Supplementary Information Supplementary Methods 1: Gene Editing in Zebrafish to Produce the Psen1 K97fs Mutation

Supplementary Information Supplementary Methods 1: Gene editing in zebrafish to produce the psen1 K97fs mutation.

A1. PRESENILIN and PS2V protein structure. The lipid bilayer is shaded grey and the transmembrane domains are indicated by black ovals. The site of the Homo sapiens (Hs) PSEN2 K115fs and Danio rerio (Dr) psen1 K97fs mutation is indicated in red.

A2. A section of the Hs PSEN2 and Dr psen1 nucleotide alignment around the K115fs (K97fs) mutation site.

A3. Section of genomic Dr psen1 sequence around the GA target site with TALEN binding sites indicated and the single stranded oligonucleotide sequence below.

B1. Flow diagram for creation of the Psen1 K97fs heterozygous mutation model. Upper agarose gel image is of amplified genomic DNA from 24hpf uninjected (u) embryos, injected (i) embryos and water negative control (-). PCR reactions containing either wild type (wt) detecting or mutation (mut) detecting primers. Lower agarose gel image is of amplified genomic DNA from adult tailclips from a wild type (w), a K97fs heterozygous germline mutant (m) and water negative control (-). The amplified DNAs are from PCR reactions containing either wild type (wt) detecting or mutation (mut) detecting primers.

B2. Chromatographs of a section of genomic psen1 sequence around the K97fs deletion site for wild type DNA and for K97fs heterozygous mutant DNA. Red sequence is the altered DNA sequence after GA dinucleotide deletion.

B3. A K97fs heterozygous mutant (red stripes) was pair mated with a wild type (Tübingen) to produce a tank of fish that is ~1:1 mutant:non-mutant. At 6 months and 24 months of age, brains were removed from female K97fs heterozygous mutant adults and non-mutant adult siblings. Mutant and non-mutant brains were subjected to either RNAseq or LC-MS/MS procedures followed by downstream analysis of the data. 6 months 24 months 6 months 24 months

Wild type A and a Mutant Wild type Mutant b Wild type Mutant Wild type Mutant rnps1 si:dkey−19f4.2 2 her4.1 2 CR847895.3 mblac1 sox9b 1 1 nfyal si:dkey−147f3.4 barhl1a 0 bada 0 U2 mastl hey1 zgc:173570 −1 si:ch211−107e6.5 Genes shown in the −1 ZNF189 (1 of many) zgc:165551 RELATIVE EXPRESSION RELATIVE EXPRESSION −2 vamp5 −2 zgc:101100 Metazoa_SRP gb:eb993102 ptger1b stard9 CR847563.1 hist1h4l tesca A. 6-month-old mutant vs. 6-

nupr1 B .

FAILURE TO DOWNREGULATE krt15 lats1 myhb ttn.2 psen1 lfng amotl2a sord zgc:114130 sertad2b rerglb CABZ01087568.1 mknk2a gtf2a1l CU984600.2 gpr146 PAMR1 erf kdm6ba zgc:158463 tiparp HEPACAM (1 of many) rtkn2a dhx58 si:dkey−85k7.12 si:ch211−108c17.2 cabz01093075.1 myhb sdhda RNPS1 (1 of many) ) siblings at 6 and 24 months of age. si:ch211−189k9.2 ms4a17a.11 ssh1a +/+ pim3 FP074874.2 cmklr1 errfi1b CABZ01015815.2 slc38a4 si:ch73−158p21.3 lgals3bpb FAILURE TO UPREGULATE tagapb usp2b ankrd45 CU681842.1 usp2a adi1 ackr3a si:ch211−195b13.1 CR847563.1 irs2b mcee rbms1a

snx22 fold change > 0.5 for the following comparisons:

gpr63 2 btg2 fam107b cyp1c2 fosab serpinh1b slc25a25b ier2a chodl nfil3−5 CR855274.1 aftphb zgc:175088 nudt19 dnajb1b hoxb3a mpdu1b hoxa4a urp1 mettl1 hoxb5a tnfrsf11a hoxa5a si:ch73−380l10.2 hapln1b pkd1l2a ) zebrafish compared to wild type ( psen1 phox2bb arrdc2 hoxc1a hoxb8a K97fs/+ fosb INAPPROPRIATELY DOWNREGULATED hoxb5b hoxc5a foxg1b rgs5a mrc1b cmklr1 stk40 fam208b adrb2b hif1al tdh si:dkey−27j5.5 hsd11b2 nt5dc2 si:dkey−13p1.4 zgc:174944 pfkfb4b fkbp5 abcf2a eif4e1c nfkbiab COQ10A rnf144ab nr1d2a plxnc1 dpp7 csrnp1a tnfrsf11a rgcc pik3r3a mcl1b foxg1b klf9 si:dkey−13p1.4 rassf9 si:dkey−201c13.2 bgnb zgc:162730 nocta rps6kb1a zgc:122979 usp2a gadd45ba si:dkey−27j5.5 zgc:110340 slc16a9a si:dkey−242g16.2 nfkbiab npas4a B. 24-month-old mutant vs. 24-month-old wild type zebrafish brains. Gene expression patterns for the opposite age group are also shown in both INVERTED ddit4 zgc:122979 slc13a5b nfil3−6 tsc22d3 nr1d2a trib3 slc16a9a plekhg5a si:ch211−232b12.5 si:dkey−242g16.2 pgp DTX3 ddit4 pik3r1 sox11b cebpd ip6k2b cdkn1d her8a siah2l unm_sa1261 si:ch211−9d9.1 stk35 ube2al jdp2b nocta csrnp1b gadd45ba csrnp1a pdk2a ghsrb klf9 si:dkey−201c13.2 pik3r3a MMADHC (1 of many) fkbp5 slc16a12b Supplementary Figure 1 . Gene expression changes in the brains of mutant ( psen1 heatmaps are restricted to those with False Discovery Rate [FDR]-adjusted p -value < 0.05 and absolute log month-old wild type zebrafish brains. B (not in black boxes). Genes are hierarchically clustered with Euclidean distance measure. Biological interpretations proposed for the clusters of genes a b 6 months 24 months 6 months 24 months

Wild type Mutant Wild type Mutant Wild type Mutant Wild type Mutant

Nuclear distribution protein nudE−like 1−B (ndel1b) Phenylalanine−tRNA ligase alpha subunit (farsa) Adenylosuccinate synthetase isozyme 1 (adssl1) Serrate RNA effector molecule homolog (srrt) Keratin, type II cytoskeletal 8 (krt8) Poly(U)−binding−splicing factor PUF60−B (puf60b) Betaine−homocysteine S−methyltransferase 1 (bhmt) Hypoxia up−regulated protein 1 (hyou1) RNA binding protein fox−1 homolog 1 (rbfox1) Laminin subunit gamma−1 (lamc1) Cytochrome c oxidase subunit 2 (mt−co2) Calcium uniporter protein, mitochondrial (mcu) V−type proton ATPase subunit d 1 (atp6v0d1) Cytochrome c oxidase subunit 2 (mt−co2) Neuronal−specific septin−3 (sept3) Probable 2−oxoglutarate dehydrogenase E1 (dhtkd1) CUGBP Elav−like family member 2 (celf2)

NADH−ubiquinone oxidoreductase chain 4 (mt−nd4) Interferon−inducible dsRNA−dependent ... activator A (prkra) Serine/threonine−protein phosphatase 4 ... subunit B (ppp4cb) ATP synthase protein 8 (mt−atp8) Protein tweety homolog 3 (ttyh3) Paraspeckle component 1 (pspc1) L−lactate dehydrogenase B−A chain (ldhba) Triosephosphate isomerase B (tpi1b) Actin−related protein 2−A (actr2a) Secernin−2 (scrn2) Guanine nucleotide−binding protein subunit beta−2−like 1 (gnb2l1) Protein LZIC (lzic) Glutaredoxin 3 (glrx3)

Protein DJ−1zDJ−1 (park7) Protein phosphatase 1 regulatory subunit 12A (ppp1r12a) Carbonic anhydrase (cahz) SH3 domain−binding ... protein 2 (sh3bgrl2) RNA−binding protein 4.1 (rbm4.1) Isoaspartyl peptidase/L−asparaginase (asrgl1) Malignant T−cell−amplified sequence 1 (mcts1) Protein C10 (si:dkey−29f10.1) Actin−related protein 2−B (actr2b) Superoxide dismutase [Cu−Zn] (sod1) Dynactin subunit 2 (dctn2)

Anamorsin (ciapin1) Anamorsin (ciapin1) Neural cell adhesion molecule L1.1 (nadl1.1) Apolipoprotein Eb (apoeb) SUMO−activating enzyme subunit 2 (uba2) SUMO−conjugating enzyme UBC9−B (ube2ib) Nascent polypeptide−associated complex subunit alpha (naca) Histone−binding protein RBBP4 (rbbp4) RELATIVE ABUNDANCE RELATIVE ABUNDANCE Phospholysine phosphohistidine ... phosphatase (lhpp) Guanine nucleotide−binding protein subunit beta−1 (gnb1) -1.5 -1 -0.5 0 0.5 1 1.5 -2 -1 0 1 2 COP9 signalosome complex subunit 3 (cops3) Prostamide/prostaglandin F synthase (fam213b) Adenylosuccinate synthetase isozyme 1 (adssl1) Apolipoprotein A−I (apoa1) KH domain−containing ... protein 2 (khdrbs2) Lissencephaly−1 homolog A (pafah1b1a) 60S ribosomal protein L24 (rpl24) Phosphatidylinositide phosphatase SAC1−B (sacm1lb) Eukaryotic translation initiation factor 3 subunit M (eif3m) 3−hydroxyisobutyryl−CoA hydrolase, mitochondrial (hibch) Dynamin−1−like protein (dnm1l) V−type proton ATPase subunit C 1−B (atp6v1c1b) SLIT−ROBO Rho GTPase−activating protein 2 (srgap2) tRNA−splicing ligase RtcB homolog (rtcb) 60S ribosomal protein L10a (rpl10a) COP9 signalosome complex subunit 5 (cops5) LanC−like protein 1 (lancl1) Citrate synthase, mitochondrial (cs) 60S ribosomal protein L27 (rpl27) Mitochondrial Rho GTPase 1−A (rhot1a) Brain acid soluble protein 1 homolog (basp1) V−type proton ATPase subunit C 1−A (atp6v1c1a) NEDD8−activating enzyme E1 regulatory subunit (nae1) Neurocalcin−delta A (ncalda) RNA−binding motif protein, X chromosome (rbmx) Neuropilin−1a (nrp1a) mRNA cap guanine−N7 methyltransferase (rnmt) Selenium−binding protein 1 (selenbp1) Nck−associated protein 1 (nckap1) WD repeat−containing protein 26 (wdr26) PI−PLC X domain−containing protein 3 (plcxd3) Casein kinase I isoform gamma−1 (csnk1g1) ADP−ribosylation factor−like protein 8B−A (arl8ba) UPF0568 protein C14orf166 homolog (zgc:56576) Synembryn−A (ric8a) Sideroflexin−4 (sfxn4)

RELATIVE ABUNDANCE RELATIVE ABUNDANCE

-2 -1 0 1 2 -2 -1 0 1 2

Supplementary Figure 2. Proteins distinguishing mutant (psen1K97fs/+) zebrafish brains from wild type (psen1+/+) siblings. A. Differentially abundant proteins between mutant and wild type brains at 6 months of age. B. Differentially abundant proteins between mutant and wild type brains at 24 months of age. Proteins are considered differentially abundant if the False Discovery Rate [FDR]-adjusted p-value of their moderated t-test < 0.05. The protein abundance of the opposite age group is also shown in A and B. Colours of heatmap cells represent relative protein abundance levels, scaled across the samples in each age group. Supplementary Figure 3. Average expression (logCPM) of gene markers for various neural cells (astrocyte, microglia, neuron, oligodendrocyte) in mutant and wild type zebrafish brains. Representative marker genes for microglia were obtained from Oosterhof et al. 1 while gene markers for the remaining neural cells were obtained from Lein et al. 2. The number of genes used to calculate the average gene expression (in logCPM) was 41 (astrocyte), 99 (microglia), 77 (neuron) and 78 (oligodendrocyte). Overall, the average expression of gene markers for the major neural cell types is approximately constant in each of the biological conditions, and no obvious outlier samples are evident Supplementary Table 3. Summary of functional enrichment in different sets of genes identified from differential gene expression analysis of zebrafish brains. Up to the top three functional terms are included in this table; see Supplementary Table 3 for full analysis results. Functional terms are considered significantly enriched when the Bonferroni-adjusted p-value < 0.05. The 18,296 genes remaining after filtering in the differential gene expression analysis form the background gene set. Only genes with Entrezgene identifiers were included in the functional enrichment analysis.

Differentially expressed in ‘aged mutant vs. aged wild type’ (177 genes)

Enriched functional Description Gene Names % % Bonferroni term Genes with Background p-value term genes with term

Homeobox Antennapedia subfamily hoxb5b, hoxb3a, 6.19% 0.12% 1.18E-06 Antennapedia CS of homeobox genes hoxb5a, hoxa5a, (IPR001827, Interpro hoxc5a, hoxb8a, domains) hoxa4a

Homeobox CS Conserved homeobox hoxb3a, hoxb5b, 9.73% 0.94% 0.000359 (IPR017970, Interpro site barhl1a, hoxc5a, domains) hoxb5a, zgc:101100, hoxc1a, hoxa4a, hoxa5a, hoxb8a, phox2bb

GSE18791 CTRL VS Genes down-regulated in amotl2a, gadd45ba, 12.86% 0.93% 0.000640 NEWCASTLE comparison of control lgals3bpb, VIRUS DC 8H DN conventional dendritic LOC566587, (MSigDB) cells (cDC) at 0 h versus dhx58, phox2bb, cDCs infected with mastl, csrnp1b, Newcastle disease virus ddit4 (NDV) at 8 h.

Differentially expressed in ‘aged wild type vs. young wild type’ (1,795 genes)

Enriched functional Description Gene Names % % Bonferroni term Genes Background p-value with genes with term term

GSE19888 Genes up-regulated in psmb9a, tnfsf10, gbp2, ikzf1, 4.00% 0.88% 4.42E-12 ADENOSINE A3R HMC-1 (mast leukemia) hck, iﬁh1, irf7, psmb8a, INH PRETREAT cells: incubated with the psme1, si:ch1073-340j19.1, AND ACT BY A3R peptide ALL1 and then zgc:171731, nmi, mxe, VS TCELL treated with Cl-IB- si:ch211-197g15.8, irf9, MEMBRANES ACT MECA versus si:ch211-160o17.2, zgc:92791, MAST CELL UP stimulation by T cell eif2ak2, tnfaip2a, lyn, (MSigDB) membranes. mpeg1.1, mxc, tap1, dhx58, xaf1, si:ch211-219a4.3, samsn1b, lipg, b4galt5, gne, il2rga, tapbp.1, parp3, npc2, pik3ap1, rnf213a, gla, edem1

GSE42021 CD24HI Genes down-regulated in abcb3l1, b2m, parp3, fosl2, 4.32% 1.27% 3.91E-08 VS CD24INT TREG thymic T reg: CD24 tapbp.1, b2ml, si:dkey- THYMUS DN high versus CD24 int. 201c1.2, mhc1uka, nfkbiab, (MSigDB) sema3fa, tap1, gpr18, mxc, bcl2l13, tfg, calcoco2, rps6ka5, gbp1, erap1b, zgc:92791, mhc1zba, irf9, si:dkey- 57h18.2, mxe, lgals3bpa, si:ch211-197g15.8, ctss2.2, hist2h2l, mhc1uba, itm2ba, psme1, erap1a, zgc:171731, nmi, iﬁt12, irf7, rbms1a, psmb8a, tnfsf10, psmb9a, iﬁh1

Antigen processing and The process in which an b2m, tap1, cd74a, si:ch73- 1.38% 0.23% 7.81E-05 presentation antigen-presenting cell 158p21.3, cd74b, mhc1uba, expresses antigen si:busm1-266f07.2, si:dkey- (GO:0019882, GO (peptide or lipid) on its 57h18.2, abcb3l1, erap1a, Biological Process) cell surface in association tapbp.1, mhc1uka, erap1b, with an MHC protein ctss2.2, mhc2dab complex.

Differentially expressed in ‘aged mutant vs. young mutant’ (1,072 genes)

Enriched functional Description Gene Names % Genes % Background Bonferroni p- term with term genes with term value

Homeobox Antennapedia subfamily hoxb5a, hoxa5a, 1.43% 0.12% 2.78E-07 Antennapedia CS of homeobox genes hoxc4a, hoxb6b, (IPR001827, Interpro hoxb6a, hoxa4a, domains) hoxb8a, hoxc5a, hoxb3a, hoxc6a, hoxb5b, hoxb8b

Circadian regulation of Any process that nr1d2a, nr1d1, 1.00% 0.08% 0.00370 gene expression modulates the frequency, ciarta, per3, nﬁl3- (GO:0032922, GO rate or extent of gene 5, bhlhe40, clocka Biological Process) expression such that an expression pattern recurs with a regularity of approximately 24 hours.

GSE21546 WT VS Genes up-regulated in psme2, b2ml, 3.67% 1.05% 0.00421 SAP1A KO DP untreated double positive zgc:171731, b2m, THYMOCYTES UP thymocytes: wild type dhx58, usp18, (MSigDB) versus ELK4 knockout. eif2ak2, mhc1uka, tjp1a, mxc, psmb9a, tapbpl, psmb8a, xaf1, nrg2a, csnk1g1, csrnp1b, vezf1b, abcb3l1, mhc1uba, pfkfb3, si:dkey- 57h18.2, arsh Age-dependent inversion between mutant and wild type brains (65 genes) (opposite direction of differential expression in ‘young mutant vs. young wild type’ and ‘aged mutant vs. aged wild type’

Enriched functional Description Gene Names % Genes % Background Bonferroni p- term with term genes with term value

CHEN LVAD Genes up-regulated in pik3r1, ddit4, 21.21% 0.73% 0.000125 SUPPORT OF left ventrical gadd45ba, FAILING HEART UP myocardium of patients nfkbiab, irs2b, (MSigDB) with heart failure fkbp5, mcl1b

GSE14769 UNSTIM Genes down-regulated in LOC566587, 24.24% 1.25 % 0.000224 VS 40MIN LPS comparison of stk35, tob1a, BMDM DN (MSigDB) unstimulated mcl1b, ifrd1, macrophage cells versus nfkbiab, macrophage cells gadd45ba, csrnp1b stimulated with LPS (TLR4 agonist) for 40 min.

GSE46606 UNSTIM Genes down-regulated in LOC566587, 21.21% 1.14% 0.00299 VS CD40L IL2 IL5 at day 0 B cell IRF4-KO ifrd1, irs2b, mcl1b, 1DAY versus CD40L and IL-2 tagapb, csrnp1b, STIMULATED IL-4 IL-5 stimulated at nfkbiab IRF4HIGH SORTED day 1 B cell IRF4high. BCELL DN (MSigDB)

Accelerated aging (65 genes) (same direction of differential expression in ‘young mutant vs. young wild type’ and ‘aged wild type vs. young wild type’)

Enriched functional Description Gene Names % Genes with % Background Bonferroni term term genes with term p-value

GSE9988 LPS VS Genes up-regulated in plk3, ddit4, 21.88% 1.00% 0.000948 VEHICLE comparison of gadd45ba, nfkbiab, TREATED monocytes treated with csrnp1b, tnfaip2a, MONOCYTE UP 1 ng/ml LPS (TLR4 adrb2b (MSigDB) agonist) versus untreated monocytes.

Inappropriately downregulated (57 genes) (down-regulated in both ‘aged mutant vs. young mutant’ and ‘aged mutant vs. aged wild type’)

Enriched functional Description Gene names % % Background Bonferroni term Genes genes with term p-value with term

Homeobox Antennapedia hoxc5a, hoxa4a, hoxb3a, 13.46% 0.12% 4.25E-09 Antennapedia CS subfamily of homeobox hoxb8a, hoxb5b, hoxa5a, (IPR001827, Interpro genes hoxb5a domains)

CHEN LVAD Genes upregulated in gadd45ba, fkbp5, 27.59% 0.73% 8.55E-07 SUPPORT OF left myocardium after rnd3a, pik3r1, FAILING HEART UP heart failure foxo3b, nfkbiab, (MSigDB) ddit4, mcl1b

Regulation of cellular Any process that nr1d2a, foxo3b, sart3, 47.92% 15.45% 0.00504 metabolic process modulates the hoxa5a, hoxc1a, tnfrsf11a, (GO:0031323, GO frequency, rate or hoxb5b, tsc22d3, phox2bb, Biological Process) extent of the chemical si:ch73-380l10.2, hoxb5a, reactions and pathways hoxb8a, gadd45ba, mcl1b, by which individual nﬁl3-5, rgcc, arntl1b, cells transform pik3r1, hoxb3a, tob1a, chemical substances. hoxa4a, hoxc5a, pik3r3a Failure to upregulate (94 genes) (up-regulated in ‘aged wild type vs. young wild type’ and not up-regulated in ‘aged mutant vs. aged wild type’)

Enriched functional Description Gene Names % Genes % Background Bonferroni p- term with term genes with term value

GSE339 EX VIVO VS Genes down-regulated in tiparp, kpna1, 16.98% 1.51% 0.00344 IN CULTURE comparison of ex vivo nfkbiab, mapk6, CD8POS DC DN CD8 dendritic cells pik3r1, ulk2, (MSigDB) versus cultured CD8 gadd45ba, mxi1, DCs. Ifrd1

GSE46606 UNSTIM Genes down-regulated in nfkbiab, irs2b, 15.09% 1.14% 0.00556 VS CD40L IL2 IL5 at day 0 B cell IRF4-KO dnajb4, tagapb, 1DAY versus CD40L and IL-2 ifrd1, STIMULATED IL-4 IL-5 stimulated at LOC566587, IRF4HIGH SORTED day 1 B cell IRF4high. stk40, csrnp1b BCELL DN (MSigDB)

GSE14769 UNSTIM Genes down-regulated in nfkbiab, tiparp, 15.09% 1.25% 0.0116 VS 40MIN LPS comparison of stk35, dnajb4, BMDM DN unstimulated LOC566587, macrophage cells versus csrnp1b, macrophage cells gadd45ba, ifrd1 stimulated with LPS (TLR4 agonist) for 40 min.

Aging signature (525 genes) (same direction of differential expression in ‘aged wild type vs. young wild type’ and ‘aged mutant vs. young mutant’)

Enriched functional Description Gene Names % Genes % Background Bonferroni p- term with term genes with term value

GO ANTIGEN The process in which an b2ml, abcb3l1, 2.80% 0.14% 5.87E-05 PROCESSING AND antigen-presenting cell mhc1zba, si:dkey- PRESENTATION OF expresses a peptide 57h18.2, ENDOGENOUS antigen of endogenous mhc1uba, PEPTIDE ANTIGEN origin on its cell surface mhc1uka, b2m, (MSigDB) in association with an tapbp.1 MHC protein complex.

GSE37533 PPARG1 Genes up-regulated in eif2ak2, vezf1b, 6.99% 1.37% 8.58E-05 FOXP3 VS FOXP3 CD4 T cells treated with mhc1uka, b2m, TRANSDUCED CD4 pioglitazone and over- nfyba, lgals9l1, TCELL expressing: FOXP3 and mhc1uba, usp25, PIOGLITAZONE PPARg1 isoform of b2ml, zgc:171731, TREATED UP PPARG versus FOXP3. psmb9a, xaf1, (MSigDB) ctnnd2b, mef2cb, ctnnd2a, scamp1, cyp2p7, mxc, dhx58, psmb8a

GSE21546 WT VS Genes up-regulated in mhc1uka, b2m, 5.94% 1.05% 0.000257 SAP1A KO DP untreated double positive vezf1b, eif2ak2, THYMOCYTES UP thymocytes: wild type b2ml, abcb3l1, (MsigDB) versus ELK4 knockout. pfkfb3, mhc1uba, xaf1, zgc:171731, psmb9a, arsh, psmb8a, csnk1g1, dhx58, si:dkey- 57h18.2, mxc Supplementary Table 5. Summary of enrichment of known zebrafish promoter motifs in sets of genes identified from differential gene expression analysis. Significantly enriched motifs with Bonferroni-adjusted p-value < 0.05 are included in this table. Background genes are the 18,296 genes used in the differential gene expression analysis.

Genes Analysed Number Enriched Motif Bonferroni % Genes % Background Name p-value with motif genes with motif

Differentially expressed in ‘young mutant 105 – – – – vs. young wild type’

Down-regulated only Up-regulated only 40 – – – – 65 – – – –

Differentially expressed in ‘aged mutant 177 GRE 0.005725 17.65% 4.86% vs. aged wild type’

Down-regulated only 139 GRE 6.736e-04 18.89% 4.88% Up-regulated only 38 – – – –

Differentially expressed in ‘aged wild type 1,795 ISRE 3.633e-05 7.77% 3.92% vs. young wild type’ IRF8 1.346e-04 21.81% 15.52% IRF1 2.044e-03 12.65% 8.29% PU.1:IRF8 3.336e-03 11.06% 7.08% IRF2 5.285e-03 10.36% 6.59%

Down-regulated only 692 – – – – Up-regulated only 1,103 ISRE 7.634e-06 9.16% 3.96% IRF8 2.509e-05 24.01% 15.6% PU.1:IRF8 4.747e-04 12.64% 7.10% IRF2 1.576e-03 11.69% 6.62% IRF1 4.879e-03 13.59% 8.36% Atf3 6.019e-03 36.65% 28.69% Fra2 0.03700 25.12% 18.86% Ets-distal 0.04120 15.01% 10.12% AP-1 0.04818 36.97% 29.9%

Differentially expressed in ‘aged mutant 1,072 – – – – vs. young mutant’

Down-regulated only 378 – – – – Up-regulated only 694 – – – –

Age-dependent inversion between mutant 65 GRE 0.004735 23.91% 4.97% and wild type brains (opposite direction of change in ‘young mutant vs. young wild type’ and ‘aged mutant vs. aged wild type’)

Accelerated aging (same direction of 65 – – – – differential expression in ‘young mutant vs. young wild type’ and ‘aged wild type vs. young wild type’)

Inappropriately down-regulated 57 GRE 0.000105 28.26% 4.99% (down-regulated in both ‘aged mutant vs. young ARE 0.0344 26.09% 7.34% mutant’ and ‘aged mutant vs. aged wild type’)

Failure to up-regulate (up-regulated in ‘aged 94 – – – – wild type vs. young wild type’ and not up- regulated in ‘aged mutant vs. aged wild type’)

Failure to downregulate (down-regulated in 26 – – – – ‘aged wild type vs. young wild type’ and not down-regulated in ‘aged mutant vs. aged wild type’)

Aging signature (same direction of differential 525 – – – – expression in ‘aged wild type vs. young wild type’ and ‘aged mutant vs. young mutant’) Supplementary Methods 2. Comparison of LC-MS/MS protein abundance and RNA- seq gene expression levels in zebrafish brains

Correlation Analysis We were not able to thoroughly compare the proteins quantified using mass spectroscopy to the RNA transcripts quantified using RNA-seq due to vast differences in sensitivity (only 323 proteins were reliably quantified compared to 18,296 genes). By using matching Ensembl gene identifiers to overlap the normalised and filtered LC-MS/MS and RNA-seq data sets, we identified 219 proteins that were reliably quantified in both data sets. Overall, the correlation between the abundance level of these proteins and their corresponding gene expression counts in the RNA- seq data was weakly positive (Spearman’s rho 0.34-0.45 for LC-MS/MS and RNA-seq data for 6- month-old zebrafish brains; Spearman’s rho 0.24-0.31 for LC-MS/MS and RNA-seq data for 24- month-old zebrafish brains) (Supplementary Fig. 4).

Supplementary Figure 4. Boxplots of Spearman’s rho (rank) correlation between RNA-seq (gene expression) and LC-MS/MS (protein abundance) data for each sample. Overall, correlation of the gene expression patterns between individual RNA-seq samples (libraries) is high (median Spearman’s rho 0.89 in 6-month- old samples, 0.87 in 24-month-old samples). The correlation of protein abundance patterns between LC-MS/ MS samples is also high (median Spearman’s rho 0.99 in 6-month-old samples, 0.99 in 24-month-old samples). However, the correlation between RNA-seq and LC-MS/MS samples is moderately weak (median Spearman’s rho 0.40 in 6-month-old samples, 0.28 in 24-month-old samples). Fisher’s Method Fisher’s method is a meta-analysis technique used to combine results (raw p-values) from independent tests with the same hypothesis. Here, we used Fisher’s method to combine p- values from the differential gene expression and differential protein abundance tests. These tests test the hypothesis that a gene or protein is differentially expressed in a particular comparison (either ‘young mutant vs. young wild type’, ‘aged mutant vs. aged wild type’, ‘aged wild type vs. young wild type, or ‘aged mutant vs. young mutant’). By combining these results, it is possible to identify genes/proteins that are likely to experience significant changes in both protein abundance and in gene expression level in particular comparisons. Below, we summarise proteins that may be of interest because they show similar changes at both the protein and RNA- seq level when comparing mutant and wild type zebrafish brains.

‘Young mutant vs. young wild type’ Raw Protein p- Raw RNA p- logFC logFC UniProtKB Gene Fisher’s Method value value (protein) (RNA) Name FDR p-value 0.0522 0.00117 -0.823 -0.886 Q6DHI5 sirt5 0. 0797

• For the ‘young mutant vs. young wild type’ comparison, no proteins were signiﬁcantly altered

(absolute log2-fold-change > 0.5 in both the protein and RNA-seq data) at both the protein and

RNA level after FDR adjustment of p-values. The only protein that showed similar log2-fold-

change at both the protein and RNA level was sirtuin-5 (sirt5) (RNA-seq log2-fold-change

0.886, LC-MS/MS log2-fold-change 0.823; Fisher’s method FDR-adjusted p-value 0.0797), although the result does not reach statistical signiﬁcance.

‘Aged mutant vs. aged wild type’ Raw Protein p- Raw RNA p- logFC logFC UniProtKB Gene Fisher’s Method value value (protein) (RNA) Name FDR p-value 0.141 4.03E-06 -0.503 -0.501 B3DJT0 sart3 0.00189 0.00213 0.0461 0.738 1.57 O42363 apoa1a 0.0272

• For the ‘aged mutant vs. aged wild type’ comparison, two proteins were signiﬁcantly altered at both the protein and RNA level after FDR adjustment, squamous cell carcinoma antigen

recognised by T-cells (sart3; RNA-seq log2-fold-change -0.501, LC-MS/MS log2-fold-change - 0.503, Fisher’s method FDR-adjusted p-value 0.00189) and apolipoprotein-A (apoa1a; RNA-

seq log2-fold-change 1.57, LC-MS/MS log2-fold-change 0.738, Fisher’s method FDR-adjusted p-value 0.0272). ‘Aged wild type vs. young wild type’ Raw Protein p- Raw RNA p- logFC logFC UniProtKB Gene Fisher’s Method value value (protein) (RNA) Name FDR p-value 5.76E-10 6.49E-04 -3.55 0.605 Q6PI20 h3f3d 2.39E-09 2.12E-08 1.36E-04 2.80 0.845 Q9MIY1 mt-nd4 5.84E-09 6.05E-06 4.87E-07 1.85 1.01 Q7T3C6 nudt21 5.84E-09 2.76E-04 2.91E-06 -1.67 0.980 Q6DGK2 ldhbb 4.23E-07 1.45E-06 8.32E-04 -2.01 0.532 Q5PRD4 csnk1g1 4.69E-07 3.46E-04 1.68E-05 0.793 0.525 Q6PH57 gnb1a 1.41E-06 7.15E-05 7.44E-04 -1.81 0.828 Q9MIY5 mt-atp6 8.86E-06 6.26E-02 4.77E-05 -1.52 1.21 Q9MIY9 mt-nd2 1.93E-04 2.10E-04 8.50E-02 2.80 0.536 Q9MIY6 mt-atp8 7.20E-04 4.22E-02 1.88E-03 0.68 0.734 Q803W1 scrn3 2.29E-03 9.68E-02 3.17E-03 -0.63 0.649 Q9MIY0 mt-nd5 6.35E-03 5.33E-03 6.26E-01 -2.62 -0.942 Q5XJ10 gapdh 3.66E-02

• For the ‘aged wild type vs. young wild type’ comparison, 12 proteins are signiﬁcantly altered at both the protein and RNA level after FDR adjustment. Many of these proteins are encoded by genes that are expressed in mitochondria (e.g. mt-nd4, mt-atp6, mt-nd2, mt-atp8, mt-nd5).

The correlation between protein and gene log2-fold-change is weaker; many proteins which are show lowered abundance with age are increased at the RNA transcript level.

‘Aged mutant vs. young mutant’ Raw Protein Raw RNA p- logFC logFC UniProtKB Gene Fisher’s Method FDR p-value value (protein) (RNA) Name p-value 0.000301 0.00117 -1.75 0.886 Q6DHI5 sirt5 6.71982E-05 0.00737 0.140 0.861 -0.640 Q90487 hbaa1 0.020333525

• For the ‘aged mutant vs. young mutant’ comparison, two proteins are signiﬁcantly altered at the protein and RNA level after FDR adjustment. These proteins do not overlap with those in

the ‘aged wild type vs. young wild type’ comparison. The log2-fold-changes in protein abundance and gene expression are in opposite directions.

Co-inertia analysis Co-inertia analysis is a multivariate statistical method that can be used to integrate transcriptomic and proteomic data to identify relationships between these datasets3. First, a multivariate analysis transformation (in this case, non-symmetric CA, NSC) was separately applied to the proteomic and transcriptomic data sets. Co-inertia analysis then uses the NSCs to identify common trends or relationships in the data sets by selecting axes that maximise the square covariance3. The first two CIA dimensions are plotted in Supplementary Fig. 5. Here, the base of the arrow represents the gene expression, the tip of the arrow represents the protein abundance, and the length of the arrow is proportional to the overall similarity between the gene and protein expression/abundance levels. The CIA in Supplementary Fig. 5 indicates that the four biological conditions (young wild type, young mutant, aged wild type, aged mutant) possess sufficiently different gene expression and protein abundance patterns to separate them from other biological conditions. However, the lengths of the arrows suggest moderate differences between gene expression and protein abundance patterns for each biological condition.

In CIA, the global correlation between data sets can be summarised by the RV coefficient which takes values from 0 (no correlation) to 1 (perfect correlation). The RV coefficient of the RNA-seq and LC-MS/MS data sets in this case is 0.428, indicating moderate global similarity for the proteins and genes that can be matched. Supplementary Figure 5. Co-inertia analysis (CIA) of matched protein-gene pairs from the LC- MS/MS (protein abundance) and RNA-seq (gene expression) data sets. In the top panel, the base of the arrows represent the overall gene expression for each biological group (young wild type, young mutant, aged wild type, aged mutant), the tip of the arrow represents the overall protein abundance for each biological condition, and the length of the arrow is proportional to the overall similarity between the gene and protein expression/abundance levels. A shorter arrow indicates that the gene expression and protein abundance data are more correlated to each other. The bottom two panels plot the matched genes (left) and proteins (right) used in the CIA. LASTROCYTE MARKERS

J CELLULAR RESPONSE TO UV KIFN-γ RESPONSE GOLIGODENDROCYTE MARKERS

SYNAPTIC TRANSMISSION ANEURONAL MARKERS

CD209 SIGNALING CmRNA SPLICING

DCELL CYCLE DEVELOPMENTAL TRANSCRIPTION F FACTORS RESPONSE TO OXIDISED PHOSPHOLIPIDS ERESPONSE TO HEAT PROTEIN FOLDING

IMMUNE RESPONSE I MICROGLIA MARKERS

CORRELATION BETWEEN MODULE EXPRESSION AND TRAIT (p-value)

−0.58 0.25 0.7 0.62 0.42 0.68 0.51 0.36 0.6 0.59 0.59 0.73 (0.006) (0.3) (4e−04) (0.003) (0.06) (7e−04) (0.02) (0.1) (0.004) (0.004) (0.005) (2e−04) Early-onset AD

−0.32 0.49 0.23 0.096 −0.18 −0.035 0.48 0.39 0.48 0.31 0.31 0.3 (0.2) (0.02) (0.3) (0.7) (0.4) (0.9) (0.03) (0.08) (0.03) (0.2) (0.2) (0.2) PSEN1 mutation

A B C D E F G H I J K L MODULE

Supplementary Figure 6. Co-expression network of genes expressed in human brain tissue (posterior cingulate region). Each node represents a gene, with node size proportional to the number of co-expressed genes. An edge between two nodes indicates those genes are co-expressed. 13 different groups of co-expressed genes (modules) are coloured. Labels indicate the top functional enrichment term(s) associated with each module. Co- expression network was constructed using gene expresion data from Antonell et al. The hybrid-robust correlation between module expression and sample traits (presence of early-onset AD, presence of PSEN1 mutation) is shown underneath the network. Supplementary Methods 3. Histological analysis of brains from two-year-old mutant and wild type zebrafish siblings

Tissue preparation Two-year-old adult zebrafish heterozygous for the psen1K97fs mutation and their wild type siblings were sacrificed by immersion in ice-water, then tails were nicked to exsanguinate the fish and prevent blood clotting on neural tissue. The dorsal neurocranium was subsequently resected to expose the brain. Fish were then decapitated and heads were incubated in a decalcification solution (100 ml 0.5M EDTA, 22 g sucrose, 11 ml 10x phosphate buffered saline solution, PBS) for four hours on a slow shaker at room temperature. Decalcified heads were then fixed overnight in 4% paraformaldehyde in phosphate buffer (PFA in PB), at 4°C on a slow shaker. Heads were then embedded in a sucrose-gelatin medium (20% sucrose, 8% cold-water fish gelatin in 1x PBS), frozen on dry ice and cryosectioned at 16 µm thickness on a Leica CM3050-S cryostat. Serial sections were mounted on SuperFrost Plus microscope slides (Menzel-Gläser). Sections were subsequently dried at room temperature for four hours, and then stored at -20°C until staining. Prior to all stains, sections were retrieved from -20°C and brought to RT, then rehydrated in 1x PBS.

Senescence-associated β-galactosidase (saβgal) staining Sections were prefixed with 4% PFA in PB for one hour in a humid chamber at room temperature, then washed twice for 15 minutes with 0.3% Triton X-100 in 1x PBS (PBS-Tx (0.3%)). The pH of sections was then equilibrated with two 30 minute washes with 1x PBS at pH 5.5 (all washes were performed in a humid chamber). Sections were then stained for 16 hours at 37°C in a humid chamber in staining solution (2 mM MgCl2, 5 mM K3Fe(CN)6, 5 mM K3Fe(CN)6·3H2O, 1 mg/ml X-gal, with the remaining volume made up of 1x PBS at pH 5.5). Following incubation, staining was arrested by a 20-minute wash in 4% PFA in PB at room temperature. Sections were then washed well in PBS at pH 5.5 and mounted in 50% glycerol. Sections were then imaged on an Olympus Provis AX70 widefield microscope with an Olympus DP70 camera, with images acquired at 4080x3072 resolution.

Congo Red staining for amyloid Following rehydration in 1x PBS, sections were stained for 20 minutes in Congo Red staining solution (0.5% Congo Red in 50% ethanol). Sections were then rinsed in distilled water, and quickly differentiated by dipping five times in an alkaline alcohol solution (1% NaOH in 50% ethanol). Sections were rinsed for 1 minute in distilled water, then mounted in 50% glycerol. Sections were imaged in both brightfield and birefringence on a Leica Abrio polarising microscope at 1024x1024 resolution. (Note that Congo Red staining was also performed on sections from 6- month-old mutant brains but no signal was detected – data not shown.)

Autofluorescent detection of lipofuscin Following rehydration in 1x PBS, sections were mounted in 50% glycerol and imaged confocally on a Leica TCS SP8 invert confocal laser scanning microscope with a Leica HyD hybrid detector using 20x and 63x oil-immersion objectives. Lipofuscin has an emission maximum at 590 nm when excited at 488 nm; samples were thus excited with a 488nm laser at power 5.00, and the detector was gated for emission wavelengths between 560-700 nm. Images were acquired at 1024x1024 resolution with gain at 137.8%. Supplementary Figure 7. saβgal staining in two-year-old psen1K97fs/+ and +/+ zebrafish brain. A. Cross section of telencephalon. Black arrow show very weak saβgal staining in the dorsal telencephalon. Red arrow shows weak saβgal staining in neurons in nuclei lateral to the tenecephalic ventricle. B. Cross section of the mesencephalon/diencephalon showing hypothalamus and the optic tectum. Black arrow shows weak saβgal staining in superficial and deep layer of the optic tectum. Staining is primarily in layers containing fibers. C. Cross section of the cerebellum. Black arrow shows weak saβgal staining in neurons in molecular and Purkinje cell layer. Red arrow shows strong saβgal staining in neurons and fibres ventral to the rhombencephalic ventricle. Scale bar for overviews in A-C = 200 μm and 40 μm for the high magnification inset areas. Supplementary Figure 8. Congo red staining in two-year-old psen1K97fs/+ and +/+ zebrafish brain. A. Cross section of telencephalon. Black arrow shows very weak congo red staining in the dorsal telencephalon (bright field and birefringence). Red arrow shows weak congo red staining in neurons in nuclei lateral to the tenecephalic ventricle. B. Cross section of the mesencephalon/diencephalon showing the optic tectum. Black arrow shows congo red staining in a deep layer of the optic tectum (bright field and birefringence). Staining is primarily in layers containing fibers. C. Cross section of the cerebellum. Black arrow shows very weak or no congo red staining in the cerebellum. Red arrow shows strong congo red staining in fibres of the median forebrain bundle ventral to the rhombencephalic ventricle. Scale bar for overviews in A-C = 100 μm. Supplementary Figure 9. Measurement of fluorescence (Lipofuscin) in two-year-old psen1K97fs/+ and +/+ zebrafish brain. Excitation was done using 488 nm laser line and fluorescence was detected between 560 and 700 nm. A. Cross section of telencephalon. Weak fluorescence was detected in the dorsal telencephalon. Boxed area shown in higher magnification. B. Cross section of the mesencephalon/diencephalon showing the optic tectum. Weak fluorescence is detected in all layers of the optic tectum. C. Cross section of the cerebellum. Weak fluorescence is detected in all layers of the cerebellum. Scale bar in A-C = 100 μm. Supplementary Data References

1. Oosterhof, N. et al. Identification of a conserved and acute neurodegeneration-specific microglial transcriptome in the zebrafish. Glia 65, 138-149, doi:10.1002/glia.23083 (2017).

2. Lein, E. S. et al. Genome-wide atlas of gene expression in the adult mouse brain. Nature 445, 168-176, doi:10.1038/nature05453 (2007).

3. Fagan, A., Culhane, A. C. & Higgins, D. G. A multivariate analysis approach to the integration of proteomic and gene expression data. Proteomics 7, 2162-2171, doi:10.1002/pmic.200600898 (2007).