Brain tissues and dissociated neurons in culture dishes represent diametrically opposed systems on the complexity scale. Between these two extremes, technological approaches of structuring substrates, both at the chemical and topographical levels, allow the in vitro implementation of specific characteristics or functions of cells and organs. These approaches emerged in the 90's with the development of micropatterning technologies (see [1] for a history in the particular case of neuroscience). They rapidly led to the control of single cell geometry, as illustrated by the seminal work of Chen et al. about the link between cell size and cell viability [2]. For neurosciences where the notion of cellular network is crucial, these technologies have led to the in vitro implementation of neuronal microcircuits. This has occurred almost concomitantly using micropatterning [3] and microfluidic techniques [4]. The aim of both approaches is to position the soma and to guide the growth of their extensions, mostly axons.

In these positioning and guidance strategies lies the notion of cellular confinement. Indeed, a single neuron forming autapses inside an adhesive circle [5], a neuronal process (neurite) guided by a thin stripe [6], or a soma constrained by a boomerang-like shape [7] are confined within the limits of adhesive patterns. Similarly, an axon guided by a microfluidic channel whose lateral dimensions are smaller than that of its growth cone (GC) will have its tip confined in both width and height.

This review is dedicated to how neurons, and more specifically the sub-cellular structures that are GCs and neurites, respond to a spatial confinement. Although I will heavily rely on in vitro micro-engineering approaches in this work, I would like to state that cellular growth in confined situations is not only a technological mean and the cell behavior under confinement an unavoidable contingency of in vitro systems. Importantly, neuronal growth occurs during the embryonic stage in a crowded non-homogeneous physico-chemical environment. Therefore, confinement is also a physical cue encountered in the developing nervous system. The developing cortex provides a paradigmatic example of neuron navigation in a packed tissue. The inside-out birth order arrangement of neurons imposes their progenitors to migrate along the thin processes of radial glia from the sub-plate to the pial surface i.e., into more and more populated cortical layers as embryonic time passes [8]. Another example in the same line concerns follower axons growing on pioneer axons which provide them with a scaffold [9]. A micron-size adhesive stripe, a fiber or a microchannel, and in general any environment implementing confinement in vitro would be a much better model of these in vivo situations than the bottom surface of a Petri dish.

In vitro technologies provide thus an exquisite control over the chemical and physical microenvironment at the micron scale, allowing to echo more faithfully in vivo situations than conventional 2D cultures and, even more interestingly, to reveal behaviors otherwise kept hidden (see [10] for a discussion on these aspects).

Cytoskeletal organization and dynamics as well as biomechanical aspects of neuronal growth are central to the response of neurons to confinement. Regarding specifically GCs, the reader can refer to recent reviews focusing on actin-based [11] and microtubule-based [12] mechanisms underlying GC motility, as well as on the crosstalk between both types of filaments [13], [14]. These reviews consider implicitly neurons evolving on 2D surfaces. In such a situation, GCs are compartmentalized into three zones dominated by the presence of one type of cytoskeletal structure: microtubules (MTs) emerging from the shaft in the central zone hitting an acto-myosin arc (defining the transition zone) beyond which the actin-rich peripheral domain extends [15]. The existence of an actin-microtubule crosstalk, supported by multiple microtubule-associated proteins (MAPs) (see [12] for a review), somehow blurs this stereotypical structural view, leading through different proposed mechanistic scenarios to MT invasion of the peripheral domain, controlling the distribution of filopodia length and lifetime along the GC perimeter.

Basically and importantly, both cytoskeleton filaments can sense or apply mechanical forces from their interaction with the local microenvironment trough adhesion complexes. The currently accepted mechanical view of neuronal growth is that the endogenous forces exerted by the GC are dissipated by adhesions along visco-elastic neurites [16], [17], [18]. In addition, neurites display an active i.e., contractile behavior [19], [20], [21], [22], [23]. A dynamic imbalance between the shaft contractility and the traction forces produced by the GC may govern the elongation rate. As an illustration of this conceptual framework, it was reported that diffusing and surface-bound repellents increase the relative part of the contractile contribution, while forward growth could be then restored by the inhibition of key molecular pathways responsible for cell contractility [24]. The readers can refer to [25] for an illuminating review about how mechanics couples with the chemical environment of neurons. How these acto-myosin mediated contractile forces are produced at the molecular level is not currently fully understood, but stems from the existence of actin rings arranged perpendicularly to the axon axis, spaced by spectrin tetramers and connected to myosin II motors [26], [27]. Several mechanical models takes into account the interplay of GC pushing and pulling forces, shaft viscoelasticity and contractility, and adhesions to describe axonal elongation (see, e.g., the detailed review of Olivery and Goriely [28]). Lastly, a complete picture of the mechanical link between the soma and the GC should include GC-like, anterogradely propagative structures produced by growing mammalian neurons [29], [30]. It was proposed that these so-called actin waves might extend the mechanical range of action of the GC during the neuron elongation phase, in addition to their contribution to slow transport [31].

Note that the mechanisms discussed above concern the existence of a GC seeking for a target. But, as stated by Rossi et al., “axons that already reached their targets have to elongate further to match the progressive expansion of the nervous system or of the whole organism” [32]. This interstitial growth modality is another evidence of axonal visco-elastic properties (for a review about interstitial growth, see [33]) and of force-mediated mass generation in the shaft [34].

The present review will cover the various experimental strategies (chemical, physical) for implementing confinement: adhesive patterns, topographies that are close enough to force neuronal extensions to squeeze between them, and hydrogels providing resistive forces distributed allover the surface of the cell.

Confinement means frontiers or spatial borders. I will review in Section 2 the phenomenon of morphological adaptation to adhesive areas delimited by frontiers with less permissive surfaces. It gives rise to an unexpected wealth of cellular responses supported by a competition between an affinity for sharp adhesion gradients and a limited spread capacity on adhesive surfaces. Once explored the modalities of interaction with frontiers, I will discuss under what conditions and how neurons cross these boundaries to reach nearby adhesive zones by passing over surfaces less favorable to cell adhesion.

The core part of this review (Section 3) describes the generic behavior of enhanced growth rate under confinement, whatever its chemical or physical nature. The strategies of confinement will be sorted by their dimensionality and practical modalities of implementation.

The next section (Section 4) will give some mechanistic clues about the responses to confinement, focusing on the sensing mechanisms of filopodia. Interestingly, confinement provides either restrictions or incentives in filopodia extension, and ultimately settles the balance between both. This section also highlights possible analogies between neuronal and non-neuronal cells which may help decipher, from the vast literature of migration mediated by pulling or pushing forces, the molecular mechanisms at work in neuronal growth in confined situations.

I will finally discuss in Section 5 the link between GC morphology and their dynamics both in vitro and in vivo.

This review will end with a list of open questions regarding cell spreading mechanisms on finite size adhesive areas and growth acceleration under spatial confinement.