In memoriam: John Lisman – commentaries on CaMKII as a memory molecule - Bear2018_Article

C O M M E N T A R Y

Open Access

In memoriam: John Lisman

–

commentaries

on CaMKII as a memory molecule

Mark F. Bear

, Sam F. Cooke

, Karl Peter Giese

, Bong-Kiun Kaang

, Mary B. Kennedy

, Ji-il Kim

Richard G. M. Morris

and Pojeong Park

Abstract

Shortly before he died in October 2017, John Lisman submitted an invited review to Molecular Brain on

‘

Criteria for

identifying the molecular basis of the engram (CaMKII, PKM

)

’

. John had no opportunity to read the referees

’

comments,

and as a mark of the regard in which he was held by the neuroscience community the Editors decided to publish his

review as submitted. This obituary takes the

form of a series of commentaries on Lisman

’

sreview.Atthesametimewe

are publishing as a separate article a longer response by Todd Sacktor and André Fenton entitled

‘

What does LTP tell us

about the roles of CaMKII and PKM

in memory?

’

which presents the case for a rival memory molecule, PKM

John Lisman 1944

–

2017

Contributed by Karl Peter Giese

Last October John Lisman sadly passed at the age of

73. John was an exceptional neuroscientist who made a

wide variety of seminal contributions. His very highly

cited work includes the development of novel theories

about the role of bursts in information processing [

1],

the storage of short-term memories in oscillatory cycles

[

2], the function of the hippocampal-VTA in long-term

memory formation [

3], and the CaMKII (Ca

/calmodu-

lin-dependent protein kinase II) hypothesis for memory

storage. The latter CaMKII hypothesis was John

’

s most

important contribution. It originated from a theoretical

paper in 1985 where he proposed that the molecular

basis for memory storage could be an autophosphorylat-

ing kinase that persistently maintains activity at synapses

[

4]. Shortly after John

’

s publication CaMKII was

biochemically purified from synapses and shown to have

an autophosphorylation switch. Together, with John

’

theoretical paper this led to the CaMKII hypothesis for

memory storage [

5]. The attractiveness of this hypothesis

motivated many eminent neuroscientists to study the

function of CaMKII and its isoforms in learning and

memory. Accumulating experimental evidences not only

supported, but also provided some problems for the

CaMKII hypothesis, which motivated refinements that

to keep the hypothesis viable [

–

9]. Last year John was

very delighted to publish experimental proof for his

CaMKII hypothesis after more than 30 years of research

[

9, 10]. Next to his pioneering work on memory storage,

John was also a true scholar who was open for discus-

sion of other viewpoints. I experienced this myself when

questioning the CaMKII hypothesis over many years.

John will be remembered not only for his seminal con-

tributions, but also for the breadth and generosity of his

academic spirit in advancing our understanding of the

brain.

The enigma of memory maintenance: Commentary

on Criteria for identifying the molecular basis of the

engram (CaMKII, PKM

ζ

)

Contributed by Sam F. Cooke and Mark F. Bear

The late John Lisman was one of the great thinkers of

neuroscience. He posed some of the biggest questions and

provided eminently testable hypotheses. Amongst his

many contributions, he is probably best known for illus-

trating how the enzyme Calcium/calmodulin-dependent

* Correspondence:

mbear@mit.edu

; samuel.cooke@kcl.ac.uk

;

karl.giese@kcl.ac.uk

; kaang@snu.ac.kr

; kennedym@caltech.edu

;

r.g.m.morris@ed.ac.uk

Picower Institute for Learning and Memory, Department of Brain and

Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA

02139, USA

King

’

s College London, Department of Basic and Clinical Neuroscience,

Institute of Psychiatry, Psychology and Neuroscience, De Crespigny Park,

London SE5 8AF, UK

Department of Biological Sciences, Seoul National University, Gwanak-gu,

Seoul, Republic of Korea

The Division of Biology and Biological Engineering, California Institute of

Technology, Pasadena, CA 91125, USA

Laboratory for Cognitive Neuroscience, Centre for Discovery Brain Sciences,

Edinburgh Neuroscience, Edinburgh EH8 9JZ, UK

Open Access

This article is distributed under the terms of the Creative Commons Attribution 4.0

International License (

http://creativecommons.org/licenses/by/4.0/

), which permits unrestricted use, distribution, and

reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to

the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver

(

http://creativecommons.org/publicdomain/zero/1.0/

) applies to the data made available in this article, unless otherwise stated.

Bear

et al. Molecular Brain

(2018) 11:76

https://doi.org/10.1186/s13041-018-0419-y

protein kinase II (CaMKII) may serve as a molecular

switch to maintain changes in synaptic strength and

long-term memory [

4, 5, 11, 12].

In the early 1980

’

s, variations on Hebbian synaptic

plasticity had started to predominate as the favoured

theoretical mechanism by which learning occurs in the

brain, several decades after the key concepts had first

been formalized by Donald Hebb [

13] and a decade after

the first example of lasting Hebbian plasticity, known as

long-term potentiation (LTP), was observed in the brain

of living organisms [

14, 15]. LTP has several remarkable

properties: It is

“

input specific

”

, meaning that it can

occur selectively at a very small fraction of the synapses

formed on a given neuron [

16]; it is rapidly induced

within seconds [

15]; and it can last months in vivo [

17]

and many hours in brain slices [

18]. These properties

are ideal for encoding episodic memories, but present

challenges for understanding the underlying cellular

biology. Longevity of LTP could be explained by stable

activation of gene expression and new protein synthesis,

but how would these new proteins be continuously sup-

plied to (or captured by) only the potentiated synapses,

and how would this mechanism account for the rapid

induction of LTP at synapses far away from the nucleus?

On the other hand, rapid induction and synapse specifi-

city could be readily accounted for by local

post-translational modification of pre-existing synaptic

proteins, but how would this be maintained in the face

of protein turnover?

Much thought has been expended on these questions

and the most influential solutions came from Lisman [

and, independently, from another deep thinker, Francis

Crick [

19]. To address these twin problems, the concept of

‘

memorase

’

was introduced, which is an enzyme com-

posed of multiple subunits t

hat can auto-regulate each

other to sustain activity and maintain synaptic change by

continually modifying key syn

aptic proteins. Thus, newly

synthesized subunits could be incorporated at any time to

replace outgoing subunits and switched into an active form

through post-translational m

odification by a neighbouring

subunit, thereby mai

ntaining the overall

‘

active

’

state of the

molecular complex. Lisman and Goldring went so far as to

posit the likeliest candidate, CaMKII [

5, 11], which is a kin-

ase that typically exists as a 12-subunit, double hexamer

structure that autophosphorylates neighbouring subunits to

maintain those subunits in an open and active state [

20].

Among many other actions, active CaMKII phosphorylates

post-synaptic AMPA receptors to enhance their efficacy,

which is a key expression mechanism for LTP and memory

[

21] and produces other syn

aptic alterations [

22]. This con-

cept was highly influential because it addressed both prob-

lems: (1) LTP could initially be induced by activation of

existing, local CaMKII doub

le hexamers by calcium influx

through the NMDA receptor and (2) new,

unphosphorylated subunits

couldbeincorporatedinto

CaMKII double hexamers without interrupting its activity

state and, thanks to autophosphorylation of neighbouring

subunits, perpetuate its activity state in a way that is im-

mune to molecular turnover and local to the synapse at

which potentiation originally occurred [

–

25].

Even the subsequent discoveries that hippocampal

synapses were bidirectionally modifiable, allowing low

frequency (1 Hz) stimulation to induce calcium-

dependent, Hebbian long-term depression (LTD) [

26,

], and that this LTD was mediated in part by protein

phosphatases [

–

30], was presaged by Lisman in a

modified CaMKII model [

12]. In this updated model,

phosphatases such as calcineurin respond to lower cal-

cium concentrations than CaMKII and act to dephos-

phorylate CaMKII targets and CaMKII itself, thereby

setting up a bistable switch at synapses that can adopt a

potentiated or a depressed state. Through its elegance

and simplicity, this overall model has been highly influ-

ential, and it became the textbook example of how last-

ing synaptic modification may occur at individual

synapses. Much experimental evidence supported the

model as well, notably the demonstration that phosphor-

ylation of the T286 residue of

CaMKII, which main-

tains CaMKII in an open state and is the key site for

sustained activity through autophosphorylation, is re-

quired for LTP and memory [

31]. However, two major

experimental observations were strongly at odds with

the model: First, blockade of CaMKII activity with se-

lective inhibitors prevents LTP induction and learning,

but does not prevent LTP maintenance or memory if ap-

plied very shortly after induction/learning [

–

34] and,

second, CaMKII activity is not sustained for more than a

few minutes after LTP induction [

–

37].

While CaMKII has failed to be fully anointed as the

molecular basis of the engram, PKM

, which is an atyp-

ical isoform of the calcium-dependent protein kinase

(PKC), has emerged as an alternative candidate to sup-

port long-lasting, synapse-specific potentiation and

memory. The model of how PKM

may mediate

synapse-specific potentiation and remain immune to

molecular turnover is somewhat different from CaMKII,

but no less elegant [

38]. In this case, mRNA transcripts

encoding PKM

(transcribed from the PRKC

gene), are

freely available within the cell, but translation into PKM

protein is blocked by a protein called PIN1 (protein

interacting with NIMA1) [

39]. Only once the LTP/learn-

ing event activates calcium-dependent synaptic signal-

ling, which includes CaMKII, is this PIN1 block on the

translation of PKM

removed [

40]. PKM

can then act

to increase the number of post-synaptic AMPA recep-

tors through interactions with NSF-dependent AMPA

receptor trafficking [

41, 42]. Critically, the newly-

synthesized PKM

then suppresses further PIN1 activity,

Bear

et al. Molecular Brain

(2018) 11:76

Page 2 of 12

perpetuating its own synthesis [

39]. For this model to

work as a synapse-specific potentiation system, it is as-

sumed that protein synthesis is localised to each synapse

due to proximal positioning of ribosome complexes, for

which there is substantial evidence [

43, 44].

A wealth of evidence is consistent with this kinase

serving as a molecular basis of the engram. Most strik-

ingly, application of an inhibitor peptide, ZIP, which was

designed to be highly selective for PKM

, reverses

long-established LTP and memory [

45, 46]. Additionally,

the synthesis of constitutively active PKM

is increased

for at least a month after LTP is induced [

47], in con-

trast to observations of CaMKII activity [

35, 36]. These

findings have had a dramatic impact on the field of

learning and memory because, until very recently, there

had not been an example of a molecule that could be

targeted to erase established LTP and memory. As with

CaMKII, however, major experimental counter-evidence

clouds the waters. First, genetic knockouts of the PRKC

gene (encoding PKC

and PKM

) do not prevent LTP or

learning and memory in a mouse [

48, 49] and, second,

the specific inhibitor ZIP is equally effective at erasing

LTP and memory in these knockout mice [

49, 50], dem-

onstrating that it is not as specific as it was designed to

be. Thus, both CaMKII and PKM

are compelling at a

conceptual level, but their supporters are left with some

major explaining to do.

In his article

‘

Criteria for Identifying the Molecular

Basis of the Engram

’

, John Lisman has built a case for

CaMKII as the strongest candidate for the molecular

basis of the engram [

10]. He does so by comparing it

directly to PKM

and asking how well the experimental

evidence for each stack up in relation to key criteria:

First, that the enzyme is necessary for LTP/memory.

Second, that its overactivity will saturate synaptic

strength and occlude LTP/memory and, third, that spe-

cific blockade after induction/learning will lead to LTP/

memory erasure. He also discusses a fourth consider-

ation, which is how well each mechanism could serve as

a local maintenance system. While Lisman has made a

compelling case, it is important to be clear that the evi-

dence has been selectively presented. First, Lisman ar-

gues that there is more evidence for CaMKII being

necessary for LTP/memory than there is for PKM

based on the observation that LTP and memory are

retained in PKC

knockout mice [

48, 49], while the

T286A

CaMKII point mutant mice, that do not main-

tain autonomous activity of

CaMKII through autophos-

phorylation, show very little LTP and highly deficient

hippocampus-dependent memory [

31]. However, this is

a somewhat unfair comparison. Lisman neglects to dis-

cuss the fact that the full

CaMKII knockout mice,

which provides the most direct comparison with the

PKM

knockout mouse, retain a substantial amount of

LTP at these synapses [

51, 52] and show slowed

hippocampus-dependent learning that can nonetheless

reach a normal asymptote over time [

53], likely due to

compensation by the beta isoform of CaMKII [

54]. This

type of genetic compensation is precisely the explanation

that seems most likely for the maintained plasticity and

memory in the PKC

knockout mouse, where the iota/

lamda isoform of PKC likely supports most duties in the

absence of PKM

[50]. Furthermore, even where LTP ap-

pears completely absent at the Schaffer collateral-CA1

synapses in the T286A point mutant, LTP can be in-

duced by alternative NMDA receptor-dependent means

at medial perforant path-granule cell synapses in the

dentate gyrus subfield of the hippocampus [

55], and last-

ing memory can be attained with overtraining [

56]. In

the strongest sense of

‘

necessity

’

for LTP/memory main-

tenance, therefore, neither CaMKII nor PKM

meets the

criterion. However, there is compelling evidence that,

under normal circumstances, both play a critical role in

many forms of synaptic plasticity and memory.

A more complex criterion to address is the stipulation

that overactivity of a memory mechanism should satur-

ate synaptic strength and thereby occlude LTP/memory.

As presented by Lisman in

‘

Criteria for Identifying the

Molecular Basis of the Engram

’

, overactivity of CaMKII

increases synaptic strength and prevents LTP induction

or learning [

9, 57], while a similar treatment for PKM

enhances LTP and learning [

58]. Thus, CaMKII appears

to meet the occlusion criterion while PKM

fails. How-

ever, if we were to select a different set of representative

studies, then overexpression of activated CaMKII has

been shown to enhance some forms of LTP [

59] while

overexpression of activated PKM

increases synaptic

transmission and blocks further LTP [

60]. Thus, there is

little consensus yet as to which of these two molecules

best meets the occlusion criterion, reflecting the fact

that, despite much excellent work, the true molecular

basis of the engram is not fully established.

Next, Lisman argues that while ZIP is highly effective

at erasing LTP and memory, his laboratory has identified

better ways of doing precisely the same for CaMKII.

They find in slices that a dose (20

M) of the

cell-permeable peptide tatCN21 that both inhibits CaM-

KII and interferes with the association of the enzyme

with postsynaptic NMDA receptors is sufficient to de-

press basal synaptic transmission and erase established

LTP [

61]. Additionally, a transiently-expressing herpes

simplex viral (HSV) vector, which temporarily produces

a dominant negative form of CaMKII, erases established

place avoidance memory [

9]. Moreover, he states that an

additional important test is included in these experi-

ments that has not been conducted in the key ZIP stud-

ies [

, 46, 60], namely that synapses are re-potentiated,

and animals re-learn after erasure, demonstrating that

Bear

et al. Molecular Brain

(2018) 11:76

Page 3 of 12

the necessary synaptic machinery remained intact des-

pite the presence of a foreign agent. This is a crucial

point, and more effort should certainly be made to

evaluate the erasing effects of ZIP, given some indication

that ZIP may damage cells [

62] or target other neural

processes than synaptic function [

63]. However, it is

worth noting that evidence exists that plasticity can be

restored after erasure by ZIP [

64] and that alternative

pharmaceutical inhibitors of PKM catalytic activity also

erase established LTP [

60].

The findings with a high dose of the inhibitor tatCN21

stand in contrast to the effects of a lower dose. Unlike

other pharmaceutical inhibitors such as KN62 and KN93,

tatCN21 at a dose of 5

M inhibits CaMKII phosphoryl-

ation of substrates even when the enzyme has been

switched into an activated, autonomous state [

33, 61].

Interestingly however, at this dose LTP maintenance is

unaffected. This finding therefore calls into question the

original model of CaMKII as a self-perpetuating

‘

memor-

ase

’

[33]. To reconcile these findings, it is necessary to

further postulate existence of a critical contribution of

the physical association of CaMKII and NMDA receptors

in the postsynaptic density. At high doses, tatCN21 and

related peptides disrupt CaMKII binding to the NMDAR

[

61, 65], and it is argued that the dose dependency of the

tatCN21 peptide on LTP maintenance is reflective of the

need for a CaMKII-NMDAR complex in LTP mainten-

ance. Sanhueza and Lisman have suggested that this

interaction can contribute to the structural modifications

that are believed to support long-term functional modifi-

cations of synapses [

65], and more work is required to

understand what these structural changes may be. While

this new model revives a critical role for CaMKII in

long-term synaptic modifications underlying memory, it

is somewhat removed from the elegance of the original

Crick/Lisman hypotheses [

4, 5, 11, 19].

A last point for discussion is Lisman

’

scontentionthatit

is difficult to justify PKM

as a synapse-specific mechanism

of maintenance given the fact that that it appears to be in-

creased in expression throughout neurons after learning

and has a major effect on gene expression at the nucleus as

aresult[

66]. Bearing in mind that PKM

canonlybeinan

active state, this contrasts dramatically with the observation

that CaMKII activity remains very local to the spine where

it is induced [

36, 37], and suggests that if PKM

were the

primary molecular basis of the engram it would likely exert

its influence on surr

ounding synapses, a prospect that is

clearly at odds with the input specificity of LTP [

16]and

Hebbian synaptic plasticity in general [

13]. Lisman certainly

raises a valid concern here, but it is also perfectly possible

that PKM

does not reach sufficient concentration in

neighbouring spines to suppress PIN1 activity and switch

the synapse into a stable

‘

memory

’

state. This is an issue

that requires serious exper

imental investigation. The

proposal from Lisman that PKM

is, for this reason, better

suited as a heterosynaptic scaling mechanism does, how-

ever, seem flawed, as scaling is proposed to be a homeo-

static mechanism that potenti

ates synapses after prolonged

periods of reduced neural activity [

67, 68], whereas it is pro-

nounced, event-related stimul

ation of neural activity that

leads to PKM

expression [

47, 69, 70]. It would be very use-

ful to image PKM

distribution and concentration as LTP is

induced at single synapses, as this would provide some

insight into its differential contributions to synapse-specific

and whole-cell memory mechanisms.

John Lisman made an extraordinary contribution to

neuroscience. He strived to pull together threads of data

from multiple sources to clear

ly articulate models for mem-

ory storage, lay out the critical tests of these models, and

challenge the field (includin

g his own laboratory) to refute

them with rigorous experiments. His voice will be sorely

missed in this debate. After much work from many scien-

tists we cannot question the idea that CaMKII autophos-

phorylation is required for learning and LTP induction,

proving that the

‘

memorase

’

property of the kinase does at

least bridge a critical period between NMDA receptor

opening and AMPA receptor modification. CaMKII may

well play a key role in memory maintenance, although

through a different mechanism

than originally envisioned,

and these are just two parts of a much larger scientific leg-

acy. For science to progress there must always be innovative

ideas, but there must always be rigorous testing and criti-

cism of these ideas to determine their validity. John Lisman

has been a powerful force in both processes, as illustrated

in polemical pieces such as

‘

Criteria for identifying the Mo-

lecular Basis of the Engram

’

. His article succeeds in illus-

trating that the evidence for CaMKII as a memory

maintenance mechanism is no weaker than that for PKM

The clearest message that we can take from this is that

much more work needs to be done to understand the mo-

lecular basis of the engram. Whether there is truly one

mechanism that governs memory remains open to ques-

tion. Typically, evolution would not arrive at such a fragile

solution, and tends to re-purpose a multitude of existing

mechanisms to fulfil similar, overlapping roles [

71]. Both

PKM

and CaMKII are conserved systems, being present

and operational in synaptic plasticity in both invertebrates

and vertebrates [

72, 73], but there are many such conserved

candidate systems. It therefor

e seems unlikely that either

CaMKII or PKM

serves as the singular molecular basis of

the engram [

74], but there is clear and compelling evidence

that each plays a central role.

Commentary on

‘

Criteria for identifying the mo-

lecular basis of the engram (CaMKII, PKM

ζ

)

’

Contributed by Karl Peter Giese.

How the brain stores memory is a fundamental ques-

tion in neuroscience. John Lisman pioneered the mo-

lecular investigation of this question. Back in 1985 he

Bear

et al. Molecular Brain

(2018) 11:76

Page 4 of 12

hypothesized that autophosphorylation of a kinase may

be a molecular switch that leads to persistent, memory-

storing signaling at synapses [4; for independent pio-

neering work see, 19]. Over the last three decades John

Lisman

’

s theoretical work inspired many experiments

that eventually gave rise to a very sophisticated CaMKII

hypothesis of memory storage [

–

7, 10]. Later, a differ-

ent hypothesis emerged, stating that PKM

could pro-

vide memory-storing signaling at synapses [

38].

Currently, it is unclear whether the CaMKII hypothesis

and the PKM

hypothesis are mutually exclusive, or

whether both kinases together mediate memory storage

(see also, [

74]). In his last review John Lisman critically

discusses the evidence for and against the two hypoth-

eses [

10]. He also explains fundamental concepts and in-

troduces criteria for the molecular basis of memory

storage. Whilst I think that this review is very important,

I disagree with some points, as follows.

One of the proposed criteria is the saturation/occlusion

test. The idea behind this test is that overexpression of an

activated form of the memory-storing protein after mem-

ory formation is completed should increase most synaptic

weights, preventing memory access [

9]. However, this test

is confounded if the protein of interest is involved in

retrieval-induced memory destabilization that precedes re-

consolidation/restabilization [

75]. Specifically, this is the

case for CaMKII. Using inducible overactivation of CaM-

KII, Joe Tsien

’

s lab showed that increased CaMKII at the

time of retrieval causes memory loss [

76]. Further, they

established that this is not due to deficits in accessing the

acquired memory, because switching off CaMKII overacti-

vation did not restore the lost memory. Therefore, it was

concluded that CaMKII overactivation enhances memory

destabilization, leading to memory erasure, and that this

process requires protein degradation [

77]. Consistent with

this idea, CaMKII regulates protein degradation at the

synapse [

78] and blocking CaMKII activity impairs

retrieval-induced protein degradation and memory

destabilization [

79]. Furthermore, increasing CaMKII ac-

tivity by knockdown of an endogenous CaMKII inhibitor

protein leads to enhanced memory destabilization [

80].

Rossetti et al. [

9] found that after memory formation over-

expression of constitutively active CaMKII impairs mem-

ory and in his review John Lisman writes that these results

would

‘

support the concept that memory is mediated by

an LTP-like process dependent on CaMKII

’

. However,

considering the evidence discussed above, the occlusion/

saturation test for CaMKII is inconclusive; it is very likely

that the memory impairment observed by Rossetti et al.

[

9] is caused by enhanced memory destabilization rather

than a deficit in memory access.

An intriguing alternative to the saturation/occlusion

test is an enhancement test. The idea behind such test is

that overexpression of the memory-storing protein after

memory formation (including consolidation) would lead

to increased memory due to elevated expression at rele-

vant synapses. Yadin Dudai

’

s lab demonstrated the en-

hancement of previously consolidated memory by PKM

overexpression, supporting the PKM

hypothesis [

81].

In addition to a saturation/occlusion test, John Lisman

proposed a necessary test as criteria. This test involves

blocking the function of a candidate memory-storing pro-

tein already at the time of training. The necessary test is

most powerful for exclusion of a candidate mechanism

when the blockade does not prevent memory storage.

Both

CaMKII and PKC

/PKM

knockout mouse lines

can form spatial memory [

50, 54]. Thus, in principle, these

knockout studies reject both the

CaMKII and the PKM

hypothesis. However, it is not known whether long-term

retention of spatial memory is impaired in these knock-

outs. Moreover, both knockout mouse lines have a com-

pensatory upregulation by related kinase isoforms [

50, 54].

Such compensation does not occur in some knockin mu-

tants. In calcium/calmodulin-binding-deficient-

CaMKII

(T305D) mutants there is no compensatory translocation

of ßCaMKII into the postsynaptic density and spatial

memory formation is prevented [

54]. Threonine-286

autophosphorylation-deficient

CaMKII (T286A) mutants

also cannot form spatial memory, even when environmen-

tal enrichment is provided [

31, 82]. Thus,

CaMKII

knockin studies pass the necessary test, unlike the

CaM-

KII knockout studies. However, it should be pointed out

that aversive, hippocampus- and amygdala-dependent

memories can be formed in the T286A knockin mutants,

indicating that the threonine-286 autophosphorylation of

CaMKII is not required for some types of memory stor-

age [

34, 56, 83, 84]. Therefore, the model provided by Ros-

setti et al. [

9], and discussed in John Lisman

’

s review, may

only apply to spatial memory storage and not storage of

other types of memory.

The key test to determine the necessity of a

memory-storing molecule is the erasure test. The eras-

ure test consists of functional inactivation after memory

has been formed and this inactivation should cease be-

fore memory is tested (to exclude an impact on re-

trieval). Many studies have shown that treatment with

ZIP (PKM

inhibitory peptide) erases established mem-

ories [

85]. Almost all types of memories seem to be sen-

sitive to ZIP erasure. However, ZIP not only blocks

PKM

, but also other molecules [

–

50]. Thus, currently

there is no erasure test to support the PKM

hypothesis.

In contrast Rossetti et al. [

9] suggest that they have car-

ried out a successful spatial memory erasure by blocking

CaMKII function. However, in my opinion more experi-

mental evidence is needed to assure that Rossetti et al.

specifically blocked CaMKII function. They overex-

pressed

CaMKII with a point mutation (K42 M) that

blocks its catalytic activity. Unfortunately, Rossetti et al.

Bear

et al. Molecular Brain

(2018) 11:76

Page 5 of 12

did not perform a biochemical characterization of this

overexpression, addressing, for example a change in

CaMKII activity in the postsynaptic density. Other stud-

ies with heterozygous K42 M knockin mutants (express-

ing 50% wild-type

CaMKII and 50% mutated

CaMKII) have not found a dominant-negative effect on

CaMKII activity in forebrain [

84]. Moreover, John Lis-

man

’

s lab showed that overexpression of the K42M mu-

tant impairs synaptic transmission in a dominant-

negative fashion in hippocampal slices [

86]. However,

K42M knockin mutants do not show impaired synaptic

transmission in the hippocampus [

84]. Taken together,

this suggests that the K42 M overexpression might lead

to unwanted side effects that might be the cause of the

memory erasure.

In conclusion, John Lisman

’

s review presents very im-

portant conceptual points regarding the molecular basis

of memory storage. It also analytically discusses the

PKM

hypothesis, but is less critical of the CaMKII hy-

pothesis. In my view, both hypotheses could be correct,

but more experimental evidence is needed to ascertain

this.

Commentary on

“

Criteria for identifying the mo-

lecular basis of the engram (CaMKII, PKM

ζ

)

”

Contributed by Ji-il Kim, Pojeong Park and Bong-Kiun

Kaang.

John Lisman was a respected scholar with a great pas-

sion for science. Throughout his life, he made outstand-

ing achievements in various scientific fields. In

particular, he devoted most of his career to answering

the question,

“

What is the molecular basis of memory?

”

He hypothesized that a molecular switch triggered by

recent activities could maintain synaptic potentiation

and memory [

4]. Once the molecular switch is activated,

memory can be maintained through a reverberating

positive feedback loop. Lisman found that Ca2+/calmo-

dulin-dependent protein kinase II (CaMKII) has many

properties that met his hypothesis and followed up on

this during his lifetime to prove that this brain-enriched

complex is the

“

memory molecule.

”

In Lisman

’

s posthumously published paper, we can ap-

preciate his unique and great ideas regarding his original

question,

“

What is the molecular basis of memory?

”

[

10]. In particular, unlike the general perspective of the

engram studied at the circuit, cell, and synapse levels

[

87], he stated that the engram constitutes the molecular

changes by which a memory is stored in the brain. In

addition, he marked a milestone in the analysis of the

molecular engram by presenting appropriate criteria in

his paper [

10].

In the first part of this paper, Lisman described the re-

lationship between memory and long-term potentiation

(LTP) at a glance and discussed why and how studying

LTP contributes to understanding the nature of memory.

In a similar vein, he also emphasized the importance of

understanding the molecular mechanism of LTP main-

tenance to identify the memory molecule.

In addition, Lisman suggested three types of tests (Ne-

cessary, Saturation/Occlusion, and Erasure tests), which

would be used to evaluate the role of molecules in the

maintenance of LTP and memory. By comparing two

major candidates of the memory molecule, protein kin-

ase M

(PKM

) and CaMKII, based on his three types

of tests, he suggested that CaMKII, rather than PKM

,is

more appropriate as the memory molecule.

Subsequently, Lisman divided the LTP process into

three different phases: Induction (the triggering of signal

cascades by, for example, calcium), Maintenance (the

throwing of a molecular switch such as autophosphoryl-

ation of CaMKII) and Expression (involving the down-

stream targets (AMPA receptors) of the activated

Maintenance molecule(s)). Memory, Lisman argues, will

only be permanently erased by manoeuvres that attack

the molecular mechanisms responsible for Maintenance.

This subdivision is of importance to understand whether

a candidate molecule is specifically responsible for LTP

maintenance, rather than for other phases. Therefore,

Lisman emphasized that before a candidate molecule is

interpreted as being important in the maintenance

process, it should be confirmed whether the impairment

of LTP is caused by a perturbation during the induction

or expression processes. Data interpretation based on

these subprocesses might reduce any confusion in

searching for the memory molecule regulating the main-

tenance process, which underlies storage of the engram.

In summary, Lisman

’

s insightful theories, such as of the

three types of tests for evaluating candidate molecules

and interpretation based on subprocesses, will serve as

significant milestones for future memory research.

Despite Lisman

’

s significant contributions, some im-

portant questions remain that future generations should

pursue to fully understand the molecular basis of

memory.

1. It is clear that synaptic potentiation occurs after ap-

propriate stimuli, and the synaptic potentiation between

engram cells underlies memory storage [

88]. Considering

the relationship between synaptic potentiation and mem-

ory, Lisman insisted that maintenance of synaptic potenti-

ation by the molecular switch (that is, CaMKII) is the

mechanism that maintains memory over a long period of

time. In other words, the enhancement of the synapses in

which the potentiation occurs during learning should be

maintained persistently to store the memory.

However, the place where memory is stored in our

brain might be highly dynamic. It is well known that epi-

sodic memories in rodents are initially stored in the

hippocampus for several weeks, but are gradually trans-

ferred to cortical areas [

89]. Moreover, recent studies

Bear

et al. Molecular Brain

(2018) 11:76

Page 6 of 12

have demonstrated that engram cells in various brain re-

gions are continuously changing [

–

93]. According to

these results, synapses that store memory are seemingly

not a static, but a highly dynamic structure after learn-

ing. If the engram cells are dynamically changing, then

the memory-storing engram synapses might also be dy-

namically changing, and the enhancement of synapses

that are initially activated during learning may not need

to be sustained for a long time to maintain memories.

Therefore, although it is necessary to study how molecu-

lar cascades could be maintained over a long period of

time at the synapse level, future studies will need to

adopt a systems approach to explain how a qualitatively

similar memory can be preserved in the light of memory

engram dynamics.

2. In addition to CaMKII, many other molecules such

as PKM

and cytoplasmic polyadenylation element-

binding protein (CPEB) are known to be involved in

memory maintenance [

38, 94]. As Lisman insisted, CaM-

KII might be the only molecule that meets the three cri-

teria, but there are considerable data that other

molecules are also involved in memory maintenance. In

particular, PKM

is of interest as a key molecule in

memory maintenance because it could be constitutively

active with the lack of a regulatory subunit [

38, 50].

Some evidence suggests that PKM

is crucial to maintain

synaptic potentiation not only in memory, but also in

chronic pain [

95] and drug addiction [

96]. In addition,

PKM

is also involved in epigenetic mechanisms, which

might be related to long-term regulation of gene expres-

sion for memory maintenance [

66]. Therefore, it is ne-

cessary to reconcile these data with an integrated view.

Interestingly, memory can also be abolished when a

constitutively active form of CaMKII is overexpressed

[

9], whereas an enhanced memory has been observed

following PKM

overexpression [

81]. Considering the

role of CaMKII, which is also important in memory in-

duction, this memory deficit could be a result of indis-

criminate spine maturation by excessive CaMKII

activity, which might destroy patterns that encode the

memory [

9]. In contrast, PKM

, which is known to be

insufficient to induce spine maturation, may be activated

by another mechanism (such as phosphoinositide-

dependent protein kinase-1) to maintain memory [

97].

Thus, PKM

might not be able to cause indiscriminate

spine maturation just by its overexpression, but it could

possibly induce over-maturation by functioning to a

greater extent in engram spines where maturation has

already occurred. Therefore, the molecular cascade in-

cluding CaMKII and PKM

should be further investi-

gated to better understand the detailed molecular

mechanisms of memory maintenance.

John Lisman

’

s contributions to science, and especially

his contributions to the molecular basis of memory were

original, wide-ranging and important. Future generations

must answer the remaining questions to better under-

stand the biological foundations of memory.

Commentary on

“

Criteria for identifying the mo-

lecular basis of the engram (CaMKII, PKM

ζ

)

”

Contributed by Mary B. Kennedy.

John Lisman was a devoted and serious neuroscientist

who pursued his chosen problems and ideas with pas-

sion. Among these was a question which he formulated

as: what mechanisms underlie the formation of

“

the en-

gram?

”

The engram is an abstraction that means slightly

different things to different people. Some consider it to

be the more or less permanent neural circuits that

underlie specific episodic memories. To John, it was

“

the

molecular changes by which a memory is stored in the

brain.

”

Of course, one can easily assume that the mo-

lecular changes he refers to are those that form and

maintain the more or less permanent neural circuits that

underlie specific episodic memories.

My own career has been devoted to understanding

what I refer to as the biochemical mechanisms that

regulate synaptic strength in excitatory synapses and

thus allow synapses to adapt their signaling patterns to

environmental input, and to store memories. The slight

differences in emphasis between my conception of syn-

aptic regulation and John

’

s statements about

“

the en-

gram

”

go a long way toward explaining why he and I so

often disagreed profoundly about the interpretation of

various experiments and about the experimental way

forward to understand synaptic regulation and memory.

When I was an assistant professor, I and my students

set out to use the protein chemistry methods that I still

revere, to discover and understand the function of regu-

latory proteins in the postsynaptic density. My first stu-

dents and I began by purifying a calmodulin-dependent

protein kinase (later named CaMKII) that, as a postdoc

in the Greengard laboratory, I had first measured in

brain homogenates and identified as located in a particu-

lar fraction of proteins eluted from a DEAE-cellulose

column. We quickly realized that it was highly abundant

in brain, and that it appeared to be a quantitatively

significant component of the postsynaptic density

[

–

100]. While studying the enzymatic properties of

the purified protein, my student, Steve Miller, noticed

that the kinase activity seemed to retain some phos-

phorylating activity in the absence of calcium and cal-

modulin after it had phosphorylated itself in the

presence of these two activators. During a visit to my

lab, John pointed out that this behavior might fit a

theoretical model he had proposed regarding a way for

an activated state to outlast turnover of individual

molecules and we mentioned this in our paper [

4, 20].

The notion of a calcium-dependent switch excited

many neuroscientists and John began assuring me that

Bear

et al. Molecular Brain

(2018) 11:76

Page 7 of 12

we might have discovered

“

the memory molecule.

”

This

notion, and the reaction of some neuroscientists to it,

began to disturb me. My conception of regulatory bio-

chemistry did not and does not include the notion that a

single molecule would constitute

“

memory.

”

I had, and

have, great respect for the complexity of molecular evo-

lution. I believe that subcellular

“

states,

”

especially

powerful synaptic connections that are important for

brain function, must be generated and maintained by in-

tricately controlled and robust mechanisms. Even the

pathways of bacterial metabolism are filled with feed-

back regulation and alternative pathways. How much

more complexity must have evolved to accurately regu-

late a process as subtle and important as mammalian

episodic memory and insure its robustness? I resisted

the idea that CaMKII was

“

the memory molecule

”

for

this reason. There is no

“

memory molecule.

”

Synapses

are regulated by the mutual interactions of a whole net-

work of proteins. It was clear then and is even clearer

now, that CaMKII is a critical component of synaptic

regulation because it is abundant in the brain and post-

synaptic density, it responds to calcium coming through

NMDA receptors, and retains its activation for a short

period that is delimited by the opposing activity of pro-

tein phosphatases. We now know that CaMKII feeds

back to AMPA and NMDA receptors by phosphorylat-

ing them, and it can initiate changes in many down-

stream regulatory enzymes including nitric oxide

synthase, synGAP, and regulators of the actin cytoskel-

eton. But it is also true that calcium and calmodulin in

the spine directly regulate a large number of other pro-

cesses, including several that depend on

cAMP-dependent protein kinases, and various forms of

protein kinase C. Thus, I have always believed that an

important way forward to understand synaptic regula-

tion is to study in vitro the mutual interactions among

the regulatory molecules in the postsynaptic density and

spine, test whether and when those interactions are im-

portant for synaptic function, and learn the quantitative

kinetic parameters that govern these interactions so that

we can gradually account completely for observed func-

tional changes by modeling the dynamic action of bio-

chemical regulatory loops and pathways (for a beginning

effort see [

101]).

John took a very different approach. His earliest train-

ing was in physics, and his primary technical expertise

was in electrophysiology. He worked by using logic to

create abstractions like

“

the engram

”

,or

“

induction, ex-

pression, and maintenance;

”

then trying to devise ways

that he and others could use pharmacology, electro-

physiology and sometimes genetics to deduce molecular

mechanisms. To my mind, these abstractions involve

myriad assumptions that are often unfounded and tend

to obscure the subtleties of actual molecular

mechanisms. Each of these

“

phases

”

of LTP that can be

measured with an electrode encompasses many under-

lying molecular mechanisms. I think that a defining dis-

agreement between John and me, and indeed between

me and a host of synaptic electrophysiologists, is that

one cannot tease out and understand the molecular

mechanisms of synaptic plasticity with electrodes,

pharmacology, and genetics alone. Each of these

methods is useful, but without an understanding of the

subtleties of biochemical interactions of proteins in a

regulatory network, the picture will always be incom-

plete, and may often be badly oversimplified. John was

surely not alone in his conviction that he could reason

through fundamental molecular mechanisms of learning

with logic and electrophysiological experiments in intact

tissues and animals. This thinking still permeates the

NIH review panels at every level and has damaged fund-

ing of in vitro biochemical studies. In turn, this situation

has damaged the ability of basic neuroscience to provide

the knowledge that pharmaceutical companies need to

identify possible therapeutics for Alzheimer

’

s disease,

mental illnesses, and autism spectrum disorders.

It is evident that both CaMKII and PKM

play crucial

roles in the mechanisms of synaptic plasticity and in

maintaining synaptic strength. There is no reason to see

their importance as mutually exclusive. A more useful

way forward will be to search out the crucial

“

substrate

proteins

”

that these two fascinating enzymes regulate. It

is even possible that CaMKII plays an important struc-

tural role in the postsynapse. In vitro biochemical work

will be needed to fully understand the interlocking func-

tions of the synaptic regulatory networks.

I am inclined to agree with a comment by Richard

Morris [

102] in a discussion of a recent paper from the

Sacktor lab on PKM

[50]. Morris reminds us that a

“

molecule implicated in memory retention really does

(not) need to be sustained throughout the lifetime of a

memory. An alternative possibility is that it may trigger

structural changes that are, in turn, mediated by other

molecules (such as actin): thus, with this job done, our

memory molecule can gracefully depart the scene to play

upon another stage. Such structural changes could then

be faithfully recycled during routine protein turnover,

with these proteins being unaware, so to speak, that they

are sustaining a memory.

”

When a synapse is made

more powerful by the addition of more AMPA recep-

tors, more active zones, a larger spine and postsynaptic

density, etc., the collective energy embodied in the myr-

iad protein affinities imparts great structural stability.

Removal or dislodging of any one or two or three pro-

teins will not disrupt the structural stability imparted by

all the other associations. Thus, normal cellular repair

and maintenance processes could continue to maintain

the larger synapse once it is established. These cellular

Bear

et al. Molecular Brain

(2018) 11:76

Page 8 of 12

processes are not perfect; yet we all know that memory

isn

’

t perfect either and memories are often lost or al-

tered as we age.

John Lisman made many contributions to neurosci-

ence. His passion and his provocative comments and

ideas will be missed.

The non-erasable memory of John Lisman.

Contributed by Richard G M Morris.

John Lisman was a man who was infectiously curious,

read voraciously, thought deeply and was genuinely in-

terested in what others had to say. He was a listener. He

was both passionate about science and generous to

others around him who were pursuing their scientific in-

terests; indeed the recent appreciations of him by friends

and colleagues [

103 ] are very moving. He was a man

with diverse scientific interests ranging from phototrans-

duction, the possibility that autophosphorylation of

CaMKII could be a molecular switch for memory, neural

oscillations and working memory storage and, later in

his career, the puzzle of schizophrenia. This diversity led

him into contact with all manner of different people in

neuroscience that don

’

t often connect. John Lisman was

a man who loved science and for whom science was -

beyond family - his life.

In his Paths to Discovery contribution to the 3rd edi-

tion of the Bear, Connors and Paradiso textbook of

neuroscience [

104], John describes how he came up with

the idea of an autophophorylation as a molecular switch

for memory. Inevitably, it was during a walk along a

beach - the best ideas often come near the sea:

“

My eureka moment

...

. was that a group of

autophosphorylating kinase molecules localized at a

synapse could make a stable switch. During LTP

induction, these molecules would become

phosphorylated, and this would make them active. If a

kinase molecule was dephosphorylated or replaced in

the course of protein turnover, it could be

phosphorylated by other members of the group. The

switch could stay on, perhaps indefinitely, and this

showed how unstable molecules could produce stable

information storage.

”

In this final paper on which we are commenting [

10],

John asserts that the definitive evidence that CaMKII au-

tophosphorylation could be a mechanism of memory

comes not from correlational data showing that this

process occurs in a lasting manner in response to neural

activity that may be associated with memory formation

in synaptic or behavioural models - for such data (im-

portant as they are) are merely suggestive of a memory

storage mechanism. Rather, it is from a particular type

of experiment from which a causal inference can be

drawn. This is his

“

erasure

”

test. The argument is

attractively simple: if CaMKII autophosphorylation is a

mechanism of memory storage, memory must disappear

if you turn it off.

The argument has depth too for, in Figure 1, he de-

fines the critical distinction between an induction

process, a maintenance process and an expression

process with respect to the making of a lasting memory

trace, and at several points in his article asserts (cor-

rectly in my view) that evidence from expression experi-

ments is not necessarily relevant to induction or

maintenance processes. We must keep these distinctions

separate. However, while not in any way demurring from

this argument, I am not convinced it is sufficient - for a

‘

top-down

’

perspective on memory would distin-

guish the dissociable processes of memory-encoding,

storage, consolidation and retrieval. Re-activation and

retrieval did not seem to be on John

’

s radar and this

omission from his thinking may be critical. From a psy-

chological perspective, the supposition is that memory

traces in long-term memory can be active or can be-

come dormant (i.e.

‘

inactive

’

) over weeks, months or

even years. And then - magically as it were - be reacti-

vated and retrieved. Does the autophosphorylation idea

require that this biochemical process is going on dis-

creetly at dormant synapse after dormant synapse across

these time-periods? Here I have a certain suspension of

disbelief for a pre- and post-synaptic structural solution

at the synapse feels (to me) much more economical. A

metaphor may be helpful here to get across the general

idea: Consider the task of a supermarket store-manager

trying to maximise his or her sales. They need to devote

sufficient length and space in an aisle to the sale of dif-

ferent items, and not allow an item that doesn

’

t sell very

well to occupy too much space. Given this - the range of

(say) coffee items might take up a metre or two, but

should not occupy more. The memory in the system

here is the length of the shelf devoted to coffee, itself a

reflection of manager

’

s experience of how quickly they

are taken off the shelves by the shop

’

s customers. But

the items - the coffee jars - are temporary and unstable.

In this way, stability and instability co-habit happily,

even though there is an analogy to the protein turnover

that rightly concerned John in his thinking about CaM-

KII. His ideas about

“

slots

”

in the PSD was very much in

the same spirit.

But this need not mean Lisman

’

s autophosphorylation

idea is wrong. To the contrary, if we build into his

framework the important distinction between cellular

(i.e. immediate) and systems consolidation (longer-term),

it seems entirely possible that sustained CaMKII auto-

phosphorylation during the initial cellular phase may be

critical for memory as he thought. What is important

about this division into two qualititatively distinct con-

solidation systems, from the perspective of evolution, is

Bear

et al. Molecular Brain

(2018) 11:76

Page 9 of 12

that a cellular consolidation process takes things

‘

off--

line

’

from the perspective of cell firing (the relevant

brain cells can get on with other new learning). How-

ever, this biochemical process serves, for a while, as an

on-line

‘

long-list

’

of what is destined to be retained in-

definitely in the brain

’

s lasting long-term memory sys-

tems. That is, post-processing, a body of information is

being retained for potentially quite long periods (hours,

certainly; days, maybe) even though not all of it may be

retained by the longer-term, and probably structural,

systems consolidation process.

The data presented in Figure 4 and 5 show Lisman

’

favoured

‘

saturation

’

and

‘

erasure

’

tests for establishing a

role for CaMKII autophosphorylation in memory storage.

In the former test, animals learn to avoid a specific place

in Andre Fenton

’

s ingenious place avoidance task (see

path data for Trial 12). At that point, an HSV vector ex-

pressing activated CaMKII is infused into the hippocam-

pus with the expected mechanistic consequence being a

disruption of the spatial pattern of synaptic weights. Mem-

ory is then poorer (but not in the control condition) on

Trial 13 (6 days later). This is more of a memory retrieval

test than any other, and I note that the inspiration for this

experiment was Vegard Brun

’

s similar study in which LTP

itself was saturated [

105 ]. The

‘

erasure

’

test is different. In

this case again, a memory is first acquired - either in a

model system such as LTP or in behaviour. The idea then

is to erase this memory using TatCN21 or HSV-K42 M

which interferes with the ability of CaMKII to bind effect-

ively with the NMDA receptor. When given after 4 days of

training, loss of memory was demonstrated in a test con-

ducted 10 days later after, apparently, the dominant nega-

tive construct was no longer present. This means that the

loss of memory is not a transient consequence of it not

being expressed well in the presence of the interfering

agent. The memory trace is really gone. But the encoding

system is not damaged and re-learning can occur. By any

standards, this is a powerful experimental design from

which strong inferences can indeed be drawn.

But where does this leave us? I share John

’

s assertion

that

“

the molecular basis of memory storage is one of

the most fundamental questions in cellular neurosci-

ence

”

. (page 9). I am less sure that we are yet in a situ-

ation where we can all go home and say

“

problem

solved

”

. Part of my reticence is that I feel uncertain

about any theory that, implicitly or explicitly, implies

that a specific molecule is the be-all and end-all of

everything. Even in the case of DNA, which comes

pretty close to meeting that sense of

“

problem solved

”

with respect to the issue of replication, there were still

all manner of issues that have sensibly occupied molecu-

lar biologists for years since. With respect to neurosci-

ence, I am not sure we are yet in a position to think that

the autophosphorylation of CaMKII is quite in the DNA

league, but also doubt that we can think of the problem

of memory in merely a cellular way. I have already noted

what feels to me to be the weakness of a theory that ap-

parently requires a biochemical autophosphorylation

process to be continuing indefinitely.

Beyond these comments, there are larger issues to do

with why some memories are remembered and not others.

My own reflections on this issue relate to

synaptic-tagging-and-capture [

106

–

108 ] and to how an

initial cellular consolidation process that provides the first

level of information

“

selection

”

interfaces with a systems

consolidation process that serves to update existing know-

ledge structures such as schemas [

109 , 110 ]. One joy of

my career was discussing these matters with John Lisman

by email, SKYPE and in person, especially in relation to

his theory with Tony Grace about the impact that novelty

might have on selective retention within the hippocampus

mediated via the neuromodulatory transmitter dopamine

[

111 ]. My laboratory was inspired by this idea and con-

ducted a number of experiments on the issue, some of

which went on to influence John

’

s own thinking. More re-

cently, our suggestion that he, and Julie Frey earlier were

right about dopamine but might not have been right about

the source of the dopamine, was as intriguing to him as to

ourselves. He made a point of attending our poster pre-

sentations at the Society for Neuroscience (2015) showing

behavioural, electrophysiological and optogenetic data

pointing to the locus coeruleus rather than the ventral teg-

mental area as possible cells of origin. These data were

eventually published a year later [

112 ], followed shortly by

data from Eric Kandel

’

s lab [

113 ] indicating the presence

of dopamine release from activated LC axons. The jury is

still out on this issue, but the suggestion is at least out

there. Adrian Duszkiewicz and Tomonori Takeuchi both

got a buzz from having John go through their posters so

constructively with them - and I listened with interest. To

the end, John cared about young scientists. Let us never

erase our memory of such a generous, graceful and gifted

scientist.

Authors

’

contribution

All authors read and approved the final manuscript.

Competing interests

The authors declare that they have no competing interests.

Publisher

’

sNote

Springer Nature remains neutral with regard to jurisdictional claims in

published maps and institutional affiliations.

Received: 3 August 2018 Accepted: 24 October 2018

References

1. Lisman JE. Bursts as a unit of neural information: making unreliable

synapses reliable. Trends Neurosci. 1997;20(1):38

–

43.

2. Lisman JE, Idiart MA. Storage of 7 +/

−

2 short-term memories in oscillatory

subcycles. Science. 1995;267(5203):1512

–

Bear

et al. Molecular Brain

(2018) 11:76

Page 10 of 12

3. Otmakhova N, Duzel E, Deutch AY, Lisman J. The hippocampal-VTA loop:

the role of novelty in controlling the entry of information into long-term

memory. In: Intrinsically Motivated Learning in Natural and Artifical

Systems; 2012. p. 235

–

54.

4. Lisman LE. A mechanism for memory storage insensitive to molecular

turnover: a bistable autophosphorylating kinase. Proc Natl Acad Sci U S A.

1985;82(9):3055

–

5. Lisman LE, Goldring MA. Feasibility of long-term storage of graded

information by the Ca

/calmodulin-dependent protein kinase molecules of

the postsynaptic density. Proc Natl Acad Sci U S A. 1988;85(14):5320

–

6. Lisman J. The CaM kinase II hypothesis for the storage of synaptic memory.

Trends Neurosci. 1994;17(10):406

–

12.

7. Lisman J, Schulman H, Cline H. The molecular basis of CaMKII function in

synaptic and behavioral memory. Nat Rev Neurosci. 2002;3(3):175

–

90.

8. Lisman J, Yasuda R, Raghavachari S. Mechanisms of CaMKII action in long-

term potentiation. Nat Rev Neurosci. 2012;13(3):169

–

82.

9. Rossetti T, Banerjee S, Kim C, Leubner M, Lamar C, Gupta P, Lee B, Neve R,

Lisman J. Memory erasure experiments indicate a critical role of CaMKII in

memory storage. Neuron. 2017;96(1):207

–

16.

10. Lisman J. Criteria for identifying the molecular basis of the engram (CaMKII,

PKM

). Mol Brain. 2017;10(1):55.

11. Lisman J, Goldring M. Evaluation of a model of long-term memory based

on the properties of the Ca2+/calmodulin-dependent protein kinase. J

Physiol Paris. 1988;83(3):187

–

97.

12. Lisman J. A mechanism for the Hebb and the anti-Hebb processes

underlying learning and memory. Proc Natl Acad Sci U S A. 1989;86(23):

9574

–

13. Hebb DO. The Organization of Behavior: a neuropsychological theory. New

York: Wiley; 1949.

14. Bliss TV, Gardner-Medwin AR. Long-lasting potentiation of synaptic

transmission in the dentate area of the unanaestetized rabbit following

stimulation of the perforant path. J Physiol. 1973;232(2):357

–

74.

15. Bliss TV, Lomo T. Long-lasting potentiation of synaptic transmission in the

dentate area of the anaesthetized rabbit following stimulation of the

perforant path. J Physiol. 1973;232(2):331

–

56.

16. Andersen P, et al. Specific long-lasting potentiation of synaptic transmission

in hippocampal slices. Nature. 1977;266(5604):736

–

17. Abraham WC. How long will long-term potentiation last? Philos Trans R Soc

Lond Ser B Biol Sci. 2003;358(1432):735

–

44.

18. Bradshaw KD, Emptage NJ, Bliss TV. A role for dendritic protein synthesis in

hippocampal late LTP. Eur J Neurosci. 2003;18(11):3150

–

19. Crick F. Memory and molecular turnover. Nature. 1984;312(5990):101.

20. Miller SG, Kennedy MB. Regulation of brain type II Ca2+/calmodulin-

dependent protein kinase by autophosphorylation: a Ca2+

−

triggered

molecular switch. Cell. 1986;44(6):861

–

70.

21. Lee HK, et al. Regulation of distinct AMPA receptor phosphorylation sites

during bidirectional synaptic plasticity. Nature. 2000;405(6789):955

–

22. Tomita S, et al. Bidirectional synaptic plasticity regulated by phosphorylation

of stargazin-like TARPs. Neuron. 2005;45(2):269

–

77.

23. Dunwiddie TV, Lynch G. The relationship between extracellular calcium

concentrations and the induction of hippocampal long-term potentiation.

Brain Res. 1979;169(1):103

–

10.

24. Collingridge GL, Kehl SJ, McLennan H. Excitatory amino acids in synaptic

transmission in the Schaffer collateral-commissural pathway of the rat

hippocampus. J Physiol. 1983;334:33

–

46.

25. Malenka RC, et al. Postsynaptic calcium is sufficient for potentiation of

hippocampal synaptic transmission. Science. 1988;242(4875):81

–

26. Dudek SM, Bear MF. Homosynaptic long-term depression in area CA1 of

hippocampus and effects of N-methyl-D-aspartate receptor blockade. Proc

Natl Acad Sci U S A. 1992;89(10):4363

–

27. Mulkey RM, Malenka RC. Mechanisms underlying induction of

homosynaptic long-term depression in area CA1 of the hippocampus.

Neuron. 1992;9(5):967

–

75.

28. Mulkey RM, Herron CE, Malenka RC. An essential role for protein

phosphatases in hippocampal long-term depression. Science. 1993;

261(5124):1051

–

29. Mulkey RM, et al. Involvement of a calcineurin/inhibitor-1 phosphatase cascade

in hippocampal long-term depression. Nature. 1994;369(6480):486

–

30. Zeng H, et al. Forebrain-specific calcineurin knockout selectively impairs

bidirectional synaptic plasticity and working/episodic-like memory. Cell.

2001;107(5):617

–

29.

31. Giese KP, et al. Autophosphorylation at Thr286 of the alpha calcium-

calmodulin kinase II in LTP and learning. Science. 1998;279(5352):870

–

32. Malinow R, Schulman H, Tsien RW. Inhibition of postsynaptic PKC or CaMKII

blocks induction but not expression of LTP. Science. 1989;245(4920):862

–

33. Buard I, et al. CaMKII

“

autonomy

”

is required for initiating but not for

maintaining neuronal long-term information storage. J Neurosci. 2010;

30(24):8214

–

20.

34. Murakoshi H, et al. Kinetics of endogenous CaMKII required for synaptic

plasticity revealed by Optogenetic kinase inhibitor. Neuron. 2017;94(3):690.

35. Lengyel I, et al. Autonomous activity of CaMKII is only transiently increased

following the induction of long-term potentiation in the rat hippocampus.

Eur J Neurosci. 2004;20(11):3063

–

72.

36. Lee SJ, et al. Activation of CaMKII in single dendritic spines during long-

term potentiation. Nature. 2009;458(7236):299

–

304.

37. Chang JY, et al. CaMKII autophosphorylation is necessary for optimal

integration of ca(2+) signals during LTP induction, but not maintenance.

Neuron. 2017;94(4):800

–

8 e4.

38. Sacktor TC. How does PKM

maintain long-term memory? Nat Rev Neurosci.

2011;12(1):9

–

15.

39. Westmark PR, et al. Pin1 and PKM

sequentially control dendritic protein

synthesis. Sci Signal. 2010;3(112):ra18.

40. Kelly MT, Crary JF, Sacktor TC. Regulation of protein kinase Mzeta synthesis by

multiple kinases in long-term potentiation. J Neurosci. 2007;27(13):3439

–

44.

41. Yao Y, et al. PKM zeta maintains late long-term potentiation by N-

ethylmaleimide-sensitive factor/GluR2-dependent trafficking of postsynaptic

AMPA receptors. J Neurosci. 2008;28(31):7820

–

42. Ling DS, Benardo LS, Sacktor TC. Protein kinase Mzeta enhances excitatory

synaptic transmission by increasing the number of active postsynaptic

AMPA receptors. Hippocampus. 2006;16(5):443

–

52.

43. Schuman EM. Synapse specificity and long-term information storage.

Neuron. 1997;18(3):339

–

42.

44. Steward O. mRNA localization in neurons: a multipurpose mechanism?

Neuron. 1997;18(1):9

–

12.

45. Pastalkova E, et al. Storage of spatial information by the maintenance

mechanism of LTP. Science. 2006;313(5790):1141

–

46. Serrano P, et al. PKM

maintains spatial, instrumental, and classically

conditioned long-term memories. PLoS Biol. 2008;6(12):2698

–

706.

47. Hsieh C, et al. Persistent increased PKM

in long-term and remote spatial

memory. Neurobiol Learn Mem. 2017;138:135

–

44.

48. Lee AM, et al. Prkcz null mice show normal learning and memory. Nature.

2013;493(7432):416

–

49. Volk LJ, et al. PKM-zeta is not required for hippocampal synaptic plasticity,

learning and memory. Nature. 2013;493(7432):420

–

50. Tsokas P, Hsieh C, Yao Y, Lesburgueres E, Wallace EJC, Tcherepanov A,

Jothianandan D, Hartley BR, Pan L, Rivard B et al. Compensation for PKMzeta

in long-term potentiation and spatial long-term memory in mutant mice.

Elife 2016;5:e14846.

51. Hinds HL, Tonegawa S, Malinow R. CA1 long-term potentiation is

diminished but present in hippocampal slices from alpha-CaMKII mutant

mice. Learn Mem. 1998;5(4

–

5):344

–

54.

52. Silva AJ, et al. Deficient hippocampal long-term potentiation in alpha-

calcium-calmodulin kinase II mutant mice. Science. 1992;257(5067):201

–

53. Silva AJ, et al. Impaired spatial learning in alpha-calcium-calmodulin kinase II

mutant mice. Science. 1992;257(5067):206

–

11.

54. Elgersma Y, et al. Inhibitory autophosphorylation of CaMKII controls PSD

association, plasticity, and learning. Neuron. 2002;36(3):493

–

505.

55. Cooke SF, et al. Autophosphorylation of alphaCaMKII is not a general

requirement for NMDA receptor-dependent LTP in the adult mouse. J

Physiol. 2006;574(Pt 3):805

–

18.

56. Irvine EE, Vernon J, Giese KP. AlphaCaMKII autophosphorylation

contributes to rapid learning but is not necessary for memory. Nat

Neurosci. 2005;8(4):411

–

57. Lledo PM, et al. Calcium/calmodulin-dependent kinase II and long-term

potentiation enhance synaptic transmission by the same mechanism. Proc

Natl Acad Sci U S A. 1995;92(24):11175

–

58. Schuette SR, et al. Overexpression of protein kinase Mzeta in the

Hippocampus enhances long-term potentiation and long-term contextual

but not cued fear memory in rats. J Neurosci. 2016;36(15):4313

–

24.

59. Bejar R, et al. Transgenic calmodulin-dependent protein kinase II activation:

dose-dependent effects on synaptic plasticity, learning, and memory. J

Neurosci. 2002;22(13):5719

–

26.

Bear

et al. Molecular Brain

(2018) 11:76

Page 11 of 12