Dopaminergic neurons establish a distinctive axonal arbor with a majority of non-synaptic terminals

Chemical neurotransmission in the brain typically occurs through synapses, which are structurally and functionally defined as sites of close apposition between an axon terminal and a postsynaptic domain. Ultrastructural examinations of axon terminals established by monoamine neurons in the brain often failed to identify a similar tight pre- and postsynaptic coupling, giving rise to the concept of “diffuse” or “volume” transmission. Whether this results from intrinsic properties of such modulatory neurons remains undefined. Using an efficient co-culture model, we find that dopaminergic neurons establish an axonal arbor that is distinctive compared to glutamatergic or GABAergic neurons in both size and propensity of terminals to avoid direct contact with target neurons. Furthermore, while most dopaminergic varicosities express key proteins involved in exocytosis such as synaptotagmin 1, only ~20% of these are synaptic. The active zone protein bassoon was found to be enriched in a subset of dopaminergic terminals that are in proximity to a target cell. Irrespective of their structure, a majority of dopaminergic terminals were found to be active. Finally, we found that the presynaptic protein Nrxn-1αSS4- and the postsynaptic protein NL-1AB, two major components involved in excitatory synapse formation, play a critical role in the formation of synapses by dopamine neurons. Taken together, our findings support the idea that dopamine neurons in the brain are endowed with a distinctive developmental program that leads them to adopt a fundamentally different mode of connectivity, compared to glutamatergic and GABAergic neurons involved in fast point-to-point signaling. SIGNIFICANCE STATEMENT Midbrain dopamine (DA) neurons regulate circuits controlling movement, motivation, and learning. The axonal connectivity of DA neurons is intriguing due to its hyperdense nature, with a particularly large number of release sites, most of which not adopting a classical synaptic structure. In this study, we provide new evidence highlighting the unique ability of DA neurons to establish a large and heterogeneous axonal arbor with terminals that, in striking contrast with glutamate and GABA neurons, actively avoid contact with the target cells. The majority of synaptic and non-synaptic terminals express proteins for exocytosis and are active. Finally, our finding suggests that, NL-1A+B and Nrxn-1αSS4-, play a critical role in the formation of synapses by DA neurons.


SIGNIFICANCE STATEMENT 52
Midbrain dopamine (DA) neurons regulate circuits controlling movement, motivation, and 53 learning. The axonal connectivity of DA neurons is intriguing due to its hyperdense nature, with a 54 particularly large number of release sites, most of which not adopting a classical synaptic 55 structure. In this study, we provide new evidence highlighting the unique ability of DA neurons to 56 establish a large and heterogeneous axonal arbor with terminals that, in striking contrast with 57 glutamate and GABA neurons, actively avoid contact with the target cells. The majority of synaptic 58 and non-synaptic terminals express proteins for exocytosis and are active. Finally, our finding 59 suggests that, NL-1 A+B and Nrxn-1 SS4-, play a critical role in the formation of synapses by DA containing the sequence of NL-1 or CD4 and incubated overnight at 37°C. HEK293T cells were 262 trypsinized (0,05%), washed and then co-cultured (30 000 cells/mL) with VTA or SNc DA neurons 263 (120 000 cells/mL) from P0-P3 DATcre-AI9 pups at 9DIV. HEK293T cells were also co-cultured 264 (30 000 cells/mL) with cortical neurons (120 000 cells/mL) as a positive control for the artificial 265 synapse formation assay. Co-cultures were fixed 24h later (10DIV) with 4% PFA for 30 min, 266 followed by permeabilization with 0.1% triton for 20 min and incubation with blocking solution 267 (5% BSA) for 10 min. Immunolabelling was performed as described previously. 268 269

Dopaminergic neurons in vitro establish mostly non-synaptic terminals 306
Previous work using transmission electron microscopy (TEM) demonstrated that in the intact 307 rodent brain, the axon terminals of DA neurons in the striatum are only rarely found in synaptic 308 contact with target cells, the directly apposed cellular membrane being typically devoid of a post 309 synaptic density (PSD) (43). Whether this represents an intrinsic property of mesencephalic DA 310 neurons or requires complex signals received in vivo during development is unknown. To begin 311 exploring this key question we first used scanning electron microscopy (SEM) to visualize the 312 external morphology of axonal varicosities in co-cultures of VTA or SNc DA neurons, obtained 313 from TH-GFP transgenic mice, growing with VS or DS neurons, respectively, as well as astrocytes. 314 Neurons were grown on surface-optimized titanium disks and examined by SEM after 14 days 315 ( Fig. 1A and 1B). We observed that in such cultures, some of the neurons, likely dopaminergic, 316 established a complex array of oblong axonal varicosities in close contact with the underlying 317 astrocytes, but without direct contact with dendrites or neuronal cell bodies ( Fig. 1D and 1E). 318 These varicosities have an average length of 1.4 µm with an average inter-varicose interval of 2.6 319 µm, highlighting the high density of varicosities along the axon (Fig. 1D). We found that only a 320 very small contingent of varicosities along such axons established synaptic-like contacts with 321 neuronal dendrites (Fig. 1F). Globally in these co-cultures, approximately 15% of neurons 322 developed axonal arbors with a majority of non-synaptic contacts, in line with the typical 323 proportion of DA neurons in such cultures. To confirm that this apparently non-synaptic 324 connectivity originated from DA neurons, we prepared a second set of cultures in which FACS-325 purified DA neurons were grown without striatal neurons. In such monocultures, we observed a 326 large predominance of free axonal varicosities ( Fig. 1G and 1H), confirming that DA neurons have 327 an intrinsic propensity to develop a large and complex axonal arbor, with most varicosities being 328 non-synaptic. As a point of comparison, we also examined striatal cultures, containing more than 329 95% GABAergic medium spiny neurons. In this system we found a large predominance of axonal 330 varicosities establishing synaptic-like contacts, with most axons being non-varicose and 331 containing only a small subset of free axonal varicosities ( Fig. 1I and 1J). Although accounting for 332 less than 5% of striatal neurons, some of these free varicosities could also originate from the 333 axons of striatal cholinergic neurons, previously reported to also establish mostly non-synaptic 334 terminals in the striatum (21). 335 We then used TEM (Fig. 1C) to visualize the internal organization of axonal varicosities in 336 co-cultures of VTA or SNc DA neurons, growing with VS or DS neurons, respectively, on astrocytes. 337 Using 3,3′-diaminobenzidine tetrahydrochloride (DAB), we performed immunostaining against 338 GFP to distinguish DA and non-DA terminals. In our in vitro system, we easily found DAB negative 339 synaptic terminals, identified by the presence of a post-synaptic domain as illustrated in Fig. 1K. 340 We then identified multiple sets of DAergic terminals, all filled with large pools of synaptic 341 vesicles, arguing that they possess the capacity to package and release neurotransmitters ( Fig.  342   1L, 1M and 1N). However, none were found in close proximity to a post-synaptic domain, 343 revealing the asynaptic nature of DAergic terminals (Fig. 1O). Due to the possibility that some of 344 these terminals may show a post-synaptic domain in another plane of the varicosity, we searched 345 for the same axon varicosity on different ultrathin sections. Even in such case, no post-synaptic 346 domains were observed in close proximity to DAergic terminals ( Fig. 1P and 1Q). 347 348

Non-synaptic dopaminergic axonal varicosities express distinct sets of presynaptic markers and 349 appear to actively avoid target cells 350
Although SEM and TEM analysis revealed the non-synaptic nature of DAergic axonal varicosities, 351 it was unclear if all such morphologically identified varicosities represented actual axon terminals 352 endowed with the molecular machinery required for neurotransmitter release. We next used 353 immunocytochemistry coupled to confocal microscopy to examine the axonal varicosities 354 established by DA neurons from TH-GFP mice. Double labelling for GFP, to visualize all varicosities 355 and the ubiquitous Ca 2+ sensor for exocytosis Syt1, we found that 89.0 ± 2.5% of varicosities 356 established by SNc DA neurons were positive for Syt1 ( Fig. 2A and 2F). This proportion was slightly 357 smaller for VTA DA neurons (69.3 ± 2.8%) (Fig. 2F). To gain further insight into the neurochemical 358 identity of these axon terminals, we next examined the presence of the vesicular monoamine 359 transporter VMAT2, necessary for the vesicular packaging and release of DA and the type 2 360 vesicular glutamate transporter VGluT2, necessary for glutamate release by DA neurons (14, 44-361 46). Double-labeling DA neuron co-cultures for GFP and VMAT2 revealed that 50.1 ± 6.1% and 362 53.8 ± 5.4% of SNc and VTA DA neuron terminals, respectively, contained VMAT2 ( Fig. 2B and 2G). 363 Similar double-labeling experiments evaluating VGluT2 expression revealed a small proportion of 364 GFP-positive glutamatergic terminals established by DA neurons (Fig. S2). These results indicate 365 that most GFP-positive axonal varicosities are likely to represent sites allowing the release of DA, 366 as well as other neurotransmitters by DA neurons. 367 As our results to this point suggested that most DAergic terminals were non-synaptic in 368 nature and thus tended to spontaneously avoid interacting with the otherwise numerous striatal 369 medium spiny neurons, we next aimed to examine more directly the interaction of DAergic 370 terminals with the somatodendritic domain of medium spiny neurons, visualized here using a 371 MAP2 antibody. Using triple-labeling for GFP, Syt1 and MAP2, we compared random fields of SNc 372 or VTA DAergic axonal domains with axonal fields acquired from striatal and cortical cultures, thus 373 providing a comparison with neuronal populations known to be essentially completely synaptic 374 in their connectivity ( Fig. 2C to 2E). We calculated the proportion of GFP/Syt1 double-positive 375 varicosities that were also in direct contact with MAP2-positive elements, thus providing an index 376 of the propensity of terminals to be in close contact with the dendritic or somatic membrane of 377 target cells. Strikingly, we found that only 14.9 ± 2.4% of axon terminals established by SNc DA 378 neurons were in contact with target cells (Fig. 2C to 2E). This proportion was not significantly 379 different for VTA DA neurons (17.8 ± 2.1 %) (Fig. 2H). In marked contrast, 84.5 ± 2.7% of terminals 380 in cortical cultures ( Fig. 2D and 2H) and 74.2 ± 7.1% of terminals in striatal cultures ( Another intriguing aspect of the connectivity of DA neurons is the very dense, highly complex and 388 branched nature of their axonal compartment (12, 15, 16, 47). As we examined the distribution 389 of presynaptic markers along the axonal domain of DA neurons, we observed that the density of 390 axonal varicosities along DAergic axons appeared to be higher compared to other types of 391 neurons. We therefore quantified the distribution of Syt1-positive axonal varicosities along the 392 axonal domain of VTA and SNc DAergic neurons and compared this to striatal GABA and cortical 393 glutamate neurons ( Fig. 3A to 3C). We found that the inter-varicosity distance was 9.0 ± 1.8μm 394 and 10.9 ± 2.2μm for SNc and VTA DA neurons, respectively. This value was two-fold larger for 395 GABA and glutamate neurons, for which the inter-varicosity distance was 22.1 ± 3.6μm and 27.6 396 ± 2.9μm, respectively (Fig. 3D). This observation further highlights the fundamentally different 397 nature of dopaminergic axonal arbors compared to those of GABA and glutamate neurons. However, it has not been determined whether synaptic and non-synaptic DA terminals 407 differentially contain such scaffolding proteins. We hypothesized that synaptic DA terminals 408 might be more likely to express active zone scaffolding proteins compared to the non-synaptic 409 DA terminals because of the necessity to maintain a stable pre-and postsynaptic complex. 410 Confirming previous results, we found that only 36.0 ± 5.5% and 32.2 ± 3.8% of SNc or VTA DA 411 terminals were positive for bassoon (Fig. 4A, 4D). A striking observation was that the majority of 412 bassoon positive DAergic terminals were either in direct contact or in close proximity to a target 413 cell. In axonal fields far removed from other neurons, axonal varicosities were typically completely 414 devoid of bassoon (Fig. 4A, upper panels, 4F). In axons closer to striatal neurons (Fig. 4A, lower 415 panels), approximately one third of varicosities in contact with MAP2-positive somatodendritic 416 domains were bassoon positive (27.3 ± 4.5% and 28.9 ± 3.0% for SNc and VTA DA neurons, 417 respectively; Fig. 4E). In these fields, bassoon was found in DA varicosities that were located 418 between 0.5μm to 40μm from MAP2-positive dendrites, as illustrated schematically ( Fig. 4G and 419

4E). 425
We found RIM1/2 to also be sparsely expressed by DAergic terminals, unrelated to the 426 presence of target cells. While in some fields, good colocalization of RIM1/2 and bassoon was 427 observed (Fig. S4A), in others, they were mostly segregated ( Fig.S4B and S4C). Globally, 31.7 ± 428 5.1% and 38.9 ± 5.1% of SNc and VTA DAergic terminals contained RIM1/2 (Fig. S4D). The 429 proportion of bassoon-positive terminals expressing RIM1/2 was 33.2 ± 5.4% and 29.6 ± 5.9%, 430 respectively for SNc and VTA DA neurons (Fig. S4E). Conversely, 37.6 ± 6.3% and 25.7 ± 5.3% of 431 bassoon positive SNc and VTA DAergic terminals contained RIM1/2 (Fig. S4F) with glutamatergic (PSD95) and GABAergic (gephyrin) synapses (53, 54). We found that a small 444 subset of Syt1-positive terminals established by SNc DA neurons was in close apposition to PSD95 445 (12.6 ± 3.3%) or gephyrin (5.2 ± 1.8%). A similar low proportion of VTA DA neuron terminals were 446 found in close proximity to PSD95 (8.0 ± 2.3%) or gephyrin (2.9 ± 0.8%) (Fig. 5A, 5E and 5F). The 447 difference between SN and VTA was not significant. To visualize at higher resolution the molecular 448 architecture of synaptic and non-synaptic terminals established by primary DA neurons, we used 449 direct stochastic optical reconstruction microscopy (dSTORM). For the synaptic DA varicosities, 450 the dSTORM images revealed a clear apposition between the DAergic varicosity, defined by GFP 451 nanocluster signal, and the postsynaptic domain of a target cell, defined by PSD95 nanocluster 452 signal (Fig. 5H). An example of non-synaptic DAergic terminal visualized by super-resolution 453 imaging is also illustrated, demonstrating the lack of proximity to a PSD domain (Fig. 5I). In stark 454 contrast, 83.6 ± 2.2% of VGluT1/Syt1 terminals established by cortical glutamate neurons 455 colocalized with PSD-95 ( Fig. 5B and 5E). Similarly, the large majority (60.6 ± 3.7%) of terminals 456 established by striatal GABAergic neurons were in close apposition to gephyrin postsynaptic 457 clusters ( Fig. 5C and 5F). Because previous work has suggested that olfactory bulb (OB) DA 458 neurons release GABA as a co-transmitter at many of their release sites, we also examined 459 synaptic release sites from such neurons. Interestingly we found that 55.4 ± 5.2% of Syt1-positive 460 DAergic terminals established by such neurons colocalized with gephyrin (Fig. S5). 461 Although evidence for DA receptor clusters at synaptic sites in the brain is limited (23, 24), 462 DA neurons could in principle also establish synaptic-like release sites that are in close apposition 463 to postsynaptic DA receptors, independently from PSD-95 or gephyrin domains. To examine this, 464 we took advantage of a knock-in mouse line expressing GFP-tagged D2 receptors and visualized 465 release sites specialized for DA release using VMAT2 immunostaining. In co-cultures of WT DA 466 neurons together with striatal neurons from GFP-D2 mice (Fig. 5D), we found that 54.0 ± 3.1% of 467 TH-positive terminals established by SNc DA neurons contained VMAT2 (Fig. 5G). Similarly, 57.2 468 ± 3.6% of TH-positive DA terminals established by VTA DA neurons were VMAT2 positive (Fig. 5G). 469 Arguing here again for a strong propensity of DAergic axons to establish release sites that fail to 470 seek out target cells, we found that only 7.6 ± 1.2% of TH/VMAT2 terminals established by SNc 471 DA neurons contacted D2 receptor clusters, while this proportion was 2.0 ± 0.8% for VTA DA 472 neurons (Fig. 5G). 473 474

FM1-43 imaging reveals a majority of active DA terminals 475
Considering the large heterogeneity of DAergic terminals identified in the present work, we 476 wished to determine whether all or only a subset of morphologically defined axonal varicosities 477 are functional and competent to release neurotransmitters by exocytosis. Here we used the 478 activity-dependent uptake and release of the well-characterized endocytotic probe FM1-43, 479 allowing us to examine activity-dependent vesicular cycling independently of the neurochemical 480 phenotype of the varicosities. After a first round of FM1-43 uptake induced by high potassium 481 induced membrane depolarization, a second round of membrane depolarization was used to 482 identify terminals that showed activity-dependent release of FM1-43 ( Fig. 6A and 6B). A total of 483 1091 SNc and 543 VTA varicosities were examined. To isolate activity-dependent release from 484 spontaneous release or probe bleaching, we used as a criterion a minimal release of 20% of the 485 initially up taken FM1-43 in response to the second round of membrane depolarization (Fig. 6C). 486 We found that 84.62 ± 4.60% of SNc and 77.58 ± 9.62% of VTA axonal varicosities, identified as 487 potential release sites by the presence of Syt1, were active (Fig. 6D). Post-experimental 488 immunostaining of the same fields that were imaged revealed that a majority of FM1-43 puncta 489 along DAergic axons contained Syt1 or VMAT2 (Fig. 6E). form of NL-1 (splice variant A + B + ; NL-1 AB ) tagged with extracellular HA or a negative control 500 membrane protein (CD4 with HA tag) and we examined whether DA axonal varicosities were 501 recruited to establish synapse-like contacts ( Fig. 7A and 7B). We found that HEK293T cells 502 expressing NL-1 AB were significantly more attractive for DA terminals, identified by the presence 503 of RFP or VMAT2, compared to HEK293T cells expressing the control protein CD4 (Fig. 7C). 504 Quantification of the total intensity of RFP and VMAT2 puncta on HEK293T cells expressing NL-505 1 AB showed a 5-fold and a 27-fold increase in signal, respectively, compared to the CD4 control 506 group, representing background signal due to the random distribution of axons ( Fig. 7D and 7E). 507 Cortical neurons were used in similar experiments to compare the results obtained with DA 508 neurons to a population of neurons known to establish mostly synaptic-type terminals. We found 509 that NL-1 AB also induced a robust recruitment of cortical terminals (Fig 7F) as previously reported 510 (6). Quantification of the total intensity of VGluT1 puncta on HEK293T cells expressing NL-1 AB 511 showed an 18-fold increase compared to the control CD4 condition (Fig. 7G). These results 512 suggest that NL-1 AB has efficient synaptogenic activity to induce presynaptic differentiation of DA 513 neurons, to a level comparable to that observed for cortical neurons. This observation suggests 514 that the limiting factor preventing most of the numerous terminals established by DA neurons to 515 form synapses is more likely to be at the presynaptic rather than at the postsynaptic level. 516 517

Overexpression of Nrxn 1 SS4promotes the formation of both excitatory and inhibitory 518 synapses by DA neurons 519
One of the most important transsynaptic binding partners of NL-1 is the presynaptic protein Nrxn-520 1. This protein is also involved in glutamatergic synapse formation and has two versions: the long 521 version, -neurexin (Nrxn 1), and the short version, ß-neurexin (Nrxn 1ß). Both Nrxn 1 and 522 Nrxn 1ß, regulate excitatory synapse formation and function (30, 57). Considering that a subset 523 of DA neurons uses glutamate as a second neurotransmitter, we first investigated the 524 endogenous expression of Nrxn 1 in DA neurons. We found that both isoforms are expressed at 525 high level in both VTA and SNc DA neurons, although Nrxn 1 was expressed at higher levels in 526 VTA DA neurons (Fig. S6). We therefore examined whether overexpression of Nrxn 1 influences 527 synapse formation by these neurons. Lentiviral overexpression of Nrxn 1 (splice version SS4-; 528 Nrxn 1 SS4-) in DA neuron co-cultures ( Fig. 8A and 8B) caused a 2-fold increase in the proportion 529 of DA terminals colocalizing with the excitatory postsynaptic organizer PSD95 compared to the 530 control group (Fig. 8C and 8D). The average size of PSD95 positive synaptic puncta or their total 531 area was unchanged ( Fig. 8E and 8F), suggesting that the overexpression of Nrxn1 SS4in DA 532 neurons triggered the establishment of more contacts with pre-existing postsynaptic clusters. 533 Overexpression of Nrxn1 SS4in DA neurons induced a similar increase in the proportion of DA 534 terminals colocalizing with the inhibitory postsynaptic protein Gephyrin (Fig. 8G and 8H), with no 535 change in Gephyrin puncta size or total Gephyrin puncta area ( Fig. 8I and 8J). 536 537 DISCUSSION 538 In this study we provide a novel perspective on the connectivity of DA neurons with an exhaustive 539 characterization of the axonal domain of this key neuromodulatory brain system. We found that 540 DA neurons establish a highly distinctive axonal domain with a majority of non-synaptic terminals, 541 in a way that is fundamentally different compared to cortical glutamatergic and striatal GABAergic 542 neurons. Both synaptic and non-synaptic DAergic varicosities express basic presynaptic proteins 543 such as Syt1, suggesting that the majority of these varicosities have the machinery to release 544 neurotransmitters. We discovered that synaptic and non-synaptic terminals differ in their 545 expression of active zone structural proteins: the active zone protein bassoon was found to be 546 mainly expressed in DA terminals located close to a target cell, while most non-synaptic terminals 547 were devoid of bassoon. The active zone protein RIM1/2 was more sparsely expressed in both 548 synaptic and non-synaptic DA terminals. Using the activity-dependent synaptic vesicle probe 549 FM1-43, we found that the majority of DA terminals, both synaptic and non-synaptic, are active. 550 Finally, providing initial insight into the mechanistic underpinnings of this unique connectivity, we 551 find that the postsynaptic trans-synaptic protein NL-1 is able to efficiently trigger formation of 552 synapse-like structures by primary DA neurons, while overexpression of the presynaptic trans-553 synaptic protein Nrxn-1 causes an increase in the proportion of terminals adopting a synaptic 554 configuration. 555 556 Heterogeneity in DA release sites 557 We found that 80% of axonal-like varicosities established by DA neurons express the exocytosis 558 calcium sensor Syt1, suggesting that the SNARE complex and associated proteins are present in 559 the majority of axonal varicosities established under our experimental conditions ( Fig. 2A and 2F). 560 Although we have not examined in detail the presence of other vesicular exocytosis proteins in 561 the present study, most of these varicosities also contain SNAP25 and SV2. The presence of Syt1 562 in terminals has previously been shown to provide a good index of functionality (1). Perhaps a bit 563 surprising, the expression of VMAT2, considered as a specific marker of terminals that release DA 564 or other monoamines was found in only 55% of DA terminals (Fig. 2B and 2G). These results 565 suggest that only half of DA terminals can package and release DA. One possible explanation of 566 this low proportion is the fact that some DA neurons are also able to package and co-release other 567 neurotransmitters at some of the axon terminals, including glutamate and GABA (11, 14, 50, 58). 568 Furthermore, recent studies have reported that co-release of DA and glutamate mainly occurs 569 from separate, segregated sets of terminals (52, 59). The present experiments were performed 570 with primary DA neurons in co-culture with striatal neurons. The medial part of the VTA and the 571 lateral part of the SNc were used. Because previous work reported that VMAT2 is more expressed 572 in DA neurons located in the lateral part of the VTA compared to the medial part (60), it is possible 573 that this may have contributed to our finding of a large proportion of terminals without 574 detectable VMAT2. Also, recent work suggested that 75% of DA neurons in the mesencephalon 575 express the mRNA encoding for VMAT2 suggesting that a small subset doesn't express VMAT2, at 576 least at this time point in vitro (52). Additional studies will be required to examine this further and 577 to determine whether this phenomenon is developmentally regulated. 578

579
The active zone proteins bassoon and RIM are mainly segregated 580 Our work provides new insights on the differential molecular make up of synaptic and non-581 synaptic DA terminals. At classical synapses, release sites are characterized by the presence of 582 active zones that contain different scaffold proteins including RIM, RIM-BP, bassoon/piccolo or 583 ELKS (48, 61). The active zone protein RIM has been identified as a key scaffold protein notably 584 involved in vesicle priming and Ca 2+ channel tethering (3, 62). Bassoon is also an important 585 scaffold protein and plays a role in synaptic vesicle clustering without participating directly in 586 synaptic vesicle exocytosis (63). In our in vitro system, we first found that only 30% of DA 587 terminals contained bassoon. Strikingly, we observed that the majority of them were in close 588 proximity or in contact with a target cell (see Fig. 4 and S3). We also found that only 30% of 589 bassoon positive DA terminals were also positive for RIM1/2 (Fig. S4F). Conversely, only a third of 590 RIM1/2 positive terminals contained bassoon (Fig. S4E). The mechanism and functional 591 implications of this differential expression of active zone proteins in DA terminals are presently 592 unclear. Interestingly, the conditional KO of RIM1 and RIM2 in DA neurons was recently shown to 593 completely block evoked DA release, while extracellular levels of DA were only partially decreased 594 (49). This finding suggests the possibility that while action potential-evoked DA release requires 595 RIM1/2, spontaneous release occurs through a separate mechanism. Spontaneous DA release 596 could thus potentially be mediated by Syt1-and SNARE-dependent exocytosis without any 597 implication of active zone proteins, as previously suggested for glutamate release from cortical 598 neurons (64). In our work, we found that approximately 80% of DA terminals contain Syt1, 599 suggesting that most of them have the capacity to release DA, spontaneously or in an activity-600 dependent manner. Our ultrastructural observations with TEM also support this conclusion as 601 most varicosities examined contained large pools of synaptic vesicles (Fig. 1L). Our observation of 602 the selective presence of the active zone scaffold protein bassoon at DA terminals located in close 603 proximity to a target cell is particularly intriguing. Previous work in hippocampus has suggested 604 that target cells can play a role in synapse maturation during synaptogenesis and that some 605 neurons also possess non-synaptic release sites that are functional and mobile and that express 606 the active zone protein bassoon (65, 66). Although we hypothesize that target-derived signals are 607 likely to be required to restrict bassoon to release sites that are in close proximity to target cells, 608 further work will be required to test this directly. varicosities are able to release DA (76). Here we took advantage of the high signal to noise ratio 651 and simplicity of our in vitro model to examine whether the large numbers of varicosities 652 established by DA neurons are active. We chose to evaluate this using a FM1-43 uptake assay, an 653 approach that labels activity-dependent, actively recycling vesicles, irrespectively of their 654 neurochemical identity. We found that a large majority of axonal varicosities established by DA 655 neurons are active, suggesting that the majority of non-synaptic terminals represent functional 656 neurotransmitter release sites, even if they are not associated with a tightly linked postsynaptic 657 specialization. The different conclusions of the present study compared to that of Pereira and 658 colleagues could result from several technical considerations. One of these is that this earlier 659 study used a false fluorescent neurotransmitter thought to selectively label DAergic terminals 660 containing VMAT2, while in the present study, we examined terminals irrespective of whether 661 they contained VMAT2, VGluT2 or potentially other neurotransmitters such as peptides. This 662 could have potentially provided a broader overview of all terminals, especially considering that 663 we found a large subset of syt1-positive terminals established by DA neurons but devoid of 664 detectable VMAT2 (see Fig.2G and Fig.5G) Using an artificial synapse co-culture assay and a lentiviral vector to overexpress a transsynaptic 672 protein on primary DA neurons, in this study we found that the postsynaptic protein NL-1 AB 673 expressed on HEK293T represents an efficient signal to induce presynaptic differentiation of 674 DAergic terminals. We also found that presynaptic Nrxn1 SS4overexpression increases the 675 number of excitatory and inhibitory synapses established by DA neurons. 676 The NL-1 protein is preferentially found at excitatory synapses and directly interacts with 677 the scaffolding protein PSD95 (55, 77, 78). The maintenance and the function of excitatory 678 synapses are regulated by NL-1 via the recruitment of AMPA and NMDA receptors at the 679 postsynaptic domain (55). In our in vitro experiments, we clearly demonstrated that HEK293T 680 cells expressing the major form of NL-1 (splice variant A + B + ; NL1 AB ) attract DA terminals, 681 suggesting that NL-1 AB has the potential to induce presynaptic differentiation of axon DA 682 terminals. An implication of NL-1 in the establishment of synapses by DA neurons is in line with 683 the fact that some of these neurons establish glutamate synapses and release glutamate as a 684 second neurotransmitter at sites that are synaptic in structure (22). This result is also compatible 685 with our observation that HEK293T cells expressing NL-1 AB are able to induce the formation of 686 artificial synapses by DA neuron terminals and with our finding of robust expression of Nrxn-1ß 687 by DA neurons (Fig. 7 and Fig. S6). Nrxn-1ß strongly interacts with NL-1 containing the splice site 688 B, whereas Nrxn-1 is not able to (30). 689 Considering this ability of NL-1 to promote synapse formation by DA neurons, it is puzzling 690 that in the intact brain, expression of NL-1 by striatal neurons appears to promote glutamate 691 synapse formation by cortical neurons but not by DA neurons. Further experiments will be needed 692 to solve this important question and to determine which others pre-and postsynaptic proteins 693 regulate synapse formation by DA neurons. It is interesting to note that recent work has provided 694 evidence supporting the possibility that NL-2 regulates the formation of synapses by DA neurons 695 a highly complex axonal arbor endowed with a large repertoire of signaling mechanisms. We 713 envision that the striking, seemingly default capacity of DA neurons to establish an extensive 714 axonal arbor endowed with a large numbers of axon terminals that appear to actively avoid direct 715 contact with target cells is likely to be due to a developmental program that is shared with other 716 key modulatory neurons including those using serotonin, norepinephrine and acetylcholine. 717 Further work will be required to determine the common mechanisms involved.