The authors have addressed an interesting question regarding the trafficking of the water channel AQP4 in astrocytes. When localised at the cell surface these channels increase water influx into astrocytes and contribute to the tightly controlled water homeostasis in the central nervous system which is essential in particular during brain swelling. The authors have mainly used human embryonic kidney cells for their experiments, an organ where AQP4 is also expressed, and have validated their experiments in human astrocytes.

Major points
• Page 8, line 17. To investigate dynamin-dependent internalisation of AQP4 it is important to not only rely on the inhibitor Dynasore. Dynasore has been shown to not be specific for the dynamin-1/2 GTPase domain but for example unspecifically reduces the cholesterol content of the plasma membrane (PMID: 25889964). Endocytosis of AQP2 is inhibited by cholesterol depletion. Here it is therefore essential to use the more specific inhibitor Dyngo-4a. In addition, the authors should consider using the established dynamin K44A mutant to validate the data generated with the pharmacological inhibitor.
• Page 8, line 20. The authors describe dynamin as a protein “mostly associated with clathrin-dependent endocytosis”. This is not correct and the experiments should be interpreted using up-to-date literature. Dynamin-dependent pathways are not limited to clathrin-mediated endocytosis, but includes a number of pathways such as caveolae, FEME/fast endocytosis, micropinocytosis, clathrin-independent receptor-mediated endocytosis. This should be reflected in the text. To investigate clathrin-mediated endocytosis the assembly of the clathrin coat should be perturbed with either an inhibitor (pitstop2) or by expression of a clathrin-sequestering protein domain such as AP180 C-terminus. It is not sufficient to inhibit dynamin since it is not specific to this pathway.
• Page 8, line 20. Filipin disrupts cholesterol-rich rafts at the plasma membrane and the Golgi but does not impair all clathrin-independent internalization pathways. The authors should take this into consideration in their interpretation of their experiments in Figure 1C.
• Figure 3. To assess co-localization of early endosomes and recycling endosomes Rab5 and Rab11 were used as markers. The fluorescence of these markers do not display the expected punctate pattern (Figure 3C, Supplementary Figure 3), but rather shows a cytoplasmic localization. Since the images do not have the required quality to analyse co-localization with endosomes they cannot be used to conclude that AQP4 colocalizes with endosomes. Analysis of higher resolution images with reduced background fluorescence is required, ideally backed up by western blots of purified organelles.

Minor points
• Please cite the relevant literature in the introduction.
Lisjak et al.2020, PMID: 32192013. In astrocytes AQP4 co-localizes with early endosomes, which is enhanced by hypotonic conditions.
Tham et al. 2016, PMID: 27788222. AQP4 internalization is regulated by dynamin, and the trafficking to the channel to the cell surface involves recruitment from a pool in the early endosome. The dynaminK44A mutant was used in this study.
• Please add a couple of sentences in the introduction about expression of AQP4 in the kidney to provide a context for the experiments done in the HEK293 cell line.
• Please include the supplier for all reagents used in the study, eg. Filipin, in the Materials and methods.
• Transient transfections were performed with PEI (page 7, line 36). Please specify whether the same protocol was used for both cell lines. Is it correct that a 1:12 transfection ratio DNA:PEI was used? A 1:3 transfection mix for HEK293 cells works very well. An increased PEI concentration can result in cell death and unspecific effects to trafficking pathways.
• For microscope objectives please add the numerical aperture of your lens and the specifications of the camera that was used to collect the images. Page 7, line 54
• Please define what the hypotonic media contained in addition to 25% DMEM. Page8, line 10.
• Cholesterol depletion inhibits endocytosis of AQP2 and furthermore strong cholesterol depletion results in delayed AQP2 exit from the trans-Golgi network. The authors should discuss this in relation to their experiments where cholesterol has been depleted by either Filipin or Dynasore.

1. The authors have used Dyngo-4a (Figure S6A) to inhibit dynamin oligomerisation and Pitstop2 to inhibit clathrin polymerisation (Figure S6b to assess endocytosis of AQP4 from the plasma membrane, which has strengthened the conclusions of the manuscript.

2. The authors have added relevant context to the filipin experiments that describes that the drug inhibits both trafficking to the plasma membrane and from the plasma membrane.

3. In the original manuscript Figure 3 and Supplementary Figure 3 showed fluorescence images of poor quality, which were not suited for co-localisation analysis. The quality of Figure 3 has been improved and a scale bar has been added. Supplementary figure 3 is still very pixelated, so much so that the scale bar is not resolved. Most concerning, no punctate endosome pattern is visible against the background, which is the marker that is used in the co-localisation analysis (Supplementary Figure 4). In its current form Supplementary Figure 3, and the associated Supplementary Figure 4, do not have the required resolution and quality for publication.

4. All minor points have been addressed.