NGG nucleotide sequence (or CCN) must be present immediately 5` of the sequence you want to target with Cas9. Else you cant target the sequence and must choose another.
If there is no such sequence you can use alternative nuclease like Cpf1 (TTN/NAA)
I recommend you have a look on the Addgene website for information about some Cas9 variants in the literature which have slightly different PAM sequence requirements: https://www.addgene.org/crispr/guide/ (scroll down to the bottom of the page).
Here is a summary:
SpCas9 = NGG
SpCas9 D1135E variant = NGG (reduced NAG binding)
SpCas9 VRER variant = NGCG
SpCas9 EQR variant = NGAG
SpCas9 VQR variant = NGAN or NGNG
SaCas9 = NNGRRT or NNGRR(N)
If any of these PAM sequences would be ideal for your targeted region, you can then search on the Addgene website for plasmids encoding the desired Cas9 enzyme. You can order the plasmids from them (they're not that expensive) and then start cloning for your sgRNAs.
PAM localisation is some quite flexible, no need to be on exactly locus of interesting. For knock-in, within +/- 100 base pairs region works fine, by providing a correct repair donor template. For knock-out would be even more flexible. Hard to imagine you could not find a NGG site within 200 base pairs of sequence.