Dendritic spines are membranous protrusions off the post synaptic dendrite. These are some ATP requiring fuctions that reside in these bodies.
- The Ca2+ storing endoplasmic retriculum. This particular Wikipedia page Dendritic spines does not spell out specifically SERCA the Ca2+-ATPas that pumps Ca2+ back.
- The Na+ Pump that uses ATP to maintain the resting membrane potential.
- Actin and myosin that also require ATP. Actin remodeling requires ATP.
It would seem that dentritic spines have their own supply of mitochondria to produce ATP for all of the work they do.  More stimulation of dentritic spines leads to more mitochondria.  The Li publication  has some excellent images that will not be reproduced in this post. The featured image is a compilation of many Internet images merged with GIMP2.
Propionic Acid and Mitochondia, new gene transcripts
A neuron like cell line, SH-SY5Y, and normal mouse embryonic neural progenitor cells were exposed for 1, 5 mM PPA for 4 or 24 h.  Key to the mitophagy post, the mitochondrial potential decreased in a dose-dependent manner after PPA treatment. Electron microscopy analysis revealed that the size of the mitochondria was significantly reduced following PPA treatment. Expression of these proteins was increased :
- PGC-1α controls the transcription of many mitochondrial genes, interacts with partner transcription factors CREB and PPAR gamma
- TFAM is the mitochondrial transcription factor
- SIRT3 sirtuin 1 as the NAD+ dependent deacetylase.
- COX4 is a mitochondrial cytochrome C oxidase subunit.
Propionic Acid (PPA) and mitophagy is ERK involved?
A study out of the Mun Lab used rat hippocampal neurons in primary culture. This study started with the premise that ERK signaling is required for the autophagy process. The extracellular signal regulated kinases (ERK) are second messengers that response to growth factors binding to receptor tyrosine kinases a well as ligands of G protein coupled receptors. Much of this signalling results in gene transcription. 
The Num group was interested in autism. A search of ERK kinase and autism in PubMed reveals 86 publications. The Num group started out just showing that PPA decreases the number of denditic spines. Panels 1a and 1b of this publication showed the effects of up to 100mM propionic acid on the medium pH and the viability of up 100mM on cell viability after 72 hours. 
c Top shows representative pyramidal neurons transfected with PLV-GFP and treated with vehicle (PBS) and PPA (100 μM) for 72 h. Bottom shows high magnification of representative secondary dendrites. Scale bar of top panel is 20 μm; scale bar of bottom panel is 1 μm. d Bar graph showing mean ± SEM for dendritic spine numbers of the representative groups (n = 50). ***p < 0.0005, ****p < 0.0001
A lot of the Western blots telling readers with scientific background have not been shown in this post. Panel A is particularly important because it relates to the previous presinillin and mitophagy post.The cleavage product LC3 has previously been reported to occur in response to CCCP that lowers the mitochondria membrane potential. LAMP1 and poly ubiquination are discussed in this previous post. These results almost suggest that PPA might be lowering the mitochondria membrane potential.
2. PPA up regulates biochemical markers of autophagy in hippocampus neurons
- (a) Representative western blot of LC3 after PPA treatment. Note the smaller, faint bands at 14 kDa and the larger band at 16 kDa. According to the UniProt entry,https://www.uniprot.org/uniprotkb/Q9H492/entry LC3i is the cytosolic form. LC3ii is the membrane form that is conjugated to lipids.
- (b) Quantification of LC3-II levels with respect to LC3-I levels (n = 4). More suggests proteolysis.
- c) Representative immunofluorescence images showing LC3 staining in hippocampal neurons and PPA-treated cells. d Quantification of number of LC3 puncta from images in (C) (n = 10). e Representative Western blot showing increased expression of beclin-1 for 72 h after PPA treatment. f Quantification of beclin-1 levels with respect to GAPDH levels (n = 4). g Representative immunofluorescence images showing LAMP1 staining in hippocampal neurons and PPA-treated cells. h Quantification of the number of LAMP1 puncta from images in (G) (n = 12). i Representative Immunofluorescence images showing poly-ubiquitin staining in hippocampal neurons and PPA-treated cells. j Quantification of the number of poly-ubiquitin puncta from the images in (I) (n = 10). Bar graph showing mean ± SEM of the representative groups. Scale bar is 10 μm. *p < 0.05, **p < 0.01, ****p < 0.0001
Figure 3 examined levels of the p62 adaptor protein in terms of total protein as measured by Western blots and in terms of puncta. Both increased by 25% to about 50% of control levels.
Figure 4 , the authors used a red and yellow fluorescent protein LC3B chimeras that allowed pH tracking of pH, be it neutral or acidic. Autophagolysosomes were also measured by scanning electron microscopy. A decrease was seen in autophagosomes, and a bigger decrease in autophaolysosomes. The authors then turned to Bafilomycin A1, a toxin that inhibits the v-ATPase H+ Pump of lysosomes. . After treatment with bafilomycin A1, the number of dendritic spines was 2.161 ± 0.5523, whereas that of untreated neurons was 4.726 ± 0.1715 in 10 μm. The reduction of spine density in bafilomycin A1-treated hippocampal neurons is similar compared with 100μM PPA-treated cells. The conclusion was that PPA dysregulates autolysosome fusion, similar to bafilomycin A1,
Figure 5 showed some Western blots for phosphorylated ERK and Akt, two cell signaling kinases. pAkt did not change in response to PPA whereas pERK increased significantly by about 25%. Figure 6 showed an ERK kinase inhibitor mitigated some of the PPA induced dendritic spine lost as well as increased the number of autophagolysosome vesicles.
Figure 6,.as a finale, demonstrated that the use of the MAPK inhibitor PD98059 rescues PPA treated hippocampal neurons from loss of dendritic spines. 
The Num group mentioned BDNF as the up stream-mediator of autophagy mediated synaptic plasticity in the hippocampus. 
A followup study from the Mun Lab
L-serine is an amino acid produced by astrocytes and is part of our diets. L-serine has shown promise in clinical trials for CNS related conditions. It crosses the blood brain barrier and is safe.  In this study HT22 hippocampus cells were used as well primary hippocampus neurons from embryonic Sprague–Dawley rats. PPA was used at 5 mM and L-serine at 10 mM.
- Pre-screening indicated that PPA did not decrease viability to ~50% until the concentration was increased to 10 mM. L-serine was not toxic at 10 mM.
- PPA essentially doubled the number of abnormal autophagocytic vacuoles and L-serine restored these numbers to something statistically the same as the control.
- PPA increased the number of lipid droplets per cell about 4x. L The total lipid droplet area as well as the lipid droplet size were increased by PPA. L-serine mitigated this increase to just under 3x the control.
- In figure 4 the areas and 3D volumes of these PPA induced lipid droplets, with and without L-serine, were compared. L-serine reduced the volumes and areas by about 50%/
- Figure 5, PPA increased the colocalization of lipid droplets with lysosome markers, and L-serine blocked this increase. Was this a failure of lysosomal acidification and failure to activate phospholipases?
- L-serine (SER) increased lysosomal activity in hippocampal neurons. The Cathepsin D (CTSD) is a lysosomal protease. It’s expression was normalized to normalized to GAPDH. (F) Analysis of Lamp1 protein expression by quantification of lamp1 levels normalized to GAPDH. Data obtained from independent experiments (n = 3 replicates) are presented as the mean ± SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001 by one-way ANOVA, Tukey’s multiple-comparison test.
Not only does L-serine increase the active form of cathepsin D, but it also increases the amounts of pre-pro cathepsin D. It should be noted that cathepsin D is only one of many lysosomal cathepsins. The Mun group was interested in cathepsin D because of a previous report linking CTSD to lipid efflux from lysosomes.  However, another study demonstrated the ability of L-serine to selectively induce the activity of lysosomal enzymes, cathepsins B and L in SH-SY5Y cells.
- Li Z, Okamoto K, Hayashi Y, Sheng M. The importance of dendritic mitochondria in the morphogenesis and plasticity of spines and synapses. Cell. 2004 Dec 17;119(6):873-87. doi: 10.1016/j.cell.2004.11.003. free article
- Kim SA, Jang EH, Mun JY, Choi H. Propionic acid induces mitochondrial dysfunction and affects gene expression for mitochondria biogenesis and neuronal differentiation in SH-SY5Y cell line. Neurotoxicology. 2019 Dec;75:116-122
- Choi H, Kim IS, Mun JY. Propionic acid induces dendritic spine loss by MAPK/ERK signaling and dysregulation of autophagic flux. Mol Brain. 2020 Jun 2;13(1):86 PMC free article
- Jeon H, Kim YJ, Hwang SK, Seo J, Mun JY. Restoration of Cathepsin D Level via L-Serine Attenuates PPA-Induced Lysosomal Dysfunction in Neuronal Cells. Int J Mol Sci. 2022 Sep 13;23(18):10613. PMC free article
- Haidar B., Kiss R.S., Sarov-Blat L., Brunet R., Harder C., McPherson R., Marcel Y.L. Cathepsin d, a lysosomal protease, regulates abca1-mediated lipid efflux*. J. Biol. Chem. 2006;281:39971–39981. doi: 10.1074/jbc.M605095200. [PubMed]
- Dunlop RA, Carney JM. Mechanisms of L-Serine-Mediated Neuroprotection Include Selective Activation of Lysosomal Cathepsins B and L. Neurotox Res. 2021 Feb;39(1):17-26. PubMed