ABSTRACT
In addition to neuronal death and elimination of synapses, the production of transient, exuberant axons, and axonal branches is a general phenomenon in development across species and systems. To understand what drives the decision of which axons are maintained and which are eliminated, it is important to monitor the interaction of juvenile axons at their target. As old and more recent work show, unlike what is claimed by Ribeiro Gomez et al. (2019), in the cerebral cortex, both classes of axons branch in the white matter near the target; axons destined to be maintained massively invade the gray matter where they develop terminal arbors and synapses. Axons destined to elimination remain in the white matter although a few transient, exploratory branches can enter the cortex. Axonal behavior and fate seem dictated by positional information probably conveyed by thalamic afferents and activity. Unlike what is suggested by Ribeiro Gomez et al. (2019), axonal selection should not be confused with synaptic reduction, which is a later event with minor or no impact on the topography of the connection.
Introduction
The paper by Ribeiro Gomes et al. (2019) correctly stresses the importance of the production of exuberant, transient projections in development. The study confirms and extends, in the fetal monkey, previous findings in other species, by reporting exuberant projections to the spinal cord from prefrontal, limbic, temporal, and occipital cortex and from several subcortical structures. The study claims that the transient exuberant axons enter the gray matter of the spinal cord and it assumes this to be the case for cortical exuberant axons as well. This disregards previous work. In doing so, it seems to confuse the elimination of axonal branches with the elimination of transient synapses. These statements prompt some additional clarifications based on old and more recent evidence that I will review in a chronological, topic-oriented order.
The radial reach of transient callosal axons from area 17 (transient axons) and from the 17/18 border in the same contralateral districts during the first and second postnatal week in the cat. Axons of both origins branch in the white matter and project both to areas 17 (which is acallosal in the adult) and to the 17/18 border. Axons from the 17/18 border enter the corresponding contralateral site where they will establish terminal arbors (Aggoun-Zouaoui et al. 1996), and their number increases from the first to the second week. Thickness of arrows is proportional to the number of axons (denoted by labels). Axons from area 17 remain mainly confined to the bottom of the cortex. Only a few axons of either origin target area 17 where they remain confined to the deep cortical layers or to the white matter. The gray matter is divided in three tiers of equal thickness (G1-G3); SP/Wm is the subplate/white matter transition (From Aggoun-Zouaoui and Innocenti 1994, with permission).
Exuberant Axonal Projections in Development
In 1977 using the retrograde transport of HRP, we described a massive loss of callosal projections originating in the primary visual areas (areas 17and 18; V1,V2) of the cat (Innocenti et al. 1977). Since the neurons which once projected into the corpus callosum (CC) were still alive long after the callosal axon was eliminated, the reduction of projections in development appeared to be due to selective axonal elimination rather than to neuronal death (Innocenti 1981; O’Leary and Stanfield 1986; Dehay et al. 1988; De León Reyes et al. 2019). Indeed, the neurons which had a transient callosal axons early in development were found to maintain a local, intracortical axon (Innocenti et al. 1986; De León Reyes et al. 2019). The loss of axonal projection was accompanied by a massive loss of axons in the CC both in cat and in monkey (Berbel and Innocenti 1988; LaMantia and Rakic 1990). Between birth and adulthood, CC axons decrease in the monkey from 200 to 56 million and in the cat from 66 to 31 million. It should be stressed that this is the only case where the elimination of transient axons was directly, quantitatively compared in different species. Thus, the statement that “the magnitude of transient connectivity has been shown to be less prominent in primates than in other species” (De León Reyes et al. 2019) is unjustified. If anything, the opposite is true.
Quantification of branches of different origin and termination during the first and second postnatal week in the cat. The cortex is subdivided as in Fig 1. A large fraction of the branches ends in the WM of the lateral gyrus, far from both areas 17 and 18. Thickness of arrows is proportional to the number of branches (in brackets) (From Aggoun-Zouaoui and Innocenti 1994, with permission).
Schematized summary of the growth of terminating callosal axons in the cat. The diagram on the left is from Bressoud et al. 1999, based on the reconstruction of axons from areas 17 and 18 to contralateral heterotopic visual areas. The diagram on the right is from Innocenti (1995) and is based on the work of Aggoun-Zouaoui and Innocenti (1994) and Aggoun-Zouaoui et al. (1996) based on reconstruction of axons between areas 17 and 18 of the two hemispheres (A: subcortical branching, B: cortical ingrowth, C: cortical branching and synaptogenesis, D: peak of synaptogenesis, E: adult; triangles denote growth cones and dots synaptic boutons). Both the exuberant, transient axons, and those destined to be maintained initially branch in the white matter. Those destined to be maintained grow into the gray matter where they form synapses at radial and tangential locations similar to the adult. The overproduction of synapses occurs after the transient projections have been eliminated and does not deviate significantly from the final distribution.
Time course and temporal relations of developmental events in the CC and visual areas of the cat. The horizontal bar at the top indicates the period of elimination of exuberant callosal projections originating in area 17 as shown by tracers; each vertical line indicates the age of an animal at the time of horseradish peroxidase injection; fast and slow elimination are denoted by heavy and light hatching, respectively. Filled dots, each corresponding to one animal are EM estimates of the total number of axons in the CC, and filled squares are the proportion of myelinated axons (Berbel and Innocenti 1988). Crosses are measurements of the sagittal section area of the CC in the same animals. The remaining graph shows the density of synapses in the visual cortex according to Cragg (1975). Several conclusions can be drawn from these temporal relations: 1. The elimination of callosal axons reported with EM corresponds to, and is probably caused by, the loss of projections shown with tracers. 2. The elimination of callosal axons corresponds to, and probably causes, a pause in the growth of the sagittal section area of the CC which increases again in parallel with myelination. 3. The bulk of axonal elimination occurs before myelination. 4. The bulk of axonal elimination occurs before the fast phase of synaptogenesis, which reaches a peak after elimination has been completed. Therefore, the two events are uncorrelated. Notice that the peak of synaptogenesis may actually be reached at 70 days of age according to Winfield (1981). Reproduced from Innocenti (1992). Comparable temporal relations were found in the monkey where the peak of CC axon number is reached at birth and drops precipitously until 3rd postnatal months largely before myelination (LaMantia and Rakic 1990), while the peak of synaptogenesis is at 2nd-4th postnatal months, and the elimination continues until the 3rd year (Rakic et al. 1986).
The loss of projections in development was found to be a general phenomenon (reviewed in Innocenti and Price 2005). In cats, it included the loss of projections from auditory cortex to the ipsilateral and contralateral visual areas, only a few of which were maintained into adulthood (Innocenti et al. 1988). Before Ribeiro Gomes et al. (2019), transient projections were described in the development of corticospinal projections, in rodents, opossum, and primates (Cabana and Martin 1985; O’Leary and Stanfield 1986; Galea and Darian Smith 1995). The loss of transient projections in development is so widespread across systems and areas that it was suggested to be permissive to cortical evolution (Innocenti 1995, 2017). Supporting this notion was the early discovery of abnormal callosal projections in the Siamese cat (Shatz 1977) as well as the recent discovery of unequivocal callosal projection from the parietal cortex to the striatum found in man but barely visible or absent in monkeys (Innocenti et al. 2016).
Juvenile Axons Reach Their Targets
To better understand the phenomenon, it was important to clarify the behavior of developing axons near their target sites. In an initial study using anterograde transport of WGA-HRP it was clear that, in the visual cortex of the cat, some juvenile CC axons were aiming at area 17, which is acallosal in the adult, but appeared not to enter it to any extent. Instead they entered the border between areas 17 and 18, which maintains callosal connections in the adult (Innocenti 1981). This finding suggested that axonal selection was occurring at the white matter/gray matter interface, possibly by interactions with the cortical subplate, a partially transient structure at the bottom of the gray matter, responsible for guiding geniculo-cortical axons into the gray matter (Gosh et al. 1990; Gosh and Shatz, 1992, reviewed in Kanold and Luhman 2010) and possibly of other aspects of cortical maturation (Kondo et al. 2015).
The resolution afforded by anterograde WGA-HRP being unsatisfactory, a subsequent study using biocytin investigated quantitatively the behavior of single axons originating at the 17/18 border destined to be maintained, and of axons originating in area 17, destined to be eliminated (Aggoun and Innocenti 1994). It was found that both axonal systems branched in the white matter. But only a few branches of the axons to be eliminated entered the gray matter, the majority remained confined to the white matter. In contrast, the axons to be maintained entered massively the cortex and developed terminal arbors (Figs 1–3). Ding and Elberger (1994) similarly found massive axonal projections to the 17/18 border at P0 and only a few axons to area 17, all of which were eliminated by P15. De León Reyes et al. (2019) confirmed the same concept by showing that CC axons originating in layers 2/3 of S1, in the mouse, enter and arborize in the cortex while only a few transient projections originating in layer 4 do, but normally fail to form terminal arbors. They also applied the precaution of injecting the CC, in order not to miss transient axons which would not enter the gray matter. This was apparently not done in other studies leading to doubtful negative findings (Dehay et al. 1988b).
Ribeiro Gomez et al. (2019) seem to be unaware of the work by Aggoun and Innocenti (1994) and by Bressoud and Innocenti (1999) in the cat. Their conclusion that transient axons enter the gray matter of the spinal cord is based on one case with accidental injection of retrograde tracer seemingly restricted to the spinal cord gray matter. This is far from providing compelling evidence since the volume of a tracer injection changes over time (Vanegas et al. 1978) and therefore the region of uptake cannot be precisely determined, particularly after 12–13 days, the survival used in Ribeiro Gomez et al. (2019) study. Also, during survival, ephemeral exploratory branches (see below) might have entered the target, loaded the retrograde tracer and retracted. The statement that transient axons “unequivocally enter the gray matter” in the same paper is based on work by the same laboratory (Dehay et al. 1988), with a low-resolution anterograde tracer (WGA-HRP) injected in the ectosylvian gyrus of the cat to label a transient projection to primary visual areas. A similar study was performed by Innocenti et al. (1988) who also described a few axons entering the gray matter from a transient projection from the auditory cortex. These results are ambiguous since a few neurons still project to the visual cortex in the adult from the deep layers of the ectosylvian cortex (Innocenti et al. 1988).
Axonal Selection at the Target
The conclusion of Ribeiro Gomez et al. is misleading since it fails to recognize the existence of fundamental mechanisms of axonal selection at the target. It also confuses the elimination of transient axonal branches and synapses. Most probably, exuberant axons enter the gray matter to some extent with exploratory branches seeking unsuccessfully for cues which might allow synaptogenesis. Indeed, that is how incoming callosal axons visualized with time-lapse video microscopy behave at the white matter/gray matter border in development (Halloran and Kalil 1994). Although it cannot be excluded that some of the exuberant axons that enter the gray matter may establish some ephemeral electrical interaction with cortical or subcortical neurons via en-passant junctions, they establish transient projections not connections. Indeed, the formation of transient projections should not to be confused with the formation and elimination of transient synapses which is a later event, both in cats and monkeys (Innocenti 1992, 1995; Bressoud and Innocenti 1999; Fig. 4), which results in minor, if any, topographical reorganization (Fig. 3). As shown in Figure 4, in the cat, the bulk of axonal elimination occurs before the fast phase of synaptogenesis, which reaches a peak at about postnatal day 40, after elimination has been completed. Therefore, the two events are uncorrelated. Notice that the peak of synaptogenesis may actually be reached at postnatal day 70 according to Winfield (1981). Comparable temporal relations were found in the monkey where the peak of CC axon number is reached at birth and drops precipitously until 3rd postnatal months largely before myelination (LaMantia and Rakic 1990), while the peak of synaptogenesis is at 2nd-4th postnatal months and the elimination continues until the 3rd year (Rakic et al. 1986).
It remains to be clarified if the selection by axon-target interactions, preceding synaptogenesis, is a general phenomenon applying also to the corticospinal projections described by Ribeiro Gomes (2019) but this is probably the case (Galea and Darian Smith 1995).
What Does It All Mean?
The most plausible explanation for the exuberant development of projections is that axons are driven to the target by incomplete or low-resolution cues (Supplementary Fig. 1). Then two mechanisms of selection occur, one based on positional tags carried by the juvenile axons as suggested by transplantation experiments (O’Leary and Stanfield 1989), possibly conveyed by thalamocortical afferents (Shatz 1977; De León Reyes 2019). The second one is based on activity (Innocenti and Frost 1979; Innocenti 1980; Zufferey et al. 1999). In the case of peripheral or central lesions, the interplay of these mechanisms can lead to the stabilization of connections normally eliminated (Innocenti and Frost 1979, Innocenti et al. 1988, Restrepo et al. 2003; De León Reyes et al. 2019; Mikellidou et al. 2019).
The need of activity to stabilize juvenile connections is important since it explains the massive regression of CC connections between primary visual areas in conditions of visual deprivation and might also explain the consequences of other types of deprivation in humans (Innocenti 2007).
