Hexokinase 2-driven glycolysis in pericytes activates their contractility leading to tumor blood vessel abnormalities

Defective pericyte-endothelial cell interaction in tumors leads to a chaotic, poorly organized and dysfunctional vasculature. However, the underlying mechanism behind this is poorly studied. Herein, we develop a method that combines magnetic beads and flow cytometry cell sorting to isolate pericytes from tumors and normal adjacent tissues from patients with non-small cell lung cancer (NSCLC) and hepatocellular carcinoma (HCC). Pericytes from tumors show defective blood vessel supporting functions when comparing to those obtained from normal tissues. Mechanistically, combined proteomics and metabolic flux analysis reveals elevated hexokinase 2(HK2)-driven glycolysis in tumor pericytes, which up-regulates their ROCK2-MLC2 mediated contractility leading to impaired blood vessel supporting function. Clinically, high percentage of HK2 positive pericytes in blood vessels correlates with poor patient overall survival in NSCLC and HCC. Administration of a HK2 inhibitor induces pericyte-MLC2 driven tumor vasculature remodeling leading to enhanced drug delivery and efficacy against tumor growth. Overall, these data suggest that glycolysis in tumor pericytes regulates their blood vessel supporting role.

1. How do authors explain that high HK2 expression reduces vessel dimeter? They claim that that HK2 regulates cell contractility via ROCK expression, but this is hardly convincing from the data provided.
2. CD146 staining seems very weak, especially in NSLCC, at least in the pictures showed in Figure  1B. Only the depicted big vessel in the inset seems to be positive for CD146 (yellow), the other two are only red (αSMA positive) and the majority of the small vessels are only CD34 positive. Given that the MACS enrichment strategy is based on CD146 expression in vascular cells, this picture generates doubts about the percentage of pericytes obtained after FACS purification ( Figure 1F and H). Also, this percentage is almost double in tumor samples compared with normal tissue. More images and a quantification of CD146 expression in tissue sections of normal and tumoral section will help to clarify this issue.
3. It seems that pericytes in vitro behave quite similarly irrespective or the origin. Are isolated pericytes losing their properties/dedifferentiating when cultured? However, when combined with endothelial cells, TPC seem to enhance tube/EC contraction compared to NPC. Again, the data are not convincing. How do authors explain that TPC express higher HK2 levels?
4. Figure 4J. Authors claim that HK2 expression is significantly increased in mural cells in tumors. Quantification of that expression in representative images should be provided. Endothelial cells seem to be contributing to HK2 overexpression too, as well as many other tumoral/stromal cells. This observation compromises the pericyte-driven hypothesis of the effects of in vivo, pharmacological inhibition of HK2. Pre-treated or genetically modified pericytes should be implanted together with tumor cells or, ideally, tumor cells should be implanted in pericyte-specific HK2 deficient mice. 5. Conclusions obtained from clinical correlation data ( Figure 4K, Suppl Fig 4E) on HK2 expression are not valid, markers used are not specific of pericytes. PDGFRβ and αSMA are expressed by cancer fibroblasts and CD146 is also expressed in endothelial cells. Figure 2B does not match the qPCR profiles, as fibroblasts in the IF image show much greater expression than qPCR graph depicts compared with pericytes Reviewer #2 (Remarks to the Author); expert on metabolism:

Data on PDGFRβ expression in cultured cells in
Pericytes play an important role in angiogenesis and tumor microenvironment formation. The authors developed a novel method to isolate pericytes from human normal and tumor tissues. This is a robust model for studying the role of pericytes in angiogenesis and cancer progression. The authors discovered that pericytes from tumor had higher glycolytic flux but lower mitochondrial activity, suggesting that the TPC also display the Warburg effect, which alters cell contractility and limit blood supply. This could further exacerbate tumor hypoxia.
A few additional experiments that could strengthen the conclusions: In 125-128, how did the authors conclude that 'CD146 was expressed in both endothelial cells and pericytes …, while CD34 and PDGFRβwas specifically expressed in endothelial cells and pericytes respectively'? Were independent markers used to confirm those cells are endothelial cells and pericytes respectively? How can they tell that CD146 was expressed in pericyte without showing the co-staining of CD146 and PDGFRβ in Figure1?
The data demonstrated a strong association between HK2/glycolysis upregulation and ROCK2 activation. However, it is not clear whether the paracrine factors from tumor cells reprogram metabolism to activate ROCK2, or activate ROCK2 to reprogram metabolism. Does ROCK2 inhibitor represses the Warburg effect in TPCs? 3BP and Dox had synergistic effect in repressing tumor growth in vivo. It is difficult to dissect whether it was due to direct inhibition of tumor cells, or through modulating TPC functions. Does 3BP and Dox combination inhibit LLC cell growth in vitro? Can the CM from 3BP and Dox treated tumor cells activate HK2 and ROCK2 in pericytes?
Reviewer #3 (Remarks to the Author); expert on ROCK2: This paper nicely demonstrates the distinct differences between normal and tumour associated pericyte function in angiogenesis and demonstrates the use of new isolation protocol to facilitate a deeper understanding of the differences between their signalling in vitro and recapitulated in vivo. Moreover, they reveal distinct differences in glycolysis and the role the ROCK2-MLC2 signalling plays in vasculature normalisation and importantly capitalise upon this to show that manipulation of this can improve chemotherapy delivery.
Can the authors perform some shRNA experiments of manipulation to show ROCK 2 effect are key to this work rather than ROCK1 off target effect-this should be easy enough to perform to demonstrate that ROCK2 rather than pan drugs such as GSK429286A are having a role in improving chemotherapy delivery beyond off target ROCK1 vascular relaxation effects?
Any other rock2 specific manipulation to address the above point would also satisfy this issue.
Lastly can the authors perform experiments that the effect they see in vivo for LLC ( ie improve chemotherapy response) also hold true for HCC cancer as this is a major ( ie half of the paper ) aspect of this work which is not show as a final in vivo experiment? 1 REVIEWER COMMENTS Reviewer #1 (Remarks to the Author); expert on pericytes and endothelial cells: Meng et al have developed a smart and elegant strategy for the isolation and culture of human-derived pericytes from normal and tumour lung and liver specimens. An approach which is lacking in the field, and it will definitely have a strong impact. They have characterized the biology of what they define as "normal" pericytes and "tumor" pericytes and their effect on angiogenesis in vitro and in vivo. They claim that HK2-driven glycolysis in cancer pericytes leads to excessive pericyte contractility, what precludes correct vessel functioning in tumors. By pharmacologically inhibition of HK2, they show an improvement of tumor perfusion and an enhanced effect of chemotherapy. We would like to thank for the reviewer's positive support about the impact of our strategy for the isolation and culture of human-derived pericytes from normal and tumor lung and liver specimens.
Overall, this manuscript is experimentally well performed, and the data are convincing. However, it often seems a bunch of interesting observations which lack a clear message/flow. Some of the parts could be simply improved by a making an effort with the wording and the rationale of the experiments. But in many others, authors should make an effort strengthen the data. We would like to thank for the reviewer's positive comment about our experimental design. We have now rewritten our revised manuscript to improve the message flow and explain the rationale of the experiments with more details. Please see the highlighted areas throughout the maintext.
1. How do authors explain that high HK2 expression reduces vessel dimeter? They claim that that HK2 regulates cell contractility via ROCK expression, but this is hardly convincing from the data provided.

ROCK2 expression in TPC, but have no effect on ROCK1 expression (Please see new supplementary fig. 6h). Together, our results prove the importance of pericyte-HK2 expression/activity in regulating ROCK2 mediated cell contractility in vitro. Please also see result and discussion sections, page 14-15 line 366-384, page 22-24 line 580-645. To confirm the role of pericyte-HK2 in modulating tumor blood vessel perfusion, diameter and subsequent drug delivery, we subcutaneously co-injected A549 (i.e. NSCLC cell line)/MHCC-LM9 (i.e. HCC cell line) cells with either HK2-depleted TPC, ROCK2 overexpressing HK2depleted TPC or scramble transfected TPC into nude mice and then treated the tumor bearing mice with placebo or doxorubicin. Our data show that administration of doxorubicin reduces tumor growth in mice co-injected with A549/MHCC-LM9 cells and HK2-depleted TPC as compared to mice co-injected with A549/MHCC-LM9 cells and scramble transfected TPC (New supplementary figure 10a-h), while co-injection of A549/MHCC-LM9 cells and ROCK2 overexpressing HK2-depleted TPC decreases the enhanced doxorubicin efficacy against tumor growth observed in mice co-injected with A549/MHCC-LM9 cells and HK2-depleted TPC (New supplementary figure 10a-h). No significant difference in total blood vessel number and pericyte coverage is observed between groups (please see new supplementary fig. 10i-k). Interestingly, the tumor blood vessel diameter, perfusion and collagen IV staining, intratumoral Hoechst and doxorubicin level is all increased in the mice co-injected with A549 cells and HK2depleted TPC as compared to the mice co-injected with A549 cells and scramble transfected TPC, while the increased blood vessel diameter, perfusion and collagen IV staining, intratumoral Hoechst and doxorubicin observed is reduced in mice co-injected with A549 cells and ROCK2 over-expressing HK2-depleted TPC (Please see new supplementary figure 10i-o, q). Furthermore, the expression of pericyte-p-MLC2 is reduced in tumors arising from the mice co-injected with A549 and HK2-depleted TPC as compared to A549 cells co-injected with scramble transfected TPC, whilst it is increased in the tumors derived from co-injection of A549 cells and ROCK2 overexpressing HK2-depleted TPC (New supplementary figure 10p). Further doppler ultrasound analysis of microbubble perfusion reveals that depletion of HK2 in TPC enhances A549 tumor blood flow and perfusion in mice as compared with scramble control group (please see new supplementary fig. 10r), while overexpression of ROCK2 in HK2depleted TPC reduces the enhanced tumor blood flow and perfusion observed (Please see new supplementary fig. 10r). Overall, these new results further demonstrate the important role of pericyte-HK2 driven ROCK2-MLC2 mediated contractility in tumor vasculature remodelling and drug delivery in both lung and liver cancers. Please also see result section page 14-15 line 366-384, page 16-20, line 422-541 and discussion section page 25-26 line 663-707.
2. CD146 staining seems very weak, especially in NSLCC, at least in the pictures showed in Figure  1B. Only the depicted big vessel in the inset seems to be positive for CD146 (yellow), the other two are only red (αSMA positive) and the majority of the small vessels are only CD34 positive. Given that the MACS enrichment strategy is based on CD146 expression in vascular cells, this picture generates doubts about the percentage of pericytes obtained after FACS purification ( Figure 1F and H). Also, this percentage is almost double in tumor samples compared with normal tissue. More images and a quantification of CD146 expression in tissue sections of normal and tumoral section will help to clarify this issue.

We apologize for the quality of our images. We now include more representative images of CD146/CD34/-SMA-triple immunostaining from HCC and NSCLC tissues and do a quantification of CD146 and -SMA-double positive (+ve) blood vessels in the sections of normal adjacent tissues and tumors from our NSCLC and HCC patient cohorts respectively (n= 10 NSCLC or HCC patients), showing that around 65-72% of the blood vessels are CD146 and -SMA-double positive in both normal adjacent tissues and tumors from NSCLC and HCC patients respectively, while no significant difference in the percentage of CD146 and -SMAdouble positive blood vessels between normal adjacent tissues and tumors is observed. Please see new figure 1a-f. For the FACS analysis, we now replace the pictures with better representative images of our FACS staging strategy in new figure 1h.
3. It seems that pericytes in vitro behave quite similarly irrespective or the origin. Are isolated pericytes losing their properties/dedifferentiating when cultured? However, when combined with endothelial cells, TPC seem to enhance tube/EC contraction compared to NPC. Again, the data are not convincing. How do authors explain that TPC express higher HK2 levels? see fig 1, 2,  3, supplementary fig. 2, 3, 4). (Fig. 5p-t, new supplementary fig. 6t, u) Hepatology. 2019, 71, 333-343) also see new supplementary fig. 6v, w,  result and discussion section, page 15-16 line 398-411 and page 24-25 line 652-661. 4. Figure 4J. Authors claim that HK2 expression is significantly increased in mural cells in tumors. Quantification of that expression in representative images should be provided. Endothelial cells seem to be contributing to HK2 overexpression too, as well as many other tumoral/stromal cells. This observation compromises the pericyte-driven hypothesis of the effects of in vivo, pharmacological inhibition of HK2. Pre-treated or genetically modified pericytes should be implanted together with tumor cells or, ideally, tumor cells should be implanted in pericyte-specific HK2 deficient mice. fig. 4j and supplementary fig. 5d. -20 line 495-541, and discussion section page 25-26 line  684-699. 5. Conclusions obtained from clinical correlation data ( Figure 4K, Suppl Fig 4E) on HK2 expression are not valid, markers used are not specific of pericytes. PDGFRβ and αSMA are expressed by cancer fibroblasts and CD146 is also expressed in endothelial cells. Figure 2B does not match the qPCR profiles, as fibroblasts in the IF image show much greater expression than qPCR graph depicts compared with pericytes

We would like to thank for the reviewer's suggestion. We have now replaced the image with a better representative picture of PDGFRB staining in fibroblasts. Please see new fig. 2b.
Reviewer #2 (Remarks to the Author); expert on metabolism: Pericytes play an important role in angiogenesis and tumor microenvironment formation. The authors developed a novel method to isolate pericytes from human normal and tumor tissues. This is a robust model for studying the role of pericytes in angiogenesis and cancer progression. The authors discovered that pericytes from tumor had higher glycolytic flux but lower mitochondrial activity, suggesting that the TPC also display the Warburg effect, which alters cell contractility and limit blood supply. This could further exacerbate tumor hypoxia.

We would like to thank for the reviewer's positive feedback about the novelty and robustness of our newly developed pericyte model.
A few additional experiments that could strengthen the conclusions: In 125-128, how did the authors conclude that 'CD146 was expressed in both endothelial cells and pericytes …, while CD34 and PDGFRβwas specifically expressed in endothelial cells and pericytes respectively'? Were independent markers used to confirm those cells are endothelial cells and pericytes respectively? How can they tell that CD146 was expressed in pericyte without showing the co-staining of CD146 and PDGFRβ in Figure1?  Fig. 1a-f, new supplementary fig. 2a, b). No significant difference in the percentage of CD146 and -SMA-double positive blood vessels is observed between normal adjacent tissues and tumors (Fig. 1c, f) 2a-c, new supplementary fig. 3a-e). supplementary fig. 2a, b and introduction page 3 line 56-63, result page 6-7 line 125-146.

Based on the reviewer's suggestion, we now also provide new triple immunostaining data with our tissue sections, indicating that CD146 is co-expressed with PDGFRβ in pericytes within normal adjacent tissues and tumors from NSCLC and HCC patients. Please see new
The data demonstrated a strong association between HK2/glycolysis upregulation and ROCK2 activation. However, it is not clear whether the paracrine factors from tumor cells reprogram metabolism to activate ROCK2, or activate ROCK2 to reprogram metabolism. Does ROCK2 inhibitor represses the Warburg effect in TPCs? see new supplementary fig. 6h, v, w).

To determine whether ROCK inhibitor represses the Warburg effect in TPCs, we now perform lactate concentration measurement and seahorse assays as well as RT-PCR and western blot analysis, showing that treatment with ROCK inhibitor does not affect HK2 expression and glycolysis (i.e. Warburg effect) in NSCLC/HCC derived TPC (please see new supplementary fig. 7a-g). For your interest, we now also show that depletion of HK2 by shRNA down-regulates ROCK2 expression in NSCLC/HCC derived TPC (New supplementary figure 6h). Altogether, these findings suggest that tumor-derived paracrine signal activates HK2 driven pericyte metabolic reprograming to promote ROCK2-ML2 mediated cell contractility. In addition, ROCK2 is the downstream effector of HK2-mediated glycolysis in TPC. Please also see new result and discussion sections page 14 line 368-373, page15-16 line 398-420 and page 24-25 line 652-661.
3BP and Dox had synergistic effect in repressing tumor growth in vivo. It is difficult to dissect whether it was due to direct inhibition of tumor cells, or through modulating TPC functions. Does 3BP and Dox combination inhibit LLC cell growth in vitro? Can the CM from 3BP and Dox treated tumor cells activate HK2 and ROCK2 in pericytes? (please see new supplementary fig. 8t)

, while exposure of NSCLC/HCC derived NPC with conditioned medium from 3BP or/and DOX pre-treated tumor cells can still up-regulate HK2 and ROCK2 expression in NPC as compared to untreated NPC (please see supplementary fig. 9). These data suggest that the synergistic inhibitory effect of 3-BP and Dox on tumor growth observed in vivo is due to the modulation of TPC function. In the revised manuscript, we also performed co-injection of cancer cells with either HK2-depleted TPC, ROCK2 overexpressing HK2 depleted TPC or scramble transfected TPC into nude mice and treated the tumor bearing mice with placebo or doxorubicin, indicating that depletion of HK2 in TPC significantly enhances blood vessel perfusion and diameter, thereby increasing doxorubicin delivery and efficacy against tumor growth, while overexpression of ROCK2 in HK2-depleted TPC reduces the enhanced vascular effect and doxorubicin delivery/efficacy against tumor growth observed, indicating that pericyte-HK2 expression regulates its blood vessel supporting function and subsequent drug delivery via ROCK2 (Please see new supplementary figure 10). In addition, our new tube formation assay also shows that depletion of HK2 in TPC increases HUVEC tube length, branch points and number as compared to HUVEC cocultured with scramble transfected TPC, while overexpression of ROCK2 in HK2 depleted TPC reduces the enhanced effect observed (Please see new supplementary fig. 6h-s). Overall, these new data demonstrate that depletion of pericyte-HK2 enhances blood vessel function and subsequent drug delivery to prohibit cancer growth. Please also see result and discussion section page 14-15 line 366-384, page 18-20 line 477-541 and page 25-26 line 663-707.
Reviewer #3 (Remarks to the Author); expert on ROCK2: This paper nicely demonstrates the distinct differences between normal and tumour associated pericyte function in angiogenesis and demonstrates the use of new isolation protocol to facilitate a deeper understanding of the differences between their signalling in vitro and recapitulated in vivo. Moreover, they reveal distinct differences in glycolysis and the role the ROCK2-MLC2 signalling plays in vasculature normalisation and importantly capitalise upon this to show that manipulation of this can improve chemotherapy delivery.

We would like to thank for the reviewer's positive comment about our findings.
Can the authors perform some shRNA experiments of manipulation to show ROCK 2 effect are key to this work rather than ROCK1 off target effect-this should be easy enough to perform to demonstrate that ROCK2 rather than pan drugs such as GSK429286A are having a role in improving chemotherapy delivery beyond off target ROCK1 vascular relaxation effects? Any other rock2 specific manipulation to address the above point would also satisfy this issue. see new supplementary fig. 10 a-h) fig. 10i-o, q-r)

. In addition, the expression of pericyte p-MLC2 is reduced in tumors arising from mice co-injected with A549 cells and HK2-depleted TPC as compared to mice co-injected with A549 cells and scrambled transfected TPC, while it is increased in the tumors derived from mice co-injected with A549 cells and ROCK2 overexpressing HK2-depleted TPC (please see new supplementary figure 10p) . Overall, our result demonstrates that ROCK2, but not ROCK1, is the downstream effector of pericyte-HK2 mediated blood vessel supporting function. Please also see result and discussion sections page 14-15 line 366-384, page 19-20 line 495-541, page 24 line 639-645, and page 25 line 684-699.
Lastly can the authors perform experiments that the effect they see in vivo for LLC ( ie improve chemotherapy response) also hold true for HCC cancer as this is a major ( ie half of the paper ) aspect of this work which is not show as a final in vivo experiment? fig. 10a-h)