Profilin 1 deficiency drives mitotic defects and reduces genome stability

Profilin 1—encoded by PFN1—is a small actin-binding protein with a tumour suppressive role in various adenocarcinomas and pagetic osteosarcomas. However, its contribution to tumour development is not fully understood. Using fix and live cell imaging, we report that Profilin 1 inactivation results in multiple mitotic defects, manifested prominently by anaphase bridges, multipolar spindles, misaligned and lagging chromosomes, and cytokinesis failures. Accordingly, next-generation sequencing technologies highlighted that Profilin 1 knock-out cells display extensive copy-number alterations, which are associated with complex genome rearrangements and chromothripsis events in primary pagetic osteosarcomas with Profilin 1 inactivation. Mechanistically, we show that Profilin 1 is recruited to the spindle midzone at anaphase, and its deficiency reduces the supply of actin filaments to the cleavage furrow during cytokinesis. The mitotic defects are also observed in mouse embryonic fibroblasts and mesenchymal cells deriving from a newly generated knock-in mouse model harbouring a Pfn1 loss-of-function mutation. Furthermore, nuclear atypia is also detected in histological sections of mutant femurs. Thus, our results indicate that Profilin 1 has a role in regulating cell division, and its inactivation triggers mitotic defects, one of the major mechanisms through which tumour cells acquire chromosomal instability.

1. It is clear that reduced PFN1 results in lowered actin in the cleavage furrow, and increased chromosome instability. What is the effect of PFN1 overexpression in a wild-type background, and could PFN1 deficiency be corrected by re-expressing PFN1 (for example by transfection)? 2. The localization of PFN1 during mitosis was shown only in RPE1 cells. It would be interesting to see the location of PFN1 in additional human cells, as well as the location of Pfn1 in the mouse cells used in the study. 3. Figure 5i,j shows increased p53 in Pfn1+/-mouse cells. This is under chronic conditions where Pfn1 levels are constitutively reduced. What would be the effect on p53 immediately after reduction of Pfn1 (for example in the shRNA RPE-1 cells). 4. If Pfn1 reduction results in aberrant rounding of mitotic cells, perhaps the authors could examine the morphology of mitotic cells when grown on different substrates (matrigel, or in low adherence 1. Perhaps I missed it, but the frequency of micronuclei formation in +/+, +/-, -/-PFN1 RPE-1 cells is not shown. 2. Line 174: "suggesting that a p53-deficient background would enhance mitotic dysfunction and tumorigenicity". "would" should be replaced with "could". 3. In general, why not examine the role of Pfn1 reduction on human fibroblasts -a more relevant model to OS than RPE-1? 4. Line 218: instead of "replication" should be "mitotic/division". 5. The level of chromosome instability in mouse tissues from Pfn1+/-would be more relevant than the measurements (low pass sequencing) performed on cultured mouse clones, as they would better reflect the effect of reduced Pfn1 on mitotic fidelity in vivo. One potential approach could be isolation of single cells from Pfn1+/-and Pfn1 +/+ mouse tissues, performing single cell DNA sequencing to score for chromosome copy number changes, and comparing the wild-type with the heterozygote. This would, obviously, require more efforts, and so if not included in the current work, the authors should consider adding such an experiment to future studies, to substantiate the Pfn1+/-mice as a CIN model. Nevertheless, the authors should include WT controls to their analysis in supplementary figure  9, and indicate the level of gain/loss of each chromosome. 6. Line 248: the authors refer to high level copy number amplifications but do not provide further details explaining what genes were amplified, and the level of amplification. 7. In the discussion, line 271: "We used both CRISPR/Cas9 and shRNA experiments to inactivate PFN1 expression in RPE1 cells, which are widely used to study mitotic defects as well as chromosomal rearrangements in a p53-deficient background". This could confuse the reader into thinking that p53-/-cells were used in this study as well (which were not). 8. In Figure 5 KO should be HET in the different panels as these are Pfn1+/-derived. 9. In supplementary Figure 5 there is no indication for % of micronuclei, although the text suggests there is. 10. The relevance of the OD/PDB patient sequencing data is not immediately clear. There is no clear comparison between samples that have PFN1 LOH and ones that do not. Could this analysis be extended using already published sequences from human cancers? It would be helpful if the authors could provide more context and explain what can be learned from this sequencing efforts with regard to PFN1 LOH. Also, the ability to determine WGD relative to PFN1 LOH is interesting, and I would encourage making this analysis more accessible to the non-computational reader (for example, "WGD relative timing" is not a clear parameter). 11. The use of the phrase "nuclear shape" can mislead to think there is something wrong with the nuclear lamina, or membrane, when in fact the authors refer to micronuclei which are separate from primary nuclei. I recommend not using this terminology. 12. It would be interesting to look at the actin filaments in mitotic PFN1+/-(compared to WT) cells under an electron microscope. We added WT-associated data in Figure 6, while for the other figures we added in the legend that control references could be found within other figures. For example, we added in the legend to Figure 4 the following sentence: "Control metaphase, anaphase, and telophase images of dividing RPE1 cells are illustrated in Figure 1."

Given that Pfn1 also interacts with microtubules (a critical player for mitosis) utilizing residues that are distinct from those required for actin and poly-prolinebinding, it would be important to perform rescue experiments in KO cells with either wild-type or mutants of Pfn1 that are impaired in binding to various ligands and assess some of the mitosis and chromosomal phenotypes.
We think this is an excellent suggestion and, in the attempt to address the Reviewer's concern, we have uncovered unexpected results about Profilin 1, which we summarise here. We have generated stable cell clones of WT, PFN1 +/-, and PFN1 -/-RPE1, each overexpressing either PFN1 WT

We gladly accepted the Reviewer's comment and performed Profilin 1 immunofluorescence in additional non-transformed cell lines, including the MC3T3 cells (used in this study), human dermal fibroblast derived from a control individual, and HK-2 cells (human kidney cells). We interestingly confirmed the enrichment of
Profilin 1 in the spindle midzone of these cell lines, using two different antibodies, and hence modified the Results and the Methods sections accordingly. These images have been included as new Supplementary Figure 2. 3. Figure 5i,j shows increased p53 in Pfn1+/-mouse cells. This is under chronic conditions where Pfn1 levels are constitutively reduced. What would be the effect on p53 immediately after reduction of Pfn1 (for example in the shRNA RPE-1 cells). We added the detailed methods of cell culture on collagen-coated coverslips and scoring of aberrant mitotic rounding in the Methods section.

If Pfn1 levels affect genome doubling, could the authors measure the levels of tetraploidy in the different culture systems (using a simple PI staining in FACS, compared to a diploid control)?
To measure the levels of tetraploidy in Profilin 1-deficient cells, we analysed the ploidy content of PFN1 -/-RPE1 clones compared with WT cells, through FACS analysis of propidium iodide incorporation. To discriminate between tetraploid peaks and cells with duplicated DNA (G2/M), cells were serum-starved in G0/G1 phase of cell cycle. To accurately identify the G0/G1 diploid peak position, we used DNA from human peripheral blood mononuclear cells (PBMCs) as a diploid internal standard. We found that 4.5 ± 0.3% of PFN1-null cells were tetraploid, which is quite in agreement with the frequency of cytokinesis failures detected. We included this result as Supplementary Figure 3b (previously, Supplementary Figure 2). To further exclude that those could represent KO cells that escaped serum starvation, aliquots of the same cell populations were every time subjected to immunofluorescence detection of phospho-histone H3, a mitotic marker. We detected a definitely negligible amount of pH3-positive cells, which is comparable to what observed in wild type cells as 4n peak.
Minor comments:

Line 218: instead of "replication" should be "mitotic/division".
We agree and we have revised it accordingly.
5. The level of chromosome instability in mouse tissues from Pfn1+/-would be more relevant than the measurements (low pass sequencing) performed on cultured mouse clones, as they would better reflect the effect of reduced Pfn1 on mitotic fidelity in vivo. One potential approach could be isolation of single cells from Pfn1+/-and Pfn1 +/+ mouse tissues, performing single cell DNA sequencing to score for chromosome copy number changes, and comparing the wild-type with the heterozygote. This would, obviously, require more efforts, and so if not included in the current work, the authors should consider adding such an experiment to future studies, to substantiate the Pfn1+/-mice as a CIN model. Thank you for this suggestion. We find it very helpful. However, as also supposed by the Reviewer him/herself, it was not feasible within the timeframe of the current revision. We are planning to include it as a different project in the next future, which will include the complete characterisation of the Pfn1 mouse model. Nevertheless, the authors should include WT controls to their analysis in supplementary figure 9, and indicate the level of gain/loss of each chromosome.
To address the Reviewer's concern, we have updated the Supplementary Figure 9 (now Supplementary Figure 11) to show the magnitude of copy number gains and losses of each chromosome relative to a wild type sample. However, a control sample cannot be added in this figure because the copy number plot shows normalised data of KO clones compared with WT control, which was therefore subtracted as baseline.
6. Line 248: the authors refer to high level copy number amplifications but do not provide further details explaining what genes were amplified, and the level of amplification.
We have not provided details about copy number amplifications in the OS/PDB samples because CNV frequently did not contain genes, but rather affected intergenic regions. In the case where genes were involved, we did not notice gene recurrency 8. In Figure 5 KO should be HET in the different panels as these are Pfn1+/derived.
We agree with the Reviewer and replaced "KO" with "Pfn1 +/-" in panels i, j, and k of Figure 5. Figure 5 there is no indication for % of micronuclei, although the text suggests there is.
10. The relevance of the OD/PDB patient sequencing data is not immediately clear. There is no clear comparison between samples that have PFN1 LOH and ones that do not. Could this analysis be extended using already published sequences from human cancers? It would be helpful if the authors could provide more context and explain what can be learned from this sequencing efforts with regard to PFN1 LOH. Also, the ability to determine WGD relative to PFN1 LOH is interesting, and I would encourage making this analysis more accessible to the non-computational reader (for example, "WGD relative timing" is not a clear parameter). We thank the Reviewer for pointing out this issue. We have now specified which samples have loss of heterozygosity of PFN1 in Supplementary Table 2 (please, see column H "PFN1 LOH"). We have also modified the main text to explain the intuition underpinning the timing analysis of WGD and make that analysis more accessible to non-expert readers. We decided to focus on osteosarcoma genomes to assess the relevance of PFN1 inactivation in human cancers due to our previous work reporting loss-of-function mutations in early onset Paget's disease of bone. In these tumours, we observe loss of heterozygosity at the PFN1 locus and multiple copies of PFN1. We believe that such configuration is more likely to occur through loss of one copy of PFN1, then followed by WGD (as we observe in vitro) and therefore duplication of the