Loss of Dendritic Complexity Precedes Neurodegeneration in a Mouse Model with Disrupted Mitochondrial Distribution in Mature Dendrites

1Although only 2% of the body’s weight, the brain consumes 20% of the body’s resting energy production, which may rise to 50% in the developing brain. Energy in the adult brain is mainly used to reverse the ion influxes underlying synaptic and action potential signaling, for pumping neurotransmitters and for supporting signal transduction and trafficking events during synaptic plasticity. During brain development energy may also be required to fuel the extensive neuronal growth necessary to build the complex neuronal morphologies that are essential for the formation of neuronal circuits and for brain computation. While changes in dendritic complexity have been reported to occur in a number of neurological and neurodegenerative diseases or during normal aging, the energy costs and cellular mechanisms underlying dendritic sustainability are poorly understood.

2Mitochondria are critical for ATP provision and play other essential roles in neurons such as buffering calcium. The very large size of many neurons suggests that mitochondrial distribution must be spatially matched to local energy usage and calcium buffering requirements during the growth and maintenance of axons and dendrites. Indeed, alterations in mitochondrial dynamics compromise synaptic function, and defective mitochondrial trafficking has been related to neurological and neurodegenerative disease. Therefore, mitochondrial distribution and trafficking may be key determinants for both the generation and the long-term maintenance of the complex neuronal morphologies essential for brain information processing. Currently, however, the relationship between mitochondrial distribution and the morphogenesis and maintenance of neuronal architecture in vivo remains unclear.

3Miro (mitochondrial Rho) GTPases are central regulators of mitochondrial trafficking that act as adaptors, linking mitochondria to motor proteins. Here, we use mouse genetics to investigate the differential role of Miro1 and Miro2 for mitochondrial trafficking, neuronal morphogenesis, and the maintenance of complex neuronal architecture. We show that Miro1 is the main regulator of mitochondrial trafficking and distribution. Depleting Miro1 levels in vivo during development disrupts neuronal morphogenesis, while postnatal Miro1 disruption in mature neurons leads to a loss of dendritic complexity starting in the distal dendrites. Our findings demonstrate that altering mitochondrial positioning in dendrites leads to a progressive loss of dendritic complexity, which, in turn, triggers neuronal death in a way that closely resembles a neurodegenerative process.

4To study the roles of Miro1 and Miro2, we characterized constitutive mouse knockout strains for Rhot1 (Miro1 gene) and Rhot2 (Miro2 gene) (Figures S1A–S1D). Protein levels within brain lysates confirmed the specific deletion of Miro1 and Miro2 proteins, respectively (Figure S1E). Rhot2 knockout animals (Miro2KO hereafter) were viable and fertile, whereas Rhot1 knockout animals (Miro1KO) were born alive at the expected Mendelian ratios but remained cyanotic and died within the first 15 to 30 min of life.

5To address the specific roles of Miro1 and Miro2 for mitochondrial trafficking, we compared hippocampal neuronal cultures from individual wild-type (WT) or knockout E16 embryos generated by heterozygous (Miro1+/− X Miro1+/− or Miro2+/− X Miro2+/−) matings. In both 6–7 days in vitro (DIV) and 14–15 DIV Miro1KO neurons expressing GFP to fill the cell and MtdsRed2 to label mitochondria, anterograde and retrograde mitochondrial trafficking was altered in dendrites (∼85% decrease) and in axons (∼65% decrease) (Figures 1 and S3A–S3D). The velocity of the remaining motile mitochondria at 14–15 DIV was unaltered in axons but was reduced by ∼50% in the anterograde direction within dendrites whereas retrograde velocity was unaffected (Figures 1E, 1K, and 1M). Interestingly, the small amount of mitochondrial trafficking that remained was still sensitive to neuronal activation induced by glutamate indicating that other mechanisms exist that can sense Ca2+ and induce Miro1-independent mitochondrial stopping (Figures S1F–S1I). Mitochondrial trafficking was fully rescued by expression of Miro1-myc (Figure 1) but only partially by Miro2 expression (Figures S2A and S2B). Unexpectedly, Miro2 deletion (Miro2KO neurons) had no substantial effect on mitochondrial trafficking (Figures 1 and S1J) suggesting Miro2 is not the main regulator of mitochondrial trafficking or that its function can be compensated by Miro1. Moreover, the trafficking of Rab5GFP positive early endosomes (Figure S1K) and axonal retrograde transport of Rab7GFP positive signaling endosomes was unaffected in Miro1KO neurons (Figures S1L–S1N) further confirming the critical importance and specificity of Miro1 deletion for mitochondrial trafficking.

6We found that in Miro1KO neurons the vast majority of mitochondria were accumulated in proximal regions of dendrites and only sparsely distributed within distal dendrites, with large dendritic segments almost entirely devoid of mitochondria (Figures 2A and 2B). To quantify this effect, we performed a mitochondrial “Sholl” distribution analysis (see Experimental Procedures), to give a “mitochondrial probability map” (MPM): the cumulative probability plot of mitochondrial distribution as a function of distance from the soma. The MPM shows a clear shift to the left in the curve for Miro1KO neurons at both 6 DIV and 14–15 DIV, confirming an accumulation of mitochondria in the proximal dendrites of Miro1KO cells and reduction in distal dendrites (Figures 2C and S3E). To account for variability in neuronal process length, we defined a normalized value, Mito60, the point at which 60% of the total mitochondrial mass is found within a normalized proximal to distal dendritic axis (where the dendritic tip is defined as 1). In Miro1KO neurons the Mito60 value was decreased by ∼40% at both 14–15 DIV (p = 0.003) and 6–7 DIV (p = 0.035), confirming a more accumulated distribution of mitochondria within proximal dendrites (Figures 2D and S3F). In contrast, in Miro2KO neurons the MPM and Mito60 values were unaltered (Figures S2C and S2D, p = 0.16). Importantly, mitochondria accumulated in proximal Miro1KO dendrites were metabolically active and kept their mitochondrial membrane potential, as revealed by tetramethylrhodamine methyl ester (TMRM) staining (Figures S3K and S3L). Thus, Miro1 control of mitochondrial trafficking is a key determinant of correct mitochondrial distribution within dendrites—a function that cannot be readily compensated by endogenous Miro2.

7Neurons lacking Miro1 appeared smaller and less developed than control neurons at 14–15 DIV (Figure 2A). Total dendritic length was ∼35% decreased in Miro1KO neurons (Figure 2E, WT 3793 ± 89 μm, Miro1KO 2447 ± 54 μm, p = 0.0003), although surprisingly the total number of branchpoints was unaffected (Figure 2F, WT 55.7 ± 2.78, Miro1KO 57.9 ± 3.53, p = 0.63). Dendritic Sholl analysis of Miro1KO neurons also revealed a marked increase in dendritic complexity in proximal regions and a decrease distally, closely correlating with the observed changes in mitochondrial distribution (Figure 2G). To account for changes in neuronal size, we defined a value that represented where 90% of the branchpoints are found within size-normalized cells (BP90). The BP90 for Miro1KO neurons is significantly less than for WT neurons as Miro1KO neurons have significantly more branchpoints near the soma (Figure 2H). In contrast Miro2KO neurons appeared similar to WT cells in terms of dendritic architecture (Figures S2E–S2H; dendritic length, p = 0.85; total branchpoints, p = 0.62) indicating that in the presence of Miro1, Miro2 is not required for correct dendritic development. Younger Miro1KO neurons (6–7 DIV) were similar in terms of dendritic length and number of primary dendrites and branchpoints compared to WT (Figures S3G–S3I). Interestingly, Sholl analysis demonstrated that they were already beginning to show less distal complexity (120 μm outward from the soma) compared to WT neurons (Figure S3J). Moreover, reconstructions of Golgi-stained neurons (Figure 2I) located in the lower levels of the cortical plate of Miro1KO brains showed a clear reduction in neurite length (p = 0.0009) and branchpoint number (p = 0.0005) (Figures 2J–2L) at an early stage of cortex development (at E18, when Miro1KO animals are viable) supporting a role for Miro1 in regulating neuronal morphogenesis in vivo. Thus local availability of mitochondrial mass is essential for correct neuronal process development.

8To test whether the loss of Miro1 altered mitochondrial distribution in vivo and impacts postnatal dendritic maintenance, we generated a conditional Miro1 allele (Miro1lox/lox, see Figure S1A) and crossed this with a strain expressing Cre recombinase under the Ca2+/calmodulin-dependent protein kinase-alpha specific promoter (CaMKIIα-Cre), which has been widely used to drive Cre-mediated genetic deletion in postnatal forebrain. Miro1Δ/Δ,Cre(+/−) animals (hereafter Miro1CKO) developed normally and appeared indistinguishable from their Miro1lox/lox,Cre(−/−) (WT hereafter) control littermates. Specific deletion of the Miro1 gene in Miro1CKO animals occurred primarily in adult principal neurons of the cortex and hippocampus (Figure S4A). Interestingly, whereas, in WT animals Miro2 levels were progressively decreased over time (Figure S4B), this was not observed in brains deleted for Miro1 (Figures S4D and S4E) suggesting that Miro2 may be upregulated to counteract the depletion of Miro1.

9By immunostaining with the mitochondrial marker Tom20 we observed in 4-month-old Miro1CKO brains a decrease in the mitochondrial index (mitochondrial area/MAP2 positive dendritic area) in distal regions of the CA1 neurons corresponding to the Stratum Lacunosum-Moleculare (SLM), compared to WT (Figures 3A and 3B, p = 0.043; t test). At 8 months, Tom20 staining revealed an even more enhanced loss of mitochondrial density in Miro1CKO CA1 distal dendrites (Figures 3A and 3B; p < 0.001; t test). To ensure that the loss in mitochondrial density was specifically in the dendrites and not in the axonal projections that populate the SLM region, we performed intracerebral injections of non-replicating Adeno-Associated Virus (AAV2) encoding MtDsRed2 and GFP into the CA1-CA3 region of the hippocampus. Neurons from control animals showed mitochondria distributed homogeneously throughout the dendritic length, whereas Miro1CKO neurons presented long segments of distal dendrites devoid of mitochondria (Figures 3C and 3D, percentage of dendritic length without mitochondria; WT 25% ± 2%, Miro1CKO 65% ± 5%; p < 0.001, t test), indicating that, also in vivo, Miro1 regulates mitochondrial distribution in mature dendrites.

10Although dendritic mitochondrial distribution was already affected at 4 months of age, no defects in dendritic complexity were apparent when comparing Golgi-stained CA1 pyramidal neurons from WT and Miro1CKO brains (Figures S5B–S5E: dendritic length: basal; WT 1321 ± 42 μm, Miro1CKO 1394 ± 69 μm, apical; WT 1183 ± 31 μm, Miro1CKO 1129 ± 86 μm; branchpoints: basal; WT 16.7 ± 0.5, Miro1CKO 15.6 ± 0.8, apical; WT 16.3 ± 0.7, Miro1CKO 14.8 ± 1.1; ANOVA and post hoc Newman-Keuls [ANOVA-NK]). In contrast, at 8 months, Miro1CKO CA1 neurons showed a significant decrease in total dendritic length in both the basal and apical dendrites, (Figures 3E and 3G, basal: WT 1307 ± 100 μm, Miro1CKO 889 ± 83 μm, p < 0.01; apical: WT 1686 ± 183 μm, Miro1CKO 1155 ± 172 μm, p < 0.05; ANOVA-NK). Interestingly, while the number of branchpoints was similar to that of WT neurons (Figure 3H), Sholl analysis revealed more branchpoints in the proximal region of the apical dendrite and reduced branching in the distal regions in Miro1CKO neurons (Figures 3F, 3I, and 3J). Age-matched Miro2KO neurons showed no morphological defects (Figures 3G and 3H). The loss of dendritic complexity at 8 months in Miro1CKO animals correlated with a clear decrease in brain size compared to WT (Figure S4C) and an increase in the size of the ventricles with a reduction in hippocampal size (Figures 3K and 3L, pial to apical surface distance of hippocampus: WT 516.8 ± 9.7 μm, Miro1CKO 363.3 ± 26.7 μm, p < 0.01; and Miro2KO 517.8 ± 14.9 μm, ANOVA-NK) and thickness of the cortex (Figures 3K and 3M, WT 960.4 ± 35.1 μm, Miro1CKO 824.3 ± 24.4 μm, p < 0.05; and Miro2KO 946.1 ± 35.4 μm, ANOVA-NK).

11Therefore, the postnatal loss of Miro1 disrupts mitochondrial distribution in the proximal to distal dendritic axis leading to a gradual decrease in dendritic complexity particularly affecting the most distal dendrites.

12As opposed to dendrites, mitochondria distributed homogeneously throughout axons of Miro1KO cultured hippocampal neurons (Figure 4A). A plot of the relative mitochondrial positions along WT and Miro1KO axons from 6-DIV neurons, transfected with GFP and MtdsRed2 and normalized for changes in length (Figure 4B), showed that mitochondria are distributed throughout the entire length of the axon, despite the fact that fast axonal mitochondrial trafficking is severely altered in Miro1KO neurons. This suggests that other mechanisms can compensate the absence of Miro1 for mitochondrial distribution within the axon. Although mitochondria distributed homogeneously, Miro1KO axons were shorter compared to WT (Figure 4C; p = 0.004, t test). As mitochondria were observed at the tip of the axon (approximately 650 μm from the soma) within both WT and Miro1KO processes, the defects in axonal growth are not due to the lack of mitochondria in the most distal regions of the axon.

13To test these observations in vivo, we analyzed the Stratum Oriens (SO) or Stratum Radiatum (SR) from the contralateral hippocampus of our AAV-injected mouse brains, which revealed that axonal mitochondrial distribution from the CA3 neurons transduced with GFP and MtdsRed2 (Figure 4D) is unaffected by Miro1 deletion even at 8 months of age (Figures 4E and 4F; number of mitochondria/10-μm axon: WT 2.31 ± 0.13, Miro1CKO 2.18 ± 0.13; p = 0.505, t test), while changes in mitochondrial distribution in dendrites were evident much earlier, at 4 months. This analysis was performed in regions of axons up to several millimeters away from the somas in the contralateral hippocampus demonstrating that the distribution of mitochondria within axons and dendrites differentially rely on Miro1 function.

14In Miro1CKO animals at 8 months of age, when a significant reorganization of dendritic architecture and a decrease in cortex and hippocampal size is observed, there is no significant loss of neurons, detected by quantification of NeuN staining in these regions (Figures 5A, 5B, and S5F). However, the loss of dendritic complexity at this time point was accompanied by a 2-fold increase in GFAP signal (p < 0.01) and infiltration of reactive astrocytes, possibly as a protection mechanism to scavenge and clear cellular material product from the initial distal dendritic loss (Figures 5A, 5E, and 5F). At a later time point, 12 months, Miro1CKO animals showed an enhanced loss of the dendritic tree in CA1 pyramidal neurons and a marked decrease in the number of neurons compared to WT brains (Figures 5C, 5D, 5G, and S5F).

15Accordingly, the absence of Miro1-mediated mitochondrial positioning leads to a progressive dendritic loss that eventually results in neuronal degeneration.

16Using constitutive and conditional mouse knockout strategies we demonstrate that Miro1 is the main regulator of mitochondrial transport and localization, important for ensuring correct dendritic development, maintenance, and neuronal survival. Miro1KO neurons have less anterogradely and retrogradely moving mitochondria in both dendrites and axons, whereas knocking out Miro2 has no significant effect suggesting that either Miro1 can compensate for loss of Miro2 or that Miro2 has evolved other roles for mitochondrial function. Loss of Miro1 leads to accumulation of mitochondria in proximal dendrites, resulting in neurons with shorter dendrites but with significantly increased branching within proximal regions. Several earlier studies have supported links between mitochondrial function and neuronal development. Disrupting the mitochondrial fission and fusion machinery (e.g., by targeting mitofusins or Drp1) disrupts dendritic growth, but it is hard to separate effects on mitochondrial dynamics, function, and trafficking in these experiments, given that targeting the fission/fusion machinery can lead to depolarized mitochondria with altered genetic content. By directly targeting the Miro trafficking machinery, we have avoided the potential non-mitochondrial effects of targeting the TRAKs, the kinesin/dynein complexes, or artificially clustering mitochondria with the Golgi, which also has a critical role in dendritic morphogenesis. Interestingly, a similar effect is seen in vivo upon conditional CaMKIIα-Cre-driven, postnatal deletion of Miro1 in hippocampus and cortex demonstrating that the relationship between mitochondrial positioning and dendritic branch distribution plays a direct role in sculpting and maintaining mature neuronal circuits.

17Remarkably mitochondrial distribution into axons is preserved in cultured neurons that lack Miro1 despite a notable 65% decrease of fast axonal mitochondrial trafficking. This is in stark contrast to the significant depletion of mitochondria from distal dendrites (which have a similar microtubule polarization as axons) suggesting that different mechanisms are involved in the localization of mitochondria within axons and dendrites independent on microtubule orientation. We show that Miro2 overexpression can, to a small extent, rescue axonal mitochondrial trafficking in Miro1KO cultured neurons. The increase in Miro2 protein levels in Miro1CKO hippocampus may compensate the lack of Miro1 in these neurons and could explain the slow progression of the neurodegeneration phenotype and/or why Miro1CKO axons are still populated with mitochondria. Coupling of mitochondria to cytoskeletal transport pathways via a Miro1-independent mechanism may also still occur, for example via Miro2 or an unknown TRAK1 adaptor or through the mitochondrial motor Kif1b. The small amount of remaining mitochondria transported in Miro1KO axons may be anchored by axonal syntaphilin or by mitochondrial presynaptic capture. The decreased motility of stabilized mitochondria in mutant axons may nonetheless impact mitochondrial dynamics and turnover resulting in mitochondria being less able to support axonal growth or branching.

18Conditional deletion of Miro1 by enolase2 promoter-driven Cre was recently proposed to cause axonal pathology in juvenile mice similar to an upper motor neuron disease. This genetic lesion is expected to affect multiple populations of neurons in brain, spinal cord, and peripheral nervous system starting as early in development as E13.5. In contrast, here we show that conditional deletion of Miro1 in mature neurons drives the loss of dendritic complexity, which is clearly evident 8 months after birth and is followed by neuronal loss (that is evident at 12 months).

19A number of neurological and neurodegenerative diseases have been related to mitochondrial dysfunction and disruption of mitochondrial trafficking, but the mechanisms underlying the pathophysiology of such disorders have frequently been difficult to identify. Axons are up to 104 times longer than dendrites and have often been the focus of attention when studying neurological diseases associated with mutations in transport components. Our study cannot rule out an axonal component participating in the neurodegenerative phenotype observed in Miro1CKO animals. However, our Rab7GFP trafficking experiments suggesting that trophic support is not affected and our analysis of mitochondrial distribution in axons from 8-month-old Miro1CKO animals suggest that an axonal pathology is unlikely the origin of the degeneration phenotype. The idea that alterations in axonal transport (mitochondrial and other organelles) are either a necessary or sufficient causative factor in the neurodegenerative progression has recently been challenged, supporting the idea that axonal transport defects and axonal degeneration may course independently of each other. It is well known, however, that the loss of dendritic complexity is a hallmark of a number of neurological and neurodegenerative diseases. Furthermore, recent work in a mouse model of Alzheimer’s disease establishes a mechanistic link between dendritic structural degeneration and neuronal hyperexcitability resulting from changes in the electrical properties of the affected neurons and that may apply to other neurodegenerative diseases where changes in dendritic morphology occur.

20Altogether we demonstrate that Miro1 function is critical in regulating the development of the dendritic tree and in maintaining and sustaining arborization. Mature neurons lacking Miro1 inevitably lose their dendritic complexity and consequently their connectivity rendering them vulnerable to undergo a neurodegenerative process. With our work, we highlight the imperative need of a correct mitochondrial distribution in dendrites to sustain their architectural complexity and to protect against neurodegeneration. This pathomechanism may contribute to the neurodegenerative outcome of many human neurological diseases providing a cellular target of intervention for future therapeutic strategies.

21Stereotaxic viral injections were performed on 3- and 7-month-old mice as described previously. Anesthetized animals were mounted on a stereotaxic apparatus, and the brain was exposed to 7 × 40-nl measures of AAV2-CAG-MtdsRed-ires-GFP (titer = 1.2 × 1012 genome copies [GC]/mL, UPenn Vector Core) injected at coordinates relative to Bregma (Medial/Lateral, Dorsal/Ventral, Rostral/Caudal; −1.4, −1.5 to −0.5, −1.8). Expression occurred in the injected brain region for approximately 4 weeks until tissues were obtained (Figure 4D). All experimental procedures were carried out in accordance with institutional animal welfare guidelines and licensed by the UK Home Office in accordance with the Animals (Scientific Procedures) Act 1986.

22Golgi staining of neurons was performed using the Rapid Golgi Stain Kit (FD NeuroTechnologies) following the manufacturer’s protocol. Neurons were imaged using a 20× objective and analyzed using Neurolucida software (MBF Bioscience).

23Mitochondria or Rab7GFP vesicles in neurons were live imaged with an Olympus BX60M upright microscope and 63× objective. Movies were processed in ImageJ, and transport parameters were obtained from the kymographs as previously described.

24Confocal images were acquired on a Zeiss LSM700 upright confocal microscope using a 63× objective. Dendrite tracing and analysis were performed in Neuronstudio. Mitochondrial-Sholl analysis was performed using a custom ImageJ plugin, which quantified the amount of MtdsRed2 pixels within shells radiating out from the soma at one-pixel intervals.

25Excel software (Microsoft) and GraphPad Prism (GraphPad) were used to analyze the data. Student’s t test or Mann-Whitney test was used to test differences between two conditions. Comparison of multiple conditions with normally distributed data was performed by one-way ANOVA followed by post hoc Newman-Keuls test and for non-parametric data by Kruskal-Wallis test followed by post hoc Dunn’s correction. Statistical significance was fixed at p < 0.05, represented as ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001. All values in text are given as average ± SEM. Error bars are SEM.

26Additional details regarding animal strains, stereotaxic brain infections, neuronal culture and transfection, histology, immunoblotting, immunofluorescence, and image acquisition and analysis can be found in the Supplemental Experimental Procedures section.

27G.L.-D., N.F.H., A.F.M., and J.T.K. designed the project and the experiments; G.L.-D., N.F.H., V.V., H.R., I.L.A.-C., and A.F.M. performed experiments. G.L.-D., N.F.H., and V.V. analyzed the data. G.L.-D., N.F.H., and J.T.K. wrote the manuscript.