of the distance between dendritic locations, with
only minimal interactions between the two hemi-
trees [fig. S14D; see also Kerlinet al.( 11 )].
Discussion
Overall, our results reveal a subclass of thick-
tufted layer 5 PTNs in M1 (type 1) that per-
forms parallel independent representations
of motor information within its tuft den-
drites. The degree of tuft compartmentaliza-
tion was primarily dependent on the nexus
and tuft tree morphology and required the
NMDA spiking mechanism. In these neurons,
motor information is integrated and ampli-
fied via NMDA spikes within adjustable seg-
ments of the tuft, ranging from small tertiary
and quarterly sister branches, independent R/L
hemi-trees, and up to a single global computa-
tional compartment in the minority of events.
Our modeling suggests that calcium spikes play
only a minimal role in tuft compartmental-
ization and local amplification. Instead, the
initiation of calcium spikes in the distal apical
trunk and proximal nexus branches probably
serves to further amplify the tuft computa-
tional products ( 46 , 47 ).
Type 2 PTNs also amplify behaviorally
relevant motor information within their tuft
but in a more global manner, mostly func-
tioning as a single computational compart-
ment ( 8 – 11 ).
The type 1 subclass constitutes a sizable
fraction of thick-tufted layer 5 PTNs (~40%), in
line with previous reports ( 24 , 26 ). Although
we based our classification solely on the apical
dendritic morphology, our physiological find-
ings supported this classification. Studies
that examined both dendritic morphology and
projection targets suggest that type 1 PTNs
may preferentially project to the medulla
( 23 , 27 , 32 ), which is consistent with our
medulla and spinal cord retrograde-labeled
PTNs. Further studies are required to exam-
ine the projection pattern and molecular
markers of type 1 and 2 PTNs ( 23 , 32 ).
Our results reconcile the differences in the
findings of prior in vitro and modeling studies,
which predicted the capacity of dendrites to
compartmentalize information ( 4 , 17 , 22 , 48 , 49 )
and recent recordings from behaving mice,
which show that tuft dendrites function primar-
ily as a single global amplification unit ( 8 – 11 ).
Several past studies also observed infrequent
local spikes that were limited to small, non-
overlapping dendritic segments in tuft den-
drites ( 9 – 11 ). Yet, this highly localized spiking
activity, reminiscent of our cluster 1 events,
cannot serve for efficiently communicating
tuft computations to the soma. It may be used
for local plasticity instead ( 6 , 21 ). A study ( 6 )
that recorded tuft dendrites of M1 layer 5
PTNs reported spatially isolated dendritic
spikes in almost all pairs of sibling distal tuft
branches (95%). The results of this study dif-
fer from ours, as we observed selective acti-
vation of sibling terminal tuft branches during
both motor tasks infrequently (<5%). These
results are especially surprising in the case of
type 2 layer 5 PTNs, which should have also
been observed in that study. The discrepancies
between our findings and those of ( 6 ) were
probably related technical issues such as the
low acquisition rate and the lack of adequate
sparse labeling in ( 6 ).
The R/L hemi-tree compartmentalization
is perhaps one of the most distinctive and
intriguing properties of type 1 PTNs, which
was not anticipated from previous work. This
hemi-tree tuft compartmentalization enables
PTNs to represent different sets of informa-
tion in parallel, with each hemi-tree routing
information to the soma independently acting
as“a neuron within a neuron.”It is conceivable
that in larger and more complex primate and
human PTNs, this property would have an even
larger impact, with a greater number of almost
isolated integrative zones in the tuft ( 50 , 51 ).
On the basis of our data, we propose a new
integration and representation scheme of
motor variables in tuft dendrites of M1 layer
5 PTNs. Motor variables are not represented
in fully compartmentalized, small, nonover-
lapping dendritic segments as previously
reported ( 6 ). Instead, a given motor behavior
or a sequence is represented by the activation
of a specific combination of distal tuft seg-
ments, which are mutually coamplified via
NMDA spikes to form spatial dendritic am-
plification maps for different motor behaviors.
In this framework, the tuft tree of type 1 layer
5 PTNs is capable of dynamic combinatorial
representation of a large number of motor
variables and sequences within the same
dendritic tuft branches ( 2 , 52 , 53 ).REFERENCES AND NOTES- T. Branco, M. Häusser,Curr. Opin. Neurobiol. 20 , 494– 502
(2010). - X. E. Wu, B. W. Mel,Neuron 62 , 31–41 (2009).
- G. J. Stuart, N. Spruston,Nat. Neurosci. 18 , 1713– 1721
(2015). - G. Major, M. E. Larkum, J. Schiller,Annu. Rev. Neurosci. 36 ,
1 – 24 (2013). - S. D. Antic, W. L. Zhou, A. R. Moore, S. M. Short, K. D. Ikonomu,
J. Neurosci. Res. 88 , 2991–3001 (2010). - J. Cichon, W. B. Gan,Nature 520 , 180–185 (2015).
- L. M. Palmeret al.,Nat. Neurosci. 17 , 383–390 (2014).
- L. Beaulieu-Laroche, E. H. S. Toloza, N. J. Brown, M. T. Harnett,
Neuron 103 , 235–241.e4 (2019). - V. Francioni, Z. Padamsey, N. L. Rochefort,eLife 8 , e49145
(2019). - J. Voigts, M. T. Harnett,Neuron 105 , 237–245.e4 (2020).
- A. Kerlinet al.,eLife 8 , e46966 (2019).
- N. L. Xuet al.,Nature 492 , 247–251 (2012).
- C. O. Lacefield, E. A. Pnevmatikakis, L. Paninski, R. M. Bruno,
Cell Rep. 26 , 2000–2008.e2 (2019). - B. B. Ujfalussy, J. K. Makara, M. Lengyel, T. Branco,Neuron
100 , 579–592.e5 (2018). - R. Naud, B. Bathellier, W. Gerstner,Front. Comput. Neurosci. 8 ,
90 (2014). - P. Poirazi, T. Brannon, B. W. Mel,Neuron 37 , 989– 999
(2003). - D. Beniaguev, I. Segev, M. London,Neuron 109 , 2727–2739.e3
(2021).
18. A. Payeur, J. C. Béïque, R. Naud,Curr. Opin. Neurobiol. 58 ,
78 – 85 (2019).
19. G. Kastellakis, P. Poirazi,Front. Mol. Neurosci. 12 , 300
(2019).
20. R. Legenstein, W. Maass,J. Neurosci. 31 , 10787– 10802
(2011).
21. J. Bono, C. Clopath,Nat. Commun. 8 , 706 (2017).
22. B. W. Mel, inAdvances in Neural Information Processing
Systems, J. Moody, S. Hanson, R. Lippmann, Eds.
(Morgan Kaufmann, 1992), vol. 4, pp. 35–42.
23. M. N. Economoet al.,Nature 563 , 79–84 (2018).
24. H. Markramet al.,Cell 163 , 456–492 (2015).
25. W. J. Gao, Z. H. Zheng,J. Comp. Neurol. 476 , 174– 185
(2004).
26. Y. Wang, M. Ye, X. Kuang, Y. Li, S. Hu,IBRO Rep. 5 , 74– 90
(2018).
27. S. Jianget al.,Sci. Rep. 10 , 7916 (2020).
28. G. M. Shepherd,Nat. Rev. Neurosci. 14 , 278–291 (2013).
29. A. Grohet al.,Cereb. Cortex 20 , 826–836 (2010).
30. E. J. Kim, A. L. Juavinett, E. M. Kyubwa, M. W. Jacobs,
E. M. Callaway,Neuron 88 , 1253–1267 (2015).
31. A. M. Hattox, S. B. Nelson,J. Neurophysiol. 98 , 3330– 3340
(2007).
32. B. Tasicet al.,Nature 563 , 72–78 (2018).
33. L. N. Fletcher, S. R. Williams,Neuron 101 , 76–90.e4 (2019).
34. S. Levyet al.,Neuron 107 , 954–971.e9 (2020).
35. F. Fuhrmannet al.,Neuron 86 , 1253–1264 (2015).
36. T. W. Chenet al.,Nature 499 , 295–300 (2013).
37. T. Deneuxet al.,Nat. Commun. 7 , 12190 (2016).
38. N. Mantel,Cancer Res. 27 , 209–220 (1967).
39. B. Manly,The Statistics of Natural Selection(Chapman & Hall,
1985).
40. P. W. Mielke,Biometrics 34 , 277 (1978).
41. B. Engelhardet al.,Nature 570 , 509–513 (2019).
42. P. Ramkumar, B. Dekleva, S. Cooler, L. Miller, K. Kording,
PLOS ONE 11 , e0160851 (2016).
43. B. A. Sauerbreiet al.,Nature 577 , 386–391 (2020).
44. A. Nashef, O. Cohen, R. Harel, Z. Israel, Y. Prut,Cell Rep. 27 ,
2608 – 2619.e4 (2019).
45. M. Lavzin, S. Rapoport, A. Polsky, L. Garion, J. Schiller,Nature
490 , 397–401 (2012).
46. M. E. Larkum, T. Nevian, M. Sandler, A. Polsky, J. Schiller,
Science 325 , 756–760 (2009).
47. M. Larkum,Trends Neurosci. 36 , 141–151 (2013).
48. H. Jia, N. L. Rochefort, X. Chen, A. Konnerth,Nature 464 ,
1307 – 1312 (2010).
49. M. E. Sheffield, D. A. Dombeck,Nature 517 , 200–204 (2015).
50. L. Beaulieu-Larocheet al.,Cell 175 , 643–651.e14 (2018).
51. H. Mohanet al.,Cereb. Cortex 25 , 4839–4853 (2015).
52. R. Urbanczik, W. Senn,Neuron 81 , 521–528 (2014).
53. J. Hawkins, S. Ahmad,Front. Neural Circuits 10 , 23 (2016).
ACKNOWLEDGMENTS
We thank S. Marom and B. Engelhard for helpful comments on the
manuscript. We also thank S. Schwartz for advice in statistical
tests and S. Gafniel for movie editing.Funding:This study was
partially supported by the Israeli Science Foundation (J.S.
and Y.S.), Prince funds (J.S. and Y.S.), Rappaport Foundation
(J.S. and Y.S.), and Zuckerman Postdoctoral Fellowship (N.C.).
Author contributions:Designed research: J.S., N.C., Y.S.
Performed experiments: Y.O., N.C., S.A., M.A. Analyzed data:
S.A., Y.O., J.S., H.B., O.B., Y.S. Writing–original draft: J.S.
Writing–review and editing: J.S., Y.O., S.A., A.P.-P., Y.S.
Performed simulations: A.P.-P., S.A., J.S.Competing interests:
The authors declare that they have no competing interests.Data
and materials availability:All data are available in the manuscript
or the supplementary materials. Computer code can be found in:
https://github.com/SchillersLab/Dynamic-compartmental-
computations-in-tuft-dendrites-.SUPPLEMENTARY MATERIALS
science.org/doi/10.1126/science.abn1421
Materials and Methods
Figs. S1 to S14
References ( 54 – 60 )
MDAR Reproducibility Checklist
Movies S1 to S107 November 2021; accepted 11 March 2022
10.1126/science.abn1421SCIENCEscience.org 15 APRIL 2022•VOL 376 ISSUE 6590 275
RESEARCH | RESEARCH ARTICLE