 16 , as has been 522 possible for T-cells 15 . Our results now demonstrate that in principle large scale and long-term 523 expansion of macrophages ex vivo is possible. Importantly, our results also show that even 524 massive and long-term expansion ex vivo does not compromise the identity of transplanted 525 macrophages in vivo, a feature that is critical both for efficacy and safety considerations. 526 Beyond cancer applications, the robustness of macrophage identity through culture will also 527 be important for multiple potential macrophage-based cellular therapies in regenerative 528 medicine and infectious disease 45 . For example, macrophage therapy has shown high promise 529 to treat degenerative liver disease 46, 47 and pulmonary alveolar proteinosis 29,48   Isolation of AM from BAL was done as previously described 19 . The expansion and maintenance 576 of exAMs ex vivo was done as described previously 19 . Briefly, cells were plated at a density of 577 1.1 -1.5 x 10 6 per 10 cm petri dish in 10 ml complete medium. The medium was supplemented 578 with murine 2-5% GM-CSF supernatant from J558L cells transfected with murine GM-CSF. 579 Cultures of exAM were passaged every 3-4 days and cultured for 2-4 months for functional 580 assays or transplantation studies. 581 582 Preparation of lung homogenates for flow cytometry analysis 583 Mice were euthanized by cervical dislocation, lungs were collected, cut in small pieces and 584 incubated with 1 mg/ml collagenase-2 (Worthington) and 0.15 mg/ml DNaseI (Sigma) at 37°C 585 for 30 min with constant agitation. Cell suspension was filtered through a 70 µm mesh and 586 erythrocytes were removed by RBC lysis (RBC Lysis buffer, Invitrogen). 587 588 Flow cytometry 589 Lung cell suspensions was pre-incubated with Fc receptor blocking antibody (clone 2.4 G2, BD 590 Pharmingen or TruStain FcXTM, Biolegend) and Zombie fixable cell viability dye (NIR 77814, 591 Aqua 77143 or UV 77474; Biolegend) or DAPI (D9542, Sigma) for 15 -20 min at 4°C. For 592 antibody staining, cells were incubated with antibody cocktail for 20 min at 4°C. FACS sorting 593 and analysis was done with FACSC anto, LSRII and LSRFortessa systems (BD). The following 594 antibodies were used for staining cells: anti-CD11b (clone M1/70, eBioscience or Biolegend), 595 anti-CD11c (clone N418, eBioscience or B-Ly6, BD), anti-F4/80 (clone BM8, eBioscience), anti-596 MHCII (clone M5/114.15.2, eBioscience), anti-B220 (clone RA3-6B2, eBioscience), anti-CD45.1 597 (clone A20, eBioscience or Biolegend), anti-CD45.2 (clone 104, eBioscience or Biolegend), anti-598 Ly6C (Clone AL-21, BD), anti-CD64 (clone X54-5/7.1, Biolegend or BD), anti-SiglecF (clone E50-599 2440, BD). Diva software was used for acquisition of data and Flowjo software V10 for data 600 analysis (TreeStar). 601 602

Intra-tracheal (i.t) and intra-nasal transplantation of alveolar macrophages (AM) 603
Mice were anesthetized by isoflurane inhalation, placed on their back, the tongue was gently 604 pulled out and 0.8 -1 x 10 6 fresh AM or exAMs in 80 -100 µl PBS were instilled intra-tracheally 605 using 1 ml syringe with a blunt 22G gavage needle. Mice were observed while recovering from 606 anesthesia and then returned to their cages for routine care and handling. Neonatal Csf2rb -/-607 mice (postnatal day 3) were anaesthetized with isoflurane and transplanted with 0.4 x 10 5 AM 608 resuspended in PBS and in a total volume of 7 µl.

610
Ex vivo alveolar macrophage assays 611 Cytospins and in-well photography 612 AM were plated at a density of approx. 0.4 x 10 5 cells/cm 2 for 4 hours for attachment at either 613 6-well or 100mm non-treated cell culture dishes. Bright-field images acquired from well center 614 using Zeiss Axio Vert A1 microscope at 10x magnification. Fresh BAL AMs or exAM (2 months 615 culture, detached and counted) were subjected to cytocentrifugation 300 -450 rpm for 4 min 616 using a Cytospin 4 (Thermo Fisher) and performed Diff-Quik staining following manufacturer's 617 protocols (9990700, Thermo Scientific TM Rapid-chrome TM Kwik-diff TM kit). Images were 618 acquired using the 3DHistech slide scanner and images were processed using Qupath v.0.2.0 619 with built-in Image J image analysis software. For Csf2rb -/transplants, brightfield images were 620 taken with an inverted microscope (Dmi1, Leica) after transfer of BAL cells to a Neubauer 621 counting chamber. Cell suspensions were cytospin at 800 rpm for 3 min using a Cytospin 4 622 (Thermo Fisher), followed by Kwik-diff staining according to the manufacturer's instructions 623 (9990700, Thermo Scientific TM Rapid-chrome TM Kwik-diff TM kit). Images were taken with a 624 Zeiss Axiolab A1 microscope.

626
Apoptosis assay 627 exAM plated at a density of 1x10 5 cells/well of a non-treated 24 well plate. (Nunc). As positive 628 control for apoptotic cells, apoptosis was induced in exAM using 1 µM Staurosporine (S5921, 629 Sigma) for 1 hour at 37°C and 5% CO2. Untreated exAM or apoptotic exAM were stained using 630 the PE AnnexinV apoptosis detection kit I (559763, BD Pharmingen) according to 631 manufacturer's protocols and analyzed by flow cytometry. Percentage of Early (AnnexinV + /7-632 AAD -) and late apoptotic (AnnexinV + /7-AAD -) cells were determined on total cells.

634
Bacterial killing assay 635 AM were plated overnight at a density of 1x10 6 cells/well of a non-treated 96 well plate. 636 (Nunc). Klebsiella Pneumoniae wild-type strain Kp52.145 was obtained from the Institut 637 Pasteur collection 51, 52 . The bacteria were grown overnight in LB medium at 37°C in a shaker 638 incubator (180 rpm). The bacteria were precultured until they reached a log phase of growth 639 after which they were incubated with the AM at MOI 100 for 15 min at 37°C and 5% CO2. The 640 supernatant was discarded and cell lysates were harvested at 0, 30 and 90 minutes post 641 incubation with bacteria. Gentamicin solution (1:1000) (G1272, Sigma) was added to the 642 medium for the 30-and 90-minute timepoints to prevent extracellular bacterial replication. 643 The cell lysates were plated in serial dilutions in LB agar and the colonies were estimated the 644 following day.

646
Cell cycle analysis 647 AM were plated at density of 0.2 x 10 6 cells/well of non-treated 12-well dishes (Nunc). The 648 cells were directly pulsed with 10 µM 5-ethynyl-2'-deoxyuridine, EdU (C10634, Thermo Fisher) 649 for 4 hours at 37°C to label proliferating cells. For expanded AMs (exAM), total RNA was extracted from exAMs after 2 months of culture. 697 exAM were transplanted into 2 months old CD45.1 mice and analyzed 4 months post 698 transplantation. In vivo AM subsets were sorted (Live, Singlets, SiglecF + CD11c + , CD45.2 + or 699 CD45.1 + ) from total lung homogenates directly into RNA lysis buffer (RLT buffer, Qiagen) for 700 RNA extraction using BD FACSAriaIII machine. RNA-seq samples were generated from a pool 701 of 3 mice or exAM cultures in duplicates. Amplified cDNA was prepared from 2 ng of total RNA 702 using the Ovation RNA-seq system V2 (NuGEN Technologies, Inc.) following manufacturer's 703 instructions. Briefly, first-strand cDNA was prepared using a combination of random and poly-704 T DNA/RNA chimeric SPIA (single primer isothermal amplification) primers and rev erse 705 transcriptase. Priming sites were used to synthesize second strand cDNA using a DNA 706 polymerase, was purified using Agencourt RNAClean XP beads (Beckman Coulter Inc.). and 707 subjected to SPIA. Amplified cDNA was purified using AMPure XP beads (Beckman Coulter Inc.) 708 and 500 ng was fragmented by sonication using a Covaris E210 instrument.   Immgen consortium 55 (GSE122108). a) Symbolic representation of AM identity OCR maintained across conditions, as well as 932 culture adaptation-dependent OCR gained or natural environment-dependent OCR 933 lost in culture but restored in the natural niche in vivo. 934