On 2018 Jan 04, Marilyn Tirard commented:

We have published a response to the comment by Wilkinson et al, which can be found with the online version of the original article: https://elifesciences.org/articles/26338#annotations:1fKgOgXREei5NxuiEWEQ0w

On 2017 Sep 20, Kevin Wilkinson commented:

Is the His6-HA-SUMO1 knock-in mouse a valid model system to study protein SUMOylation?

K.A. Wilkinson^1* , S. Martin² , S.K. Tyagarajan³ , O Arancio⁴ , T.J. Craig⁵ , C. Guo⁶ , P.E. Fraser⁷ , S.A.N. Goldstein⁸ , J.M. Henley^1* .

¹ School of Biochemistry, Centre for Synaptic Plasticity, University of Bristol, Bristol, UK. ² Université Côte d’Azur, INSERM, CNRS, IPMC, France. ³ Institute of Pharmacology and Toxicology, University of Zurich, Switzerland. ⁴ Taub Institute & Dept of Pathology and Cell Biology, Columbia University, New York, NY, USA. ⁵ Centre for Research in Biosciences, University of the West of England, Bristol, UK. ⁶ Department of Biomedical Science, University of Sheffield, Sheffield, UK. ⁷ Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada. ⁸ Stritch School of Medicine, Loyola University, Chicago, USA.

*Address for correspondence: kevin.wilkinson@bristol.ac.uk or J.M.Henley@bristol.ac.uk.

Introduction

There is a large and growing literature on protein SUMOylation in neurons and other cell types. While there is a consensus that most protein SUMOylation occurs within the nucleus, SUMOylation of many classes of extranuclear proteins has been identified and, importantly, functionally validated. Notably, in neurons these include neurotransmitter receptors, transporters, sodium and potassium channels, mitochondrial proteins, and numerous key pre- and post-synaptic proteins (for reviews see Henley JM, 2014, Wasik U, 2014, Peng J, 2016, Martin S, 2007, Luo J, 2013, Craig TJ, 2012, Scheschonka A, 2007, Guo C, 2014, Wu H, 2016, Schorova L, 2016). Furthermore, several groups have reported SUMO1-ylated proteins in synaptic fractions using biochemical subcellular fractionation approaches, using a range of different validated anti-SUMO1 antibodies (Martin S, 2007, Feligioni M, 2009, Marcelli S, 2017, Loriol C, 2012) and many studies have independently observed colocalisation of SUMO1 immunoreactivity with synaptic markers (Konopacki FA, 2011, Ghosh H, 2016, Gwizdek C, 2013, Jaafari N, 2013, Hasegawa Y, 2014). Tirard and co-workers (Daniel JA, 2017) directly challenge this wealth of compelling evidence. Primarily using a His6-HA-SUMO1 knock-in (KI) mouse the authors contest any significant involvement of post-translational modification by SUMO1 in the function of synaptic proteins.

On what basis do Daniel et al. argue against synaptic SUMOylation?

Most of the experiments reported by Daniel et al. use a knock-in (KI) mouse that expresses His6-HA-SUMO1 in place of endogenous SUMO1. Using tissue from these mice, followed by immunoprecipitation experiments, they fail to biochemically identify SUMOylation of the previously validated SUMO targets synapsin1a (Tang LT, 2015), gephyrin (Ghosh H, 2016), GluK2 (Martin S, 2007, Konopacki FA, 2011, Chamberlain SE, 2012, Zhu QJ, 2012), syntaxin1a (Craig TJ, 2015), RIM1α (Girach F, 2013), mGluR7 (Wilkinson KA, 2011, Choi JH, 2016), and synaptotagmin1 (Matsuzaki S, 2015). Moreover, by staining and subcellular fractionation, they also fail to detect protein SUMOylation in synaptic fractions or colocalisation of specific anti-SUMO1 signal with synaptic markers. On this basis, they conclude there is essentially no functionally relevant SUMOylation of synaptic proteins.

What are the reasons for these discrepancies?

• Inefficiency of His6-HA-SUMO1 conjugation and compensation by SUMO2/3

A major cause for concern is that there is 20-30% less SUMO1-ylation in His6-HA-SUMO1 KI mice than in wild-type (WT) mice (Daniel JA, 2017, Tirard M, 2012). Moreover, in the paper initially characterising these KI mice, Tirard et al. showed that while total protein SUMO1-ylation is reduced, total SUMO2/3-ylation is correspondingly increased (Tirard M, 2012). Thus, His6-HA-SUMO1 conjugation is significantly impaired and most likely compensated for by increased conjugation by SUMO2/3. Crucially, however, Daniel et al. do not examine modification by SUMO2/3 at any point in their recent study.<br> Given that SUMO modification is notoriously difficult to detect the 20-30% reduction in His6-HA-SUMO1 compared to wild-type SUMO1 conjugation will make it even more technically challenging. Moreover, this deficit in SUMO1-ylation may well be offset by an increase in SUMO2/3-ylation of individual proteins, but this likely compensation was not tested. Since these deficits alone could explain why Daniel et al. failed to detect SUMO1 modification of the previously characterised synaptic substrate proteins it is surprising that they did not attempt to recapitulate SUMO1-ylation of the target proteins under the endogenous conditions in wild-type systems used in the original papers.

• Lack of functional studies on the substrates they examine

Daniel et al. confine their studies to immunoblotting and immunolabelling. However, these techniques address only one aspect of validating a bone fide SUMO substrate. It is at least as important to examine the effects of target protein SUMOylation in functional assays. Function-based approaches such as electrophysiology or neurotransmitter release assays are not reported or even discussed by Daniel et al. This is an extremely important omission. We argue that simply because SUMO1-ylation of a protein is beneath the detection sensitivity in a model system that exhibits sub-endogenous levels of SUMO1-ylation, does not mean that protein is not a functionally important and physiologically relevant SUMO1 substrate.

• Insensitivity or inadequate use of assay systems

Failure to detect GluK2 SUMOylation

GluK2 is a prototypic synaptic SUMO1 substrate that has been validated in exogenous expression systems, neuronal cultures and rat brain (Martin S, 2007, Konopacki FA, 2011, Chamberlain SE, 2012, Zhu QJ, 2012). Daniel et al. attempt to detect SUMOylation of GluK2 using immunoprecipitation experiments from the His6-HA-SUMO1 KI mice. However, a key flaw in this experiment is that the C-terminal anti-GluK2 monoclonal rabbit antibody used does not recognise SUMOylated GluK2 because its epitope is masked by SUMO conjugation. Thus, due to technical reasons, the experiment shown could not possibly detect SUMOylated GluK2 whether or not it occurs in the KI mice.

Subcellular fractionation and immunolabelling

Daniel et al. perform subcellular fractionation and anti-SUMO1 Western blots to compare His6-HA-SUMO1 KI and SUMO1 knockout (KO) mice. In the KI mice they fail to detect SUMO1-ylated proteins in synaptic fractions. Importantly, however, they do not address what happens in WT mice, which, unlike the KI mice, exhibit normal levels of SUMO1-ylation. While the authors provide beautiful images of SUMO1 immunolabelling in neurons cultured from WT, His6-HA-SUMO1 KI mice and SUMO1 KO mice, in stark contrast to previous reports using rat cultures (Martin S, 2007, Konopacki FA, 2011, Gwizdek C, 2013, Jaafari N, 2013), they detect no specific synaptic SUMO1 immunoreactivity in neurons prepared from WT mice. We note, however, that the nuclear SUMO1 staining in neurons from His6-HA-SUMO1 KI mice is weak, and even weaker in WT neurons. Given that a very large proportion of SUMO1 staining is nuclear, these low detection levels would almost certainly rule out visualisation of the far less abundant, but nonetheless functionally important, extranuclear SUMO1 immunoreactivity.

In conclusion

Given these caveats we suggest that the failure of Daniel et al. to detect synaptic protein SUMO1-ylation in His6-HA-SUMO1 KI mice is due to intrinsic deficiencies in this model system that prevent it from reporting the low, yet physiologically relevant, levels of synaptic protein modification by endogenous SUMO1. In consequence, we question the conclusions reached and the usefulness of this model for investigation of previously identified and novel SUMO1 substrates.

On 2017 Sep 20, Kevin Wilkinson commented:

Is the His6-HA-SUMO1 knock-in mouse a valid model system to study protein SUMOylation?

K.A. Wilkinson^1* , S. Martin² , S.K. Tyagarajan³ , O Arancio⁴ , T.J. Craig⁵ , C. Guo⁶ , P.E. Fraser⁷ , S.A.N. Goldstein⁸ , J.M. Henley^1* .

¹ School of Biochemistry, Centre for Synaptic Plasticity, University of Bristol, Bristol, UK. ² Université Côte d’Azur, INSERM, CNRS, IPMC, France. ³ Institute of Pharmacology and Toxicology, University of Zurich, Switzerland. ⁴ Taub Institute & Dept of Pathology and Cell Biology, Columbia University, New York, NY, USA. ⁵ Centre for Research in Biosciences, University of the West of England, Bristol, UK. ⁶ Department of Biomedical Science, University of Sheffield, Sheffield, UK. ⁷ Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada. ⁸ Stritch School of Medicine, Loyola University, Chicago, USA.

*Address for correspondence: kevin.wilkinson@bristol.ac.uk or J.M.Henley@bristol.ac.uk.

Introduction

There is a large and growing literature on protein SUMOylation in neurons and other cell types. While there is a consensus that most protein SUMOylation occurs within the nucleus, SUMOylation of many classes of extranuclear proteins has been identified and, importantly, functionally validated. Notably, in neurons these include neurotransmitter receptors, transporters, sodium and potassium channels, mitochondrial proteins, and numerous key pre- and post-synaptic proteins (for reviews see Henley JM, 2014, Wasik U, 2014, Peng J, 2016, Martin S, 2007, Luo J, 2013, Craig TJ, 2012, Scheschonka A, 2007, Guo C, 2014, Wu H, 2016, Schorova L, 2016). Furthermore, several groups have reported SUMO1-ylated proteins in synaptic fractions using biochemical subcellular fractionation approaches, using a range of different validated anti-SUMO1 antibodies (Martin S, 2007, Feligioni M, 2009, Marcelli S, 2017, Loriol C, 2012) and many studies have independently observed colocalisation of SUMO1 immunoreactivity with synaptic markers (Konopacki FA, 2011, Ghosh H, 2016, Gwizdek C, 2013, Jaafari N, 2013, Hasegawa Y, 2014). Tirard and co-workers (Daniel JA, 2017) directly challenge this wealth of compelling evidence. Primarily using a His6-HA-SUMO1 knock-in (KI) mouse the authors contest any significant involvement of post-translational modification by SUMO1 in the function of synaptic proteins.

On what basis do Daniel et al. argue against synaptic SUMOylation?

Most of the experiments reported by Daniel et al. use a knock-in (KI) mouse that expresses His6-HA-SUMO1 in place of endogenous SUMO1. Using tissue from these mice, followed by immunoprecipitation experiments, they fail to biochemically identify SUMOylation of the previously validated SUMO targets synapsin1a (Tang LT, 2015), gephyrin (Ghosh H, 2016), GluK2 (Martin S, 2007, Konopacki FA, 2011, Chamberlain SE, 2012, Zhu QJ, 2012), syntaxin1a (Craig TJ, 2015), RIM1α (Girach F, 2013), mGluR7 (Wilkinson KA, 2011, Choi JH, 2016), and synaptotagmin1 (Matsuzaki S, 2015). Moreover, by staining and subcellular fractionation, they also fail to detect protein SUMOylation in synaptic fractions or colocalisation of specific anti-SUMO1 signal with synaptic markers. On this basis, they conclude there is essentially no functionally relevant SUMOylation of synaptic proteins.

What are the reasons for these discrepancies?

• Inefficiency of His6-HA-SUMO1 conjugation and compensation by SUMO2/3

A major cause for concern is that there is 20-30% less SUMO1-ylation in His6-HA-SUMO1 KI mice than in wild-type (WT) mice (Daniel JA, 2017, Tirard M, 2012). Moreover, in the paper initially characterising these KI mice, Tirard et al. showed that while total protein SUMO1-ylation is reduced, total SUMO2/3-ylation is correspondingly increased (Tirard M, 2012). Thus, His6-HA-SUMO1 conjugation is significantly impaired and most likely compensated for by increased conjugation by SUMO2/3. Crucially, however, Daniel et al. do not examine modification by SUMO2/3 at any point in their recent study.<br> Given that SUMO modification is notoriously difficult to detect the 20-30% reduction in His6-HA-SUMO1 compared to wild-type SUMO1 conjugation will make it even more technically challenging. Moreover, this deficit in SUMO1-ylation may well be offset by an increase in SUMO2/3-ylation of individual proteins, but this likely compensation was not tested. Since these deficits alone could explain why Daniel et al. failed to detect SUMO1 modification of the previously characterised synaptic substrate proteins it is surprising that they did not attempt to recapitulate SUMO1-ylation of the target proteins under the endogenous conditions in wild-type systems used in the original papers.

• Lack of functional studies on the substrates they examine

Daniel et al. confine their studies to immunoblotting and immunolabelling. However, these techniques address only one aspect of validating a bone fide SUMO substrate. It is at least as important to examine the effects of target protein SUMOylation in functional assays. Function-based approaches such as electrophysiology or neurotransmitter release assays are not reported or even discussed by Daniel et al. This is an extremely important omission. We argue that simply because SUMO1-ylation of a protein is beneath the detection sensitivity in a model system that exhibits sub-endogenous levels of SUMO1-ylation, does not mean that protein is not a functionally important and physiologically relevant SUMO1 substrate.

• Insensitivity or inadequate use of assay systems

Failure to detect GluK2 SUMOylation

GluK2 is a prototypic synaptic SUMO1 substrate that has been validated in exogenous expression systems, neuronal cultures and rat brain (Martin S, 2007, Konopacki FA, 2011, Chamberlain SE, 2012, Zhu QJ, 2012). Daniel et al. attempt to detect SUMOylation of GluK2 using immunoprecipitation experiments from the His6-HA-SUMO1 KI mice. However, a key flaw in this experiment is that the C-terminal anti-GluK2 monoclonal rabbit antibody used does not recognise SUMOylated GluK2 because its epitope is masked by SUMO conjugation. Thus, due to technical reasons, the experiment shown could not possibly detect SUMOylated GluK2 whether or not it occurs in the KI mice.

Subcellular fractionation and immunolabelling

Daniel et al. perform subcellular fractionation and anti-SUMO1 Western blots to compare His6-HA-SUMO1 KI and SUMO1 knockout (KO) mice. In the KI mice they fail to detect SUMO1-ylated proteins in synaptic fractions. Importantly, however, they do not address what happens in WT mice, which, unlike the KI mice, exhibit normal levels of SUMO1-ylation. While the authors provide beautiful images of SUMO1 immunolabelling in neurons cultured from WT, His6-HA-SUMO1 KI mice and SUMO1 KO mice, in stark contrast to previous reports using rat cultures (Martin S, 2007, Konopacki FA, 2011, Gwizdek C, 2013, Jaafari N, 2013), they detect no specific synaptic SUMO1 immunoreactivity in neurons prepared from WT mice. We note, however, that the nuclear SUMO1 staining in neurons from His6-HA-SUMO1 KI mice is weak, and even weaker in WT neurons. Given that a very large proportion of SUMO1 staining is nuclear, these low detection levels would almost certainly rule out visualisation of the far less abundant, but nonetheless functionally important, extranuclear SUMO1 immunoreactivity.

In conclusion

Given these caveats we suggest that the failure of Daniel et al. to detect synaptic protein SUMO1-ylation in His6-HA-SUMO1 KI mice is due to intrinsic deficiencies in this model system that prevent it from reporting the low, yet physiologically relevant, levels of synaptic protein modification by endogenous SUMO1. In consequence, we question the conclusions reached and the usefulness of this model for investigation of previously identified and novel SUMO1 substrates.

On 2018 Jan 04, Marilyn Tirard commented:

We have published a response to the comment by Wilkinson et al, which can be found with the online version of the original article: https://elifesciences.org/articles/26338#annotations:1fKgOgXREei5NxuiEWEQ0w