An RFC4/Notch1 signaling feedback loop promotes NSCLC metastasis and stemness

Notch signaling represents a key mechanism mediating cancer metastasis and stemness. To understand how Notch signaling is overactivated to couple tumor metastasis and self-renewal in NSCLC cells, we performed the current study and showed that RFC4, a DNA replication factor amplified in more than 40% of NSCLC tissues, directly binds to the Notch1 intracellular domain (NICD1) to competitively abrogate CDK8/FBXW7-mediated degradation of NICD1. Moreover, RFC4 is a functional transcriptional target gene of Notch1 signaling, forming a positive feedback loop between high RFC4 and NICD1 levels and sustained overactivation of Notch signaling, which not only leads to NSCLC tumorigenicity and metastasis but also confers NSCLC cell resistance to treatment with the clinically tested drug DAPT against NICD1 synthesis. Furthermore, together with our study, analysis of two public datasets involving more than 1500 NSCLC patients showed that RFC4 gene amplification, and high RFC4 and NICD1 levels were tightly correlated with NSCLC metastasis, progression and poor patient prognosis. Therefore, our study characterizes the pivotal roles of the positive feedback loop between RFC4 and NICD1 in coupling NSCLC metastasis and stemness properties and suggests its therapeutic and diagnostic/prognostic potential for NSCLC therapy.

The authors provide evidence to support a novel role for RFC4 in blocking CDK8-mediated degradation of NICD, and that this results in enhanced invasiveness/stem-ness in NSCLC cells in vitro/vivo. The mechanistic link between RFC4 over-expression (common in NSCLC)and increased NICD-dependent gene expression is strongly suggested to be a result of direct binding of RFC4 to NICD. A significant piece of data absent from the manuscript is the impact of genetic manipulation e.g RFC4 over-expression on cell growth rates/cell cycle distribution (see comments below). The experimental data on the impact of high RFC4 expression have strong correlates with prognosis in NSCLC clinical datasets. A limitation of the current data is that expression of RFC4 strongly correlates with expression of other DNA replication and proliferation associated genes (reviewer's own analysis of TCGA datasets used in this study), and therefore a causative effect of RFC4 over-expression on clinical outcomes cannot be definitively made, despite the strong correlations. Nonetheless, this is the first report that this reviewer is aware of of this unexpected role of RFC4, and it potentially has impact across other disease sites where RFC4 is overexpressed. The manuscript would benefit from careful scientific proofreading There are a number of specific comments below. The authors state that the expression levels of RFC4 positively correlated with levels of several canonical downstream genes of Notch signalling (Fig 4j). However, the authors do not report which of the ~20,000 genes most correlate with RFC4 expression. In particular, are genes associated with DNA synthesis, cell cycle etc better correlated with RFC4 than canonical Notch signalling Results Page 10 It's not clear whether increased expression of RFC4 is associated with increased expression of CDK8 in clinical samples. If they are co-upregulated, then the hypothesis that increased RFC4 expression leads to increased NICD1 levels becomes weaker Results Page 11 RFC4 is an essential gene for growth/survival of many cell lines (https://depmap.org/) and therefore proliferation within the sh-RFC4 xenograft (Fig 6d) may have selected for cells with only marginally reduced RFC4 levels which would affect how these data are interpreted. The manuscript titled "A RFC4/Notch1 signaling feedback loop promotes NSCLC metastasis and stemness" by Liu et al., demonstrated a RFC4 mediated regulation of Notch1 activity which is a novel finding has a high clinical significance. This study identified RFC4 as a novel regulator of Notch1 and showed a pathological role associated with NSCLC metastasis and stemness. In silico analysis of data obtained from tumor samples has identified an upregulation of RFC4 and NICD1 in metastatic tumors that are significantly correlated with patient overall survival. Overall, this work is novel and highly interesting however addressing the below mentioned concerns definitely will increase the quality of the work.
Major weaknesses of this study include: 1. Using only a single lung tumor cell line as well as 293 cells to demonstrate the majority of these effects. Experiments should include a panel of lines with different expression of these proteins. 2. Using transfected highly overexpressed NICD to study the role of RFC4 mediated Notch1 regulation and associated pathological functions. It would be difficult to envision a role for RFC4 in Notch1 regulation in significant patient population without a demonstration under physiological conditions in multiple cancer cell lines. 3. In this study, using patient data, the authors identified a strong correlation between RFC4 expression to metastasis which needs to be validated in a panel of cell lines with low vs high endogenous RFC4 expression and xenograft experiments. Cell lines with high RFC4 expression should correlate in vivo metastasis and Notch1 regulation compared to low RFC4 expression in a physiological states. Alternatively, a demonstration of endogenous RFC4 knockdown causes a decrease in metastasis and stemness in multiple high-expressing NSCLC cell lines would be important. 4. 3q amplification is mostly seen in squamous cancers where Notch is thought to be a tumor suppressor. This should be addressed in the discussion.
Minor concerns: Methods are not adequately reported to really understand some of the key findings presented in this paper.
The statement about the 1% 5 year survival of patients with metastatic disease on page 3 is likely outdated in the era of immunotherapy in which we are seeing 15 -20% 5-year survivals. Figure.1b: How was NICD1 levels were determined in IHC staining? Scoring of IHC was not clear. Figure.1b: NICD1 low and NICD1 high levels where these total Notch1 levels or NICD1 levels. How was NICD1 levels were determined?   The grammar needs editing.
The figures have too many panels.
Reviewer #4 (Remarks to the Author): Aberrant Notch signaling has long been established to play a key role in tumor formation. Notch is normally activated and cleaved at the membrane to release the intra-cellular domain (ICD) transcription factor. The most prominent oncogenic downstream target of Notch1 is probably c-Myc, a driver of nearly every tumor. But also critical inhibitors of differentiation are activated by Notch. Hyper-activation of Notch1 usually occurs via point mutations that either constitutively activate or stabilize the protein. The ICD is normally degraded by the Fbxw7 ubiquitin ligase via a degron in the C-terminus of Notch. In cancers, this regulatory mechanism is often eliminated by frame-shift mutations that truncate the protein prematurely, especially in T-cell leukemia. Therefore, inhibition of Notch signaling has received great attention in patient treatment. However, while gamma-secretase inhibitors work well in cell culture, they are intolerable in clinical settings, and alternative approaches are needed.
RFC4 is one of several subunits of the replication factor C complex and functions in DNA replication and repair as a polymerase assessory protein. It has been found over-expressed in many cancer studies and is associated with tumor progression and poor prognosis.
In this manuscript, the authors identify RFC4 as Notch1 activated gene that cooperates with the Notch ICD in tumor metastasis. The RFC4 gene is amplified in a large percentage of non-small cell lung cancers and directly activated by Notch1. Mechanistically, RFC4 is shown here to directly bind and stabilize the Notch1 ICD by preventing its phosphorylation, presumably at the degron site, and thereby evade recognition by the Fbxw7 ubiquitin ligase. This is an impressive study, and the data presented are very interesting and important to the field. Indeed, a quick look at available cancer data in c-Bioportal for NSCLC reveals near mutually exclusive RFC4 amplification, activating Notch1 mutations, and inactivating Fbw7 mutations, suggesting they function complementary on the same pathway. But while some data nicely establish the authors's model, other data do not sufficiently support their claims and need additional work, especially Figures 4 & 5. Moreover, it is difficult to reconcile how a clamp loader (RFC4) for a DNA polymerase would have such stabilizing effect on the NICD. While not impossible, this discrepancy should be discussed better.
Other comments: 1. Several stemness genes were analyzed for RFC4 responsiveness upon NICD expression (Fig.  1J). Surprisingly, c-Myc is not presented here (nor L-Myc). Why not? How about other genes not implicated in stemness? I would also like to see a control in the set that is not inhibited by RFC4 knockdown to ensure that such knockdown doesn't interfere with (NICD-dependent) transcription in general. As this figure appears, virtually every gene they looked at is regulated by RFC4. Fig. 4a it is shown that RFC4 specifically induces the NICD, but not full-length Notch1, suggesting a post-translational mechanism like protein stabilization of the NICD. The full-length band of Notch is rather poorly shown. How do the authors know this band is indeed Notch1? Please, provide a knock-down to establish this is the correct band. CHX can interfere with the assay itself, and it should first be established that it doesn't by using a standard pulse chase assay. Fig. 5b and 5c are completely uncontrolled. Even though IgG lanes are widely used as controls, I much prefer bait knock-downs, and the authors should use knock-downs instead in 5b and leave out the bait in 5c to demonstrate specificity of their interactions. They could also coexpress both proteins in cells and do the IPs without using recombinant protein, along with the according controls.

The IPs in
6. Fig 5f is insufficient to make the point of competitive binding between CDK8 and RFC4 to NICD. No cyclin was used in this assay, which is known to greatly alter the conformation of the kinase. And how about the other way around? Do increasing amounts of CDK8 displace RFC4? Best is to examine the NICD -RFC4 interaction as in 5b and knock-down CDK8, or examine the NICD -CDK8 interaction with RFC4 knock-down. 7. The NICD is full of potential serine phosphorylation sites. By using a general p-Ser but otherwise not site-specific antibody, the authors have no idea what they are looking at. Sure, the site seems CDK8 responsive (Fig 5h & i). But the relevant site in the Notch degron is actually a threonine (T2512). The contributing effects of S2514 and S2517 are controversial at best. Without knowing the site in question, there is no evidence for a mechanism.
8. The point of Fig 5k is to demonstrate that enforced RFC4 expression displaces CDK8 from Notch complexes, which then would abolish degron phosphorylation and therefore diminish Fbxw7 binding. We have extensive experience with Fbxw7 in my lab. I therefore find these data unbelievable. For one, Fbxw7 protein is notoriously hard to detect by itself, let along in co-IPs. But more importantly: without uncoupling of binding from turnover, a ubiquitin ligase, like Fbw7, cannot be co-immunoprecipitated by a substrate as it would become degraded upon contact. Now, there are some exceptions. But Notch isn't one of them. In addition, it is our experience that the antibody used in this study (3A9) makes an Fbxw7-independent cross-reaction right underneath where untagged Fbw7 would migrate on a gel. Therefore, this experiment needs to be better controlled using Fbxw7 knock-downs, or better crispr-mediated knockout. There are also cell lines available with targeted deletion of Fbw7. Overall, this is good work. The functional link of NICD and RFC4 is well illuminated. But the mechanism is still rather unclear.

Reviewer #2 (Remarks to the Author):
The authors provide evidence to support a novel role for RFC4 in blocking CDK8-mediated degradation of NICD, and that this results in enhanced invasiveness/stem-ness in NSCLC cells in vitro/vivo. The mechanistic link between RFC4 over-expression (common in NSCLC) and increased NICD-dependent gene expression is strongly suggested to be a result of direct binding of RFC4 to NICD. A significant piece of data absent from the manuscript is the impact of genetic manipulation e.g RFC4 over-expression on cell growth rates/cell cycle distribution (see comments below). The experimental data on the impact of high RFC4 expression have strong correlates with prognosis in NSCLC clinical datasets. A limitation of the current data is that expression of RFC4 strongly correlates with expression of other DNA replication and proliferation associated genes (reviewer's own analysis of TCGA datasets used in this study), and therefore a causative effect of RFC4 overexpression on clinical outcomes cannot be definitively made, despite the strong correlations. Nonetheless, this is the first report that this reviewer is aware of this unexpected role of RFC4, and it potentially has impact across other disease sites where RFC4 is over-expressed. The manuscript would benefit from careful scientific proofreading There are a number of specific comments below. Response: We thank the reviewer for raising these issues. The Gene Set Enrichment Analysis (GSEA) was employed to analyze whether the priori defined set of genes (Notch signaling up-regulated genes) shows statistically significant, concordant differences between different lymph node metastasis statuses of patients in TCGA NSCLC cohort.
These NSCLC samples were divided into two subgroups, namely, with lymph node metastasis (LN metastasis) subgroup or without lymph node metastasis (non-LNmetastasis) subgroup, which were derived from 302 and 407 patients, respectively, according to their lymph node metastasis status. In the GSEA analysis, the genes were ranked according to the fold-changes of their expression between the two subgroups from high to low (the graded scale from left-to-right). The red scale represents genes highly expressed in the LN metastasis subgroup and the blue scale represents genes highly expressed in the non-LN metastasis subgroup. Each black straight line represents a gene in priori defined gene set and the position of line represents its ranking in the ordered gene list. In addition, an enrichment score (ES) of each gene in the ordered gene list was calculated to create a running ES score. The running ES score increases when the current gene is present in the Notch signaling up gene set and highly-expressed, otherwise it deceases. The final ES score was the maximum value of the running ES score (i.e., 0.49 in Figure 1a). Therefore, the data analyzed in Figure 1a showed that the majority of Notch signaling up-regulated genes were enriched and upregulated in the LN metastasis subgroup as compared to the non-LN-metastasis subgroup, suggesting that Notch signaling is activated in NSCLC samples with lymph node metastasis. In this revision, we have reanalyzed the updated RNA-seq profiles of the TCGA NSCLC cohort using the GSEA assay and found significant correlation between Notch signaling activation and LN metastasis.
Corresponding data have been incorporated into the revised manuscript on Page 6. Moreover, silencing RFC4 significantly compromised the invasive and self-renewal abilities of NICD1-overexpressing NSCLC cells even when cell proliferation was inhibited by mitomycin C (a DNA synthesis inhibitor) treatment or cell death was induced by cisplatin treatment, respectively, evaluated in the cell invasion assay and tumor sphere culture assay (new Supplementary Figure 1, h and i). These data suggest that RFC4 plays crucial roles in promoting overactivated Notch1-driven tumor metastasis and stemness, which can be independent of the promoting effects of NICD1 on tumor cell proliferation and survival. Also notably, we also proved that the potent pro-metastatic and pro-selfrenewal effects of RFC4 can be independent of its role in DNA replication and repair and its ability to enhance cell proliferation/cell cycle/survival as described below (new Supplementary Figures 4 and 5). Moreover, primary NSCLC tumors exhibiting LN metastasis expressed much higher levels of NICD1 or RFC4 than non-metastatic tumors (new Figure 1b and Figure 7b). Therefore, our study focuses on the important regulatory roles of a RFC4/Notch1 signaling feedback loop in NSCLC metastasis and stemness.
Corresponding data have been incorporated into the revised manuscript on Pages 6, 10 and 15.

Fig 1i
What is scale showing? Where is the list of significantly altered genes?

Response:
The graded red/blue scale for the heatmap in Figure 1i shows the range of log10 Response: The authors thank the reviewer for raising these important questions. As measured by western blotting and calculated by gray scan of immunoblots, shRNA against RFC4 in RFC4-silenced A549-NICD1 cells efficiently decreased protein levels of RFC4 to less than 30%, of those in A549-NICD1 and A549-Vector cells (new Supplementary Figure   1, a and j). As expected, silencing RFC4, which is a DNA replication factor, not only markedly reversed the effects of NICD1 overexpression on promoting invasion/migration and cancer stemness, but also reversed NICD1-enhanced tumor cell growth rate, proliferation and survival (new Supplementary Figure 1 To address the reviewer's concern, we had evaluated the biological importance of the top 10 NICD1 up-regulated genes, namely, RFC4, BOLA2B, PROK1, SLUG, KCNA5, HES5, TFF1, ANKRD1, SPRY4 and CALCB in NICD1-induced malignancy. As shown in Supplementary Authors use the term 'silenced' for RF4C expression, but do not indicate what level of knockdown (RF4C protein) that this corresponds to.

Response:
To address the reviewer's concern, western blotting analysis showed that protein levels of RFC4 in A549-luci-NICD1 cells silenced for RFC4 expression were decreased to about 17% and 43%, respectively, of those in A549-luci-NICD1-Scramble and A549-luci-Scramble. Similarly, RFC4 protein levels in subcutaneous tumors formed by RFC4silenced A549-luci-NICD1 cells were decreased to 16% and 47%, respectively, of those in subcutaneous tumors formed by A549-luci-NICD1-Scramble and A549-luci-Scramble cells

Results Page 9
The authors state that the expression levels of RFC4 positively correlated with levels of several canonical downstream genes of Notch signaling (Fig 4j). However, the authors do not report which of the ~20,000 genes most correlate with RFC4 expression. In particular, are genes associated with DNA synthesis, cell cycle etc better correlated with RFC4 than canonical Notch signaling. those with genes associated with DNA synthesis and cell cycle, such as CCNB2 (r = 0.73; P < 0.001), PCNA (r = 0.68; P < 0.001), CDK2 (r = 0.69; P < 0.001), POLR2H (r = 0.70; P < 0.001) and MKI67 (r = 0.41; P < 0.001). Notably, the number of metastatic NSCLC samples is quite small relative to the number of non-metastatic NSCLC samples, which might largely affect the analysis that whether genes associated with DNA synthesis and cell cycle are better correlated with RFC4 than canonical Notch signaling. Due to too many data in this revision, we prefer to not present the correlation data in the revised manuscript, and we will present them if the reviewer feels necessary.

Results Page 10
It's not clear whether increased expression of RFC4 is associated with increased expression of CDK8 in clinical samples. If they are co-upregulated, then the hypothesis that increased RFC4 expression leads to increased NICD1 levels becomes weaker. samples. Importantly, the SPR kinetic analysis showed that the binding affinity between RFC4 and NICD1 was approximately five-folds higher than that between CDK8 and NICD1 (new Figure 5c). Moreover, the amounts of CDK8 or its binding partner Cyclin C, were gradually impaired and even totally diminished in the pulled down proteins when HAtagged NICD1 proteins were immunoprecipitated following addition of purified RFC4 in a dose-dependent manner; interestingly, although the binding of NICD1 to RFC4 could be impaired by the increasing amounts of purified CDK8 in a dose-dependent manner, it appears that large amounts of CDK8 proteins are required to significantly abrogate the interaction between NICD1 and RFC4 (new Figure 5d). Furthermore, while silencing RFC4 mildly enhanced the interaction between NICD1 and CDK8, silencing CDK8 significantly enhanced the interaction between NICD1 and RFC4; in parallel, overexpressing RFC4 significantly abrogated the binding of NICD1 to CDK8 whereas CDK8 overexpression mildly impaired the binding of NICD1 to RFC4 (new Figure 5e).
These data strongly suggest that both higher binding affinity of RFC4 to NICD1 and increased expression levels of RFC4 in NSCLC tissue make RFC4 more competitive than CDK8 to bind to NICD1. Therefore, in NSCLC increased expression of RFC4 able to directly bind NICD1 should competitively abrogate CDK8/FBXW7-mediated degradation of NICD1 and thus lead to increased NICD1 levels. Corresponding data have been incorporated into the revised manuscript on 12 and 13.

Results Page 11
RFC4 is an essential gene for growth/survival of many cell lines (https://depmap.org/) and therefore proliferation within the sh-RFC4 xenograft (Fig 6d) may have selected for cells with only marginally reduced RFC4 levels which would affect how these data are interpreted.

Response:
We thank the reviewer for this interesting comment. To address the reviewer's concern, in this revision, we evaluated the levels of RFC4 and NICD1 expression and cell proliferation/apoptosis in subcutaneous tumors formed by A549-Vector-Scramble, A549-NICD1mut-Scamble, or A549-NICD1mut-sh-RFC4 cells. As shown in new Figure 6d 17. Figure 7b It is not standard to display data for just top and bottom 25% of cases. Data for all quartiles should be shown. Appropriate description has been incorporated into the figure legends in the revised manuscript.

Reviewer #3 (Remarks to the Author):
The manuscript titled "A RFC4/Notch1 signaling feedback loop promotes NSCLC metastasis and stemness" by Liu et al., demonstrated a RFC4 mediated regulation of Notch1 activity which is a novel finding has a high clinical significance. This study identified RFC4 as a novel regulator of Notch1 and showed a pathological role associated with NSCLC metastasis and stemness. In silico analysis of data obtained from tumor samples has identified an upregulation of RFC4 and NICD1 in metastatic tumors that are significantly correlated with patient overall survival. Overall, this work is novel and highly interesting however addressing the below mentioned concerns definitely will increase the quality of the work.  Supplementary Figures 6 and 7). These data strongly support our conclusion that RFC4 is a new target gene of the Notch signaling and that high levels of RFC4 stabilize NICD1 to form a positive feedback and are enough to cause overactivation of the Notch1 signaling, probably without requiring γ-secretasedependent cleavage of Notch1 into NICD1, in NSCLC progression. Corresponding data have been incorporated into the revised manuscript. Of note, due to too many panels in figures, we have removed the majority of data derived from 293FT cells in this revision and we would like to present them if the reviewer feels necessary.
2. Using transfected highly overexpressed NICD to study the role of RFC4 mediated in these metastatic lung cancer cells significantly decreased NICD1 protein levels in the nucleus without affecting Notch1 mRNA levels or the quantities of full-length Notch1 proteins or cytoplasmic NICD1 proteins, increased K48-linked polyubiquitination of NICD1, caused shorted half-lives of NICD1 proteins and thus inhibited transcriptional activity of the Notch signaling. Moreover, RFC4 could tightly bind with NICD1 and abrogated the interaction between CDK8 or FBXW7 with NICD1; therefore, silencing RFC4 significantly increased CDK8-dependent serine and threonine phosphorylation of nuclear NICD1 proteins and failed to cause K48-linked polyubiquitination of NICD1 when FBXW7 or CDK8 was pre-silenced or depleted in H1975, LC1 and LLC cells (new Figure   5, a-k; Supplementary Figure 7, a-f). Notably, per Reviewer 4's request, we also prove that both higher binding affinity of RFC4 to NICD1 and increased expression levels of RFC4 in NSCLC tissue make RFC4 more competitive than CDK8 to bind to NICD1 using various experimental models (new Figure 5 Moreover, we used immunocompetent mice (C57BL/6) to establish experimental metastasis and tumorigenicity models of RFC4-silenced and vector-control LLC cells.
Our results showed that silencing RFC4 significantly suppressed the ability of highly metastatic murine lung cancer cells to form lung metastases when injected intravenously or to develop subcutaneous tumors when injected with various cell numbers ranging from

Response:
We are sorry for the careless error and thank the reviewer for pointing out it.
Corrections have been made in the revised manuscript on Page 3. 7. Figure.1b: How was NICD1 levels were determined in IHC staining? Scoring of IHC was not clear. Figure.1b: NICD1 low and NICD1 high levels where these total Notch1 levels or NICD1 levels. How was NICD1 levels were determined?

Response:
We thank the reviewer for raising these important questions. In our IHC staining assays, the antibody (ab8925, abcam) only detects the activated Notch1 fragment (aa 1755-1767) as the corresponding epitope, which is exposed after γ-secretase cleavage of Notch1 into NICD1, and thus only detects NICD1 but is unable to recognize the un-cleaved Corresponding information has been provided in the revised manuscript.
8. Figure   Other comments: 1. Several stemness genes were analyzed for RFC4 responsiveness upon NICD expression (Fig. 1J). Surprisingly, c-Myc is not presented here (nor L-Myc). Why not? How about other genes not implicated in stemness? I would also like to see a control in the set that is not inhibited by RFC4 knockdown to ensure that such knockdown doesn't interfere with (NICD-dependent) transcription in general. As this figure appears, virtually every gene they looked at is regulated by RFC4. 2. In Fig. 4a it is shown that RFC4 specifically induces the NICD, but not full-length Notch1, suggesting a post-translational mechanism like protein stabilization of the NICD.

Response
The full-length band of Notch is rather poorly shown. How do the authors know this band is indeed Notch1? Please, provide a knock-down to establish this is the correct band. They could also co-express both proteins in cells and do the IPs without using recombinant protein, along with the according controls.

Response:
The reviewer's suggestion is well taken. In this revision, knock-down of RFC4 or Notch1 was used to confirm the specific interaction between RFC4 and NICD1.
As shown in new Figure 5b, the IP assays immunoprecipitated with anti-NICD1 or anti-RFC4 antibodies revealed that the interaction between RFC4 and NICD1 was diminished when RFC4 and Notch1 was separately silenced. In parallel, when HA-tagged NICD1 was co-expressed with Flag-tagged RFC4, Flag-tagged GAPDH or the empty vector, the IP assays immunoprecipitated with anti-HA affinity agarose showed that only Flag-tagged RFC4 was pulled down; similarly, when Flag-tagged RFC4 was co-expressed with HAtagged NICD1, HA-tagged p84 or the empty vector, the IP assays immunoprecipitated with anti-Flag affinity agarose showed that only HA-tagged NICD1 was pulled down (new Supplementary Figure 7c). These data further demonstrate specificity of the interaction between RFC4 and NICD1. Corresponding data have been incorporated into the revised manuscript on Page 12.
6. Fig 5f is insufficient to make the point of competitive binding between CDK8 and RFC4 to NICD. No cyclin was used in this assay, which is known to greatly alter the conformation of the kinase. And how about the other way around? Do increasing amounts of CDK8 displace RFC4? Best is to examine the NICD -RFC4 interaction as in 5b and knock-down CDK8, or examine the NICD -CDK8 interaction with RFC4 knock-down.

Response:
The authors thank the reviewer for raising the important advice. The surface plasmon resonance (SPR) kinetic analysis showed that the binding affinity between RFC4 and NICD1 was approximately five-folds higher than that between CDK8 and NICD1 (new Figure 5c). In consistence, similar to CDK8, the amounts of Cyclin C, which is a binding partner for CDK8, were also gradually decreased and even totally vanished in the pulled down proteins when HA-tagged NICD1 proteins were immunoprecipitated following addition of purified RFC4 in a dose-dependent manner (new Figure 5d). Interestingly, although the binding of NICD1 to RFC4 could be impaired by the increasing amounts of purified CDK8 in a dose-dependent manner, it appears that large amounts of CDK8 proteins are required to significantly abrogate the interaction between NICD1 and RFC4 (new Figure 5d). Moreover, while silencing RFC4 mildly enhanced the interaction between NICD1 and CDK8, silencing CDK8 significantly enhanced the interaction between NICD1 and RFC4 (new Figure 5e). In parallel, overexpressing RFC4 significantly abrogated the binding of NICD1 to CDK8 whereas CDK8 overexpression moderately impaired the binding of NICD1 to RFC4 (new Figure 5e). It is also important to note that from the TCGA lung cancer datasets the mRNA levels of both CDK8 and  Figure 5a). These data suggest that both higher binding affinity of RFC4 to NICD1 and increased expression levels of RFC4 in NSCLC tissue make RFC4 more competitive than CDK8 to bind to NICD1, leading to abrogated phosphorylation and ubiquitination of NICD1 and thus stabilization of NICD1 proteins. Corresponding data have been incorporated into the revised manuscript on Pages 12 and 13.
7. The NICD is full of potential serine phosphorylation sites. By using a general p-Ser but otherwise not site-specific antibody, the authors have no idea what they are looking at.
Sure, the site seems CDK8 responsive (Fig 5h & i). But the relevant site in the Notch degron is actually a threonine (T2512). The contributing effects of S2514 and S2517 are controversial at best. Without knowing the site in question, there is no evidence for a mechanism.
Response: The authors thank the reviewer for raising these important comments. Since Based on these data, we guess that the binding of RFC4 to the PEST domain of NICD1  Figure 5j presented as original Figure 5l in the initial version of manuscript), the IP assays immunoprecipitated with anti-HA affinity agarose showed that the binding of HA-tagged K48-linked ubiquitin (Ubk48-HA) to Flag-tagged NICD1 (NICD1-Flag) was greatly impaired when CDK8 or FBXW7 was silenced, whereas silencing RFC4 failed to reverse the binding of HA-tagged K48-linked ubiquitin to Flag-tagged NICD1 when CDK8 or FBXW7 was pre-silenced (new Supplementary Figure 7i), further supporting the conclusion that RFC4 abrogates CDK8/FBXW7-mediated polyubiquitination of NICD1.
Corresponding data have been incorporated into the revised manuscript on Page 14.
Overall, this is good work. The functional link of NICD and RFC4 is well illuminated. But the mechanism is still rather unclear.

Response:
The authors greatly appreciate the reviewer's inspiring comments and very constructive suggestion.
The authors have adequately addressed the main issues raised during the first review process.
The mechanistic data presented regarding the interaction of RFC4 and notch signalling provide strong evidence for a novel link between RFC4 expression and Notch signalling that will be of interest in the field -but the relationship between RCF4 expression and outcomes in NSCLC remain correlative and therefore need to be interpreted cautiously, in particular as RCF4 expression correlates with cell proliferation markers.

Response:
The authors greatly appreciate the reviewer's encouraging and constructive comments. Our data strongly suggest that through activating Notch signaling and forming a positive feedback loop with NICD1, RFC4 potently promotes tumor invasion/metastasis and stemness/tumorigenicity, which can be independent of its role in DNA replication and repair and of its ability to enhance cell proliferation/cell cycle/survival. Indeed, as RCF4 expression significantly correlates with cell proliferation markers in human NSCLC tissue, the oncogenic effects of RFC4 on cell proliferation, tumorigenicity and tumor metastasis could together contribute to the worse outcomes of NSCLC patient prognosis. To ease the reviewer's concern, in this revision we have modified the interpretation of the roles of RFC4 in NSCLC appropriately and appropriate description has also been incorporated into the Discussion Section in the revised manuscript on Page 18.
Specific points line 72. The response rates of patients with metastatic NSCLC are not 'extremely low' Provide more specific information

Response:
We thank the reviewer for pointing out the inappropriate description. The response rates of patients with metastatic NSCLC to first-line chemotherapy, targeted therapy and immunotherapy can be ranging from 15%-30%, 25%-60%, and 14%-65%, respectively 9, 10, 11 . As response rates might be impacted by selected therapy types, prescribed medication, enrolled patients and so on, it is hard to give a definitive calculation of the response rates of metastatic NSCLC patients to clinical therapies. Thus, we prefer to change the original description to "Despite the significant advancement in currently available therapies, the unsatisfying response rates of metastatic NSCLC patients to the initial anti-cancer treatments and the fairly high frequencies of tumor recurrence posttreatment remain to be the most serious challenge in the clinic." lines 103/104 The text 'which almost resulted in therapeutic failures' does not make sense Response: In this revision, "which almost resulted in therapeutic failures" has been changed into "which almost ended in therapeutic failures".
lines 168-175 The terms 'High' and 'Low' metastatic lines (for A549 and H1975) are not defined Response: According to a recent paper, A549 cell line is moderate metastatic, while H1975 cell line presents highly metastatic potential (Nature. 2020 Dec;588(7837):331-336.). To avoid misunderstanding, in this revision, we have deleted these terms and made appropriate modification according to the context of this paragraph on page 7.
Comments: In NSCLC, the RFC4 gene is commonly amplified in squamous cell but not adenocarcinoma (TCGA), but this is not discussed, and the human cell lines used in the study are adenocarcinoma.
The specificity of these observations to NSCLC are nor clear as (e.g.) esophageal and ovarian cancers also have a high proportion of RFC4 gene amplification, and mRNA expression is highly correlated with markers of cell proliferation (TCGA) for these diseases too -these limitations should be addressed in the discussion as the data could (potentially wrongly) indicate selective effects for NSCLC, in particular as the most common NSCLC histological subtype (adenocarcinoma) rarely has RCF4 gene amplification.

Response:
We thank the reviewer for this important suggestion. In fact, we have employed a LUSC cell line (H1703) to prove the pro-invasive and pro-tumorigenic effects of RFC4 and RFC4-directed stabilization of NCD1 proteins and thus activation of Notch signaling ( Fig. 3a, 6f, 6g, S1 a-d, S3 e-h, S4 a-g and S8 a-c, e and f), which are consistent with the results obtained from multiple LUAD cell lines. properties. In addition, other cancer types, such as esophageal and ovarian cancers, in which aberrantly activated Notch1 signaling plays important roles, also have distinct high proportions of RFC4 gene amplification, and high RFC4 levels significantly correlates with poor prognosis of patients with these cancers, indicating that the oncogenic effects of RFC4 are not limited to NSCLC. Appropriate information has been incorporated into the Discussion Section in the revised manuscript on Pages 19 and 20.

Reviewer #3 (Remarks to the Author):
The authors have carefully addressed and extensively revised the manuscript in response to the reviewers' comments and added a substantial amount of new data in support of their findings.
Two minor comments: LLC is not really a lung cancer cell line, but a sarcoma, and the We prefer to present the data obtained from LLC in this revised manuscript and are also willing to remove them if the reviewer/editor feels necessary. In addition, the misspelled "FALG" in the Figure 5 legend has been corrected to "FLAG".

Reviewer #4 (Remarks to the Author):
The authors have addressed all my points of concern adequately. My last comment concerns the migration of Fbxw7 on SDS-PAGE. In figure 5 as well as supplemental figure   7 the migration of Fbxw7 is marked at just under 70 kDa and in my reviewers figure 4 Fbxw7 migrates just above 40 kDa. The authors should revisit their migration pattern as the most common Fbw7 alpha isoform is established to migrate at around 110 kDa (~30 kDa larger than calculated). There are indeed 2 smaller isoforms, which would migrate around the marked 70 kDa range (Fbxw7 beta is about 69 kDa). But to my knowledge this isoform is much harder to detect than the alpha isoform, and it is cytoplasmic.